The AI Decision Platform for F&B
Open any solution and watch it run: the till, the label printer, the scale, the guest's phone. 149 live solutions across quick service, casual dining, café, cloud kitchen and six more sectors.
10 sectors, one AI — configured per format.
Kitchen Display & Station Routing
Every order broken into station work and sequenced by promise time — grill, fry, assembly and drinks working to one clock instead of four.
- Order Split to Stations One ticket becomes station work: patty to grill, fries to the fry bank, drink to the beverage station. Nobody reads the whole order and decides for themselves what to start.
- Sequenced by Promise, Not Arrival A drive-through order promised in 90 seconds outranks a delivery order promised in 12 minutes, even if the delivery order came first. The screen enforces the sequence.
- Cook-Time Synchronisation Items start so they finish together. Fries do not sit under the lamp for four minutes waiting on a grill item that had not been started.
- Bump, Recall & Remake A bumped item is timestamped; a recall brings it back with the reason attached. Remake rates by station and by hour become visible rather than argued about.
- Station Load Balancing When the fry bank is the constraint, Iris says so during the rush — not in a report on Monday when the queue has already gone somewhere else.
“What is our average order-to-hand-off time at peak, and which station causes the delay?”
Self-Service Kiosk
Full menu browse, allergen and calorie display, upsell that actually converts, and payment — with the kiosk writing into the same order book as the counter.
- Browse and Build The full menu, browsable without a queue behind you. Guests explore more on a kiosk than at a counter, which is most of why the basket is bigger.
- Allergen and Calorie Display Every item carries its allergens and calories from the recipe, not from a laminated sheet updated last year. It is a legal requirement in most markets and a trust signal in all of them.
- Upsell That Converts Iris offers the add-on with the highest modelled acceptance for that basket and that day-part — not the most expensive item, which is what a fixed rule would push.
- Payment and Order Number Card, wallet or cash-to-counter, then a number the guest watches on the board. The order is already at the stations before they have pocketed their phone.
- Accessibility and Language Screen height, contrast, text size and language are switchable per guest. A kiosk that only serves the confident is a kiosk that loses covers.
“What is the average kiosk basket against the counter basket, and how much of the gap is upsell?”
Drive-Through Management
Lane depth, order-to-window time and the second window measured to the second — with the sequence protected when a large order would otherwise block the lane.
- Lane Timing to the Second Arrival, order, payment, hand-off. Each segment timed separately, because a slow lane is almost never slow in the place people assume.
- Order Ahead in the Lane At peak, orders are taken further back so the window becomes hand-off rather than conversation. The lane keeps moving even when the kitchen is at capacity.
- Large-Order Pull-Forward A twenty-item order is routed to a waiting bay rather than holding six cars behind it. Iris flags the order at the point of capture, not at the window.
- Second Window Discipline Payment and hand-off separated properly, with the time between them measured. This is where most drive-through seconds are quietly lost.
- Peak Forecast to the Half-Hour Tomorrow's lane volume is forecast in thirty-minute blocks, so staffing covers the rush instead of arriving with it.
“What is our order-to-window time by hour, and where in the lane do we actually lose seconds?”
Recipe-Based Inventory Depletion
Every sale removes ingredients in recipe proportion — patties, buns, oil, packaging — so stock is a consequence of trading rather than a Sunday-night count.
- Bill of Material per Item A burger is not one line of stock. It is a patty, a bun, sauce, lettuce, packaging and a portion of oil — each depleting at its own rate.
- Depletes as You Sell Selling removes ingredients immediately. A stock figure that is four days old is a story; this one is a fact.
- Yield and Portion Reality Cooked yields, trim and portion tolerance are modelled, so the cost is per sellable item rather than per theoretical one.
- Theoretical vs Actual The gap between what recipes say and what actually left the store is where food cost goes. It is reported daily, not discovered at stocktake.
- Auto-Reorder at Par When projected stock breaches par before the next delivery, a draft order is raised with quantity and timing already worked out.
“How much beef did we actually use last week against what the recipes say we should have?”
Daypart Menu Auto-Switch
Breakfast, lunch, late-night and weekend menus that change themselves — on the till, the kiosk, the app and every aggregator, at the same second.
- Daypart Calendar Each menu has a start and end time per branch, because breakfast in a station outlet ends at a different hour than in a retail park.
- Switches Everywhere at Once Counter, kiosk, app, web and every aggregator flip together. The failure mode this prevents is an aggregator still selling breakfast at 12:30.
- Grace Period on In-Flight Orders An order started before the switch completes on the old menu. Guests are not told mid-transaction that the thing they chose no longer exists.
- Prep Aligned to the Switch The kitchen is told what to run down and what to start ahead, so the changeover does not create either a shortage or a bin full of breakfast.
- Measured at the Boundary Revenue in the fifteen minutes either side of each switch is reported, which is how you learn whether your breakfast should end at 11:00 or 11:30.
“How much revenue do we lose in the fifteen minutes around each daypart change?”
Real-Time Out of Stock — ‘86’ an Item Everywhere at Once
An item goes 86 once and disappears from every channel in the same second — with substitutes offered and the kitchen told before the guest is disappointed.
- Called Once, Applied Everywhere 86 an item on the KDS and it leaves the till, the kiosk, the app, the web store and all five aggregators together. This single behaviour prevents most refund-and-apologise cycles.
- Predicted Before It Happens Iris projects which items will run out before close at the current rate, so the call is made deliberately rather than in the middle of a rush.
- Substitute Offered, Not Silence The guest is offered the nearest equivalent with the reason given, rather than discovering at the window that half their order is missing.
- Auto-Restore on Replenish When the item comes back, it returns everywhere automatically. Forgotten 86s that stay off the menu for a week are a real and invisible revenue loss.
- 86 Log and Root Cause Every 86 is logged with cause — under-prepped, under-ordered, supplier short, equipment down — because the pattern is what you actually need to fix.
“How many orders did we refuse last month because an item was 86 on one channel but still selling on another?”
Delivery Aggregator Sync
DoorDash, Uber Eats, Grubhub, Deliveroo and Just Eat on one menu and one live stock, with orders landing in the same queue as the counter.
- One Menu, Every App Publish once and it lands on all five, with the aggregator price uplift applied per platform rather than maintained by hand in five portals.
- Orders in the Same Queue Aggregator orders sequence alongside counter and drive-through by promise time. No tablet farm, no re-keying, no order lost behind a screen nobody was watching.
- Availability Held Honest An 86 removes the item from every app in the same second, which is the single biggest driver of aggregator refunds and rating damage.
- Rider Timing Respected Food is made when a rider is genuinely close. Bags sitting on a shelf are why delivery ratings fall even when the kitchen did everything right.
- True Margin by Platform Commission, packaging and refund rate are attributed per aggregator, so you can see which platform is growth and which is an expensive habit.
“Which aggregator has the worst true margin once commission and packaging are counted?”
Combo & Modifier Engine
Combos, upsizes, swaps and no-onions handled at speed — priced correctly, costed correctly, and printed clearly enough that the line does not remake it.
- Combo Build and Pricing A combo is a priced bundle with its own cost and margin, not a discount typed in at the till. Upsizes and swaps reprice automatically.
- Modifier Speed at the Counter No pickles, extra sauce, sub the fries — two taps, not a conversation. At a hundred orders an hour, seconds per order is the whole argument.
- Cost and Margin per Variant A combo with a swapped side has a different cost. The platform knows it, so the reported margin is real rather than nominal.
- Printed So the Line Reads It Modifiers print in a fixed position and order, in the same place every time. Most remakes are a reading error, not a cooking error.
- Combo Mix Intelligence Iris reports which combos staff actually push against which ones make money, and where the two disagree.
“Which combo has the best margin, and which one are staff defaulting to instead?”
Hourly Sales Heatmap & Staffing AI
Sales and labour matched by the half-hour, so the rush is covered and the lull is not paid for — with the roster built from the forecast rather than last week's copy.
- Heatmap by Half-Hour Sales plotted against labour hours in thirty-minute blocks. The mismatch is almost always visible immediately and almost never where managers expect.
- Roster Built from Forecast Shift shape comes from predicted trade, so the counter is covered at 12:30 and not overstaffed at 15:00.
- Live Labour Ratio Labour against sales is visible during the shift, not discovered at month-end when nothing can be done.
- Break Placement That Works Breaks are scheduled into the troughs automatically, which is a small thing that quietly protects the peak.
- Overtime Before It Happens Projected overtime and working-time breaches are flagged while the roster is still a draft rather than after it is worked.
“What was labour as a percentage of sales by half-hour last Saturday?”
Speed of Service Analytics
Every second between arrival and hand-off measured, segmented and attributed — because in quick service, speed is not a metric, it is the product.
- Segment the Journey Arrival to order, order to start, start to ready, ready to hand-off. A single average tells you nothing; the four segments tell you everything.
- Compare Like With Like A drive-through at noon is not a counter at 15:00. Speed targets are set per channel and per day-part rather than one number for the whole estate.
- The Cost of a Second Iris converts seconds into covers and covers into revenue, so a fifteen-second improvement has a number attached and a case for the investment.
- Outlier Detection One branch at 210 seconds while its cluster runs at 140 is a specific problem with a specific cause — usually layout or one under-staffed station.
- Shift-Level Attribution Speed is tracked to shift and team, which makes coaching a conversation about a number rather than an impression.
“Where exactly are we losing seconds at Friday lunch, and what is it costing in lost covers?”
Loyalty, App & Offer Engine
App orders, points, targeted offers and the returning-guest economics that decide whether quick service is a transaction or a habit.
- Identify Without Friction A scan, a phone number or the app. Anything that adds ten seconds at a QSR counter will not be used, so identification has to be effectively free.
- Offers That Move Behaviour Iris targets the offer at the guest whose behaviour it can actually change, rather than discounting people who were coming anyway.
- Frequency Over Basket In quick service the win is one more visit a month, not two dollars more per visit. The engine optimises for visit frequency.
- App Order Ahead Ordered on the app, collected at a shelf or the lane, timed against the guest's real arrival rather than a fixed estimate.
- Lapse Detection A twice-weekly guest who has not been seen in three weeks is worth an intervention now, while the habit is still recoverable.
“What share of sales comes from identified guests, and how much more do they spend?”
Food Safety & Temperature Compliance
Fry oil, hot holding, chillers and cook temperatures logged automatically to a standard that survives an inspection — and flags a failing unit before it spoils stock.
- Automatic Temperature Logging Chillers, freezers, hot holds and fry vats report themselves. A log that depends on someone remembering is a log an inspector will not trust.
- Fry Oil Quality Oil is tracked by hours, cook count and polar compound reading, so it is changed on condition rather than on habit — which is both a quality and a cost decision.
- Cook Temperature Verification Probe readings tie to the batch and the operator, giving a defensible record for the products where it matters most.
- Breach Response, Automatic A breach quarantines the affected stock, notifies the food safety lead and opens an incident — rather than being noticed at the next manual check.
- Inspection-Ready Export Two years of records exportable in the format an inspector asks for, which turns an unannounced visit into an administrative event rather than a crisis.
“Which branch has the most temperature exceptions, and which unit is causing them?”
Multi-Site Rollout & Standards
Menus, prices, promotions and procedures pushed to a hundred branches at once — with local variation where it is justified and nowhere else.
- Push Once, Land Everywhere A price change, a new item or a promotion goes to every branch and every channel from one place, with confirmation that it actually landed.
- Justified Local Variation Airport pricing, a franchise exception, a market-specific item — permitted by rule and visible, rather than a spreadsheet nobody can find.
- Drift Detection Branches running off-standard prices or menus are flagged with how long they have been drifting. Most drift is accidental and none of it is harmless.
- New-Site Opening Pack A new branch inherits menu, recipes, prices, rules and standards on day one rather than being configured by hand and getting it 90% right.
- Standards Scored, Not Assumed Compliance with the operating standard is measured per branch and reported to the area manager who can actually act on it.
“Which branches are running off-standard menus or prices, and for how long?”
Table Management & Floor Plan
A live digital floor plan showing every table's status — occupied, free, reserved, or being cleared. The host always knows where to seat the next guest without checking physically.
- Live Floor Map Drag-and-drop floor plan matches your actual layout — tables update in real time as orders are placed and bills paid. Table turn predictor estimates when each occupied table will free up — host sees expected availability 15 minutes ahead.
- Table Status Colour Coding Free (green), occupied (amber), reserved (blue), cleaning (grey) — visible at a glance from any device.
- Cover Count per Table
- Multi-Room / Multi-Floor Indoor, outdoor terrace, private dining room — all on one screen, switchable by tab.
- Table Merge & Split Merge adjacent tables for large parties; split a table's bill for groups paying separately.
“What is the average table turn time on Friday dinner service?”
Server Assignment & Multi-Course Coursing
Tables assigned to specific servers. Orders coursed to the kitchen in sequence — starter, main, dessert — kitchen receives each course only when the previous one is cleared.
- Server-to-Table Assignment Each table assigned to a named server — sales, tips, and performance tracked per server. Server performance scorer tracks AOV, upsell rate, and tip % per server — identifies coaching opportunities without bias.
- Course Sequencing Starter fires at order time. Main fires when staff 'course' it from the POS — kitchen never gets ahead.
- Hold & Fire Control
- KDS Course Awareness Kitchen display shows course number and sequence — no confusion between tables' starters and mains.
- Course Timing Log Time between courses tracked per table — identifies where kitchen or service is causing delays.
“Which server had the highest average order value this week?”
Split Bill — By Item or Equally
A table's bill can be split any way the guests choose — equally per head, by individual items ordered, or a custom combination. No mental arithmetic for staff.
- Equal Split Total divided by number of guests — each person pays their share via card or cash. No ML model — pure transactional logic. AOV and tip analysis run post-payment on split data.
- Item-by-Item Split Each guest selects their own items from the bill — pays exactly what they ordered.
- Custom Split
- Multi-Tender per Split Each split can be paid differently — Guest A pays cash, Guest B taps card.
- Service Charge Allocation Service charge distributed proportionally across splits — no manual calculation.
“What percentage of our tables use split billing on weekends?”
Reservation & Waitlist Management
Phone and walk-in reservations captured in one system — with deposit tracking, no-show flags, and a live waitlist for busy periods. The host is always in control.
- Advance Booking Reservations taken by phone or online — date, time, covers, special requests, and contact recorded. No-show predictor scores each reservation 0–100 for likelihood of arriving — allows intelligent overbooking buffer.
- Deposit Collection Deposit required for large parties or peak nights — linked to the reservation, deducted at billing.
- Walk-In Waitlist
- No-Show Logging Reservations that don't arrive are logged — feeds the no-show predictor for future bookings.
- Cover Forecast Expected cover count by hour — helps kitchen plan prep quantities in advance.
“Which day of the week has the highest reservation no-show rate?”
Open Tab Management
A tab opened when guests are seated stays open until they ask for the bill. Staff can add items at any time without re-opening an order.
- Tab Opening at Seating Tab created automatically when a table is assigned — no separate step required. High-spend tab alert — AI flags tabs approaching a configurable threshold for manager awareness.
- Add Items at Any Time Additional orders added to the running tab — no new order needed, no duplication.
- Running Total Visible
- Tab Transfer Tab can be moved to another table or transferred to a bar tab if guests move.
- Partial Payment Guests can pay part of the tab early — remaining balance stays open.
“What is the average tab value on Saturday night vs Sunday lunch?”
Tip Handling & Service Charge
Tip prompts on the card terminal at checkout — with staff tip allocation, pooling options, and full reporting. Service charge applied per your policy and distributed correctly.
- Card Terminal Tip Prompt Configurable tip % options displayed on the payment terminal at checkout — 10%, 15%, 20%, or custom. Server tip rate tracked alongside AOV and cover count — scoring model surfaces performance outliers.
- Cash Tip Recording Cash tips logged against the server — included in their performance report.
- Tip Pooling Options
- Service Charge Automation Fixed % service charge added to bills automatically — exempt categories can be excluded.
- Tip & Service Charge Report Total tips per server, per shift, per week — full audit trail for payroll integration.
“What was the average tip percentage by server this month?”
Dietary Preference Profiling AI
Over multiple visits, the system builds a preference profile for returning guests — dietary restrictions, favourite dishes, allergen flags — enabling personalised service without asking every time.
- Allergy & Dietary Logging Guest's dietary requirements logged at first visit — vegan, nut allergy, gluten-free — attached to profile. Customer memory model builds preference vectors per guest. Recommendation model scores current menu items against profile.
- Favourite Dish Memory Items ordered repeatedly are flagged — staff can suggest 'the usual' without being prompted.
- Allergen Conflict Alert
- Preference-Based Suggestions AI suggests dishes that match the guest's known preferences from the current menu.
- Special Occasion Flags Birthday, anniversary notes attached to booking — front of house prompted at seating.
“How many of our regular guests have a dietary restriction logged?”
KDS — Course-Aware Kitchen Routing
The kitchen display knows which course each item belongs to — starters, mains, and desserts displayed separately, fired in sequence, and colour-coded by table urgency.
- Course Separation on Display Starters and mains shown in separate sections — kitchen never fires both at once accidentally. Kitchen load predictor models order volume 20 minutes ahead — alerts head chef to incoming surge.
- Table Urgency Colour Coding Tables approaching wait target turn amber, then red — kitchen can prioritise without checking a clock.
- Station Routing
- Bump & Confirm Chef bumps an item when complete — front of house notified to collect immediately.
- Course Timing Analytics Average time from order to starter, starter to main, main to dessert — per day, per service.
“What was the average time from starter to main on Saturday evening?”
Table Turn & Pacing Intelligence
Turn time measured from seat to settle, broken into seat, order, fire, serve and pay — with Iris pacing courses so a table finishes on time without ever feeling rushed.
- Turn Broken Into Stages Seat to order, order to fire, fire to serve, serve to settle. A 92-minute turn is not one problem; it is usually one stage of it, and until you split them you are guessing.
- Paced, Not Rushed Iris nudges the fire time for the next course based on how the table is actually eating. The aim is a table that never waits and never feels hurried — those are different failures with the same cause.
- The Settle Gap The longest unmanaged stage in casual dining is finishing dessert to paying. Payment offered at the table the moment plates clear routinely recovers eight to twelve minutes per turn.
- Turn Target by Daypart A weekday lunch and a Saturday dinner are different businesses. Targets are set per daypart and per table size rather than one estimate applied to everything.
- What a Turn Is Worth On a full Saturday, six minutes off the average turn is roughly one extra sitting on the smaller tables. Iris quantifies it so the pacing conversation has a number attached.
“Which table sizes turn slowest on a Saturday, and where in the meal is the time actually lost?”
Waitlist, Reservations & No-Show Prediction
Bookings, walk-ins and the waitlist on one floor plan — with no-show risk scored per booking and overbooking calculated rather than guessed.
- One Book, Every Channel Phone, web, Google, the app and walk-ins all write into the same floor plan. Double-booking a table across two channels is not a mistake staff should have to prevent manually.
- No-Show Risk Scored Lead time, party size, channel, prior history and whether a card was held all predict no-show. A 22:00 booking for eight made three weeks ago through an aggregator is not the same risk as a regular booking on Tuesday.
- Overbooking With Arithmetic If Saturday 20:00 carries a modelled 14% no-show, a calculated overbook recovers covers without creating a queue of angry booked guests. Guesswork does the opposite.
- Waitlist That Tells the Truth Quoted wait times come from live turn data rather than optimism. A guest told 40 minutes who waits 35 is happy; one told 20 who waits 35 is not — same wait, opposite outcome.
- Confirmation That Reduces Loss Automatic confirmation and a one-tap release recover tables that would otherwise sit empty. Most no-shows are people who could not be bothered to cancel, not people who meant to waste your table.
“What is our no-show rate by booking channel, and how much cover did it cost us last month?”
Menu Engineering & Course Mix
Every dish graded on contribution and velocity, with course mix, attachment and the true cost of the free bread tracked — so the menu earns rather than merely pleases.
- Stars, Puzzles, Plough-Horses, Dogs Each dish graded on contribution margin against velocity. The chef's favourite that sells 14 a week at 42% margin is a conversation the platform can start with evidence rather than opinion.
- Course Mix Is the Real Margin Starter and dessert attachment moves profit far more than main-course price. A table that takes three courses is worth roughly 60% more than one that takes a main and a soft drink.
- Attachment Prompted, Not Nagged Iris tells the server which starter this table is most likely to take, based on their order so far and comparable tables — one suggestion, not a script.
- The Free Things Counted Bread, olives, sauces and refills are real cost that most casual restaurants never attribute to the dish that triggered them. Counted properly, some “profitable” mains are not.
- Menu Position Tested Placement, description and photography change order rates measurably. Changes are tested against a matched control rather than rolled out on instinct.
“Which mains carry the highest margin but the lowest order rate, and what sits next to them on the menu?”
Server Performance & Section Balance
Sales per server hour, attachment rate, course accuracy and section load — measured fairly, with sections balanced so the same person is not carrying the room every Saturday.
- Measured Fairly Sales per server hour normalised for section, daypart and table size. Judging a window-section server against a bar-section server on raw revenue is measuring the floor plan, not the person.
- Attachment as the Real Skill Starter, side, wine and dessert attachment is what separates strong service from adequate service, and it is coachable in a way that charm is not.
- Sections Balanced Deliberately Iris proposes section allocation so covers and revenue potential are shared. The strongest server quietly carrying the hardest section every weekend is how good people leave.
- Course Accuracy Tracked Wrong course, wrong table, missing modifier — recorded against the shift rather than the individual first, because most course errors are a pass problem before they are a server problem.
- Coaching, Not Ranking The output is a specific behaviour to work on, not a league table on the wall. Ranking servers publicly reliably damages the teamwork the floor depends on.
“Which servers have the highest attachment rate, and are our sections balanced by covers or by revenue?”
Loyalty, CRM & Guest Recognition
Guest profiles built from every visit — usual table, allergies, anniversaries, wine preference — surfaced to the server before the guest sits down.
- Recognised, Not Interrogated The server sees the usual table, the allergy, the wine they liked and the fact it is an anniversary before they reach the table. Recognition is the cheapest luxury a restaurant can offer.
- Allergies Travel With the Guest A recorded allergy follows the profile and blocks the dish at the point of order rather than depending on one server remembering one conversation.
- Occasions Captured Birthdays and anniversaries booked once become a prompt every year. A dessert with a candle for a returning couple costs very little and is remembered for a long time.
- Lapsed Regulars Flagged A weekly couple who have not appeared in six weeks is a recoverable relationship if you notice. Iris flags it while it is still recoverable, with what they usually ordered.
- Privacy Handled Properly Guests can see and delete what is held. Notes are professional and factual — preferences and allergies, not personal remarks about the guest.
“Which regulars have not visited in 60 days, and what did they usually order?”
Recipe Costing & Portion Control
Dish specs with real plate cost, portion weights verified at the pass, and margin recomputed live as ingredient prices move.
- A Plate Cost That Is Real Protein, garnish, sauce, side and the trimmings that never make the recipe card. Most casual-dining plate costs are understated by 8–12% for exactly this reason.
- Portions Verified at the Pass Scale-verified proteins on high-value dishes. Steak portioned by eye drifts upward under pressure — not through carelessness, but because a generous plate is the path of least resistance.
- Yield Recorded Honestly Trim, cook loss and waste captured as real yields, so cost is per plate served rather than per kilo purchased.
- Recosted as Prices Move When beef moves, every dish built on it recosts in the same second and any crossing its margin floor is flagged with the gap — before the quarter closes.
- Variance Traced to Service Portion variance attributed to shift and section turns an abstract food-cost percentage into a specific coaching conversation about a specific station.
“Which dishes fell below margin floor this quarter, and was it price or portion?”
Omni-Channel — Dine-In, Takeaway & Delivery
Dine-in, collection, delivery aggregators and table QR on one order book — with kitchen capacity protected so delivery never quietly destroys the dining room.
- One Kitchen, Several Demands Dine-in courses, collection orders and three aggregators all draw on the same pass. Sequencing them by promise time rather than arrival is the difference between a controlled service and a shouting match.
- Delivery Capped by Capacity When the dining room is at capacity, delivery slots close automatically rather than accepting orders the kitchen cannot honour. Protecting the room the guest is sitting in is the right default.
- A Different Menu Where It Makes Sense Dishes that travel badly are hidden from delivery. Publishing a dish that arrives cold is a reputational cost that no commission rate justifies.
- True Margin by Channel Aggregator commission, packaging and the kitchen time consumed are attributed per channel. Delivery frequently carries the highest revenue and the lowest contribution.
- Table QR Without Losing Service Guests can order and pay from the table while servers keep working the room. Used well it removes the settle gap; used badly it removes the service — so it is configurable per section.
“How many dine-in tables waited longer because delivery orders were firing at the same time?”
Waste, Comps & Void Control
Every void, comp, remake and wastage recorded with reason and authoriser — so generosity is a decision rather than a leak.
- Reason and Authoriser, Always A void with no reason is the single most common cover for both error and theft. Requiring both takes four seconds and changes behaviour immediately.
- Comps That Recover a Guest Comping a dish for a genuine mistake is good business. The platform separates recovery comps from habitual ones so the first is protected and the second is questioned.
- Remakes Traced to Cause Wrong order, wrong cook, sent back, dropped. Remake reasons point at the pass, the section or the spec — three very different fixes.
- Prep Waste Against Forecast Over-prep is the quiet cost of a variable service. Prep quantities driven by the covers forecast rather than by yesterday reduce it without risking a sell-out.
- Patterns Surfaced, Not Accusations Iris raises the pattern — a section, a shift, a dish — and leaves the judgement to a manager who knows the people. That is the only responsible way to use this data.
“Which section generated the most comps last month, and what were the stated reasons?”
Integration Hub — Payments, Delivery & Tax
Card, wallet and split payments, the delivery aggregators, reservation platforms, local tax rules and the accounting ledger — connected, health-checked and configured per market.
- Payments — Including the Awkward Ones Stripe, Adyen, Square and PayPal, plus Apple Pay and Google Pay. Split by item, split equally, split across four cards and one person paying cash — all settling into one reconciliation.
- Reservations — Platform Agnostic OpenTable, TheFork, Google Reserve and direct bookings all write into the same floor plan, so the book is never a screen someone forgot to check.
- Delivery Apps — Capacity Aware Uber Eats, Deliveroo, Just Eat and DoorDash through Deliverect, with a menu that respects what travels and slots that respect what the kitchen can do.
- Tax — Eat-In Versus Take-Away Rates differ by market and by whether the guest sits down. The till has to know before the guest answers, and the rule is configured per market rather than coded per customer.
- Connection Health — Watched, Not Assumed Every connection health-checked with last-sync times and error rates, and Iris quantifies the covers at risk when a reservation or delivery API degrades.
“Which integration failed most often last month, and how many covers did it cost us?”
Tasting Menu Coursing & Pacing
Multi-course tasting menus require precise pacing — each course fired to the kitchen only when the previous one has been cleared, with timing logged and enforced by the POS.
- Course-by-Course Fire Control Staff fires each course manually from the POS — kitchen never receives the next course until the table is ready. Course timing regression model benchmarks each course against historical averages — flags deviations in real time.
- Tasting Menu Templates Pre-set course sequences — 5-course, 7-course, 10-course loaded as templates.
- Per-Course Timing
- Paired Beverage Sequencing Wine pairings linked to each course — sommelier prompted to serve before the food arrives.
- Simultaneous Service When a course is ready for all guests, kitchen notifies all stations to plate at once.
“What is the average duration of a 7-course tasting menu from first course to bill?”
Wine & Beverage Pairing Module
A curated pairing guide built into the POS — sommelier or server can suggest the right wine for every dish, log the recommendation, and track what was accepted.
- Dish-to-Wine Pairing Library Each menu item has recommended pairings pre-loaded — visible to staff on the POS item card. Recommendation model scores pairing acceptance by guest profile and dish — surfaces the suggestion most likely to be accepted.
- Vintage & Bin Number Tracking Wine inventory tracked by vintage, bin, and remaining bottles — staff know availability before recommending.
- Pairing Acceptance Rate
- Sommelier Commission Tracking Beverage upsells attributed to the recommending staff member — drives incentive programmes.
- By-the-Glass Depletion Open bottle tracked by glass — system knows when a bottle runs out without a physical count.
“Which wine pairing had the highest acceptance rate with the lamb main this season?”
Reservation & Deposit Management
Fine dining reservations carry deposit requirements, special requests, and occasion notes. The system manages the full lifecycle from booking to arrival to post-dining follow-up.
- Online & Phone Booking Reservations accepted via website widget or phone — all land in the same system with full guest details. No-show predictor scores reservation risk 48 hours ahead — allows proactive confirmation call for high-risk bookings.
- Deposit Collection & Escrow Deposit charged at booking — held until the reservation is honoured, then deducted from the final bill.
- Cancellation Policy Enforcement
- Occasion & Preference Notes Birthday, anniversary, dietary requirement, preferred table — all captured and surfaced at arrival.
- No-Show Deposit Retention If a guest no-shows, deposit is retained automatically with a record of the non-arrival.
“What is our no-show rate for weekend dinner reservations over the last 3 months?”
Clienteling — Guest Preference History
Every visit, preference, and note recorded against a guest profile. Return guests are recognised, preferences surfaced to the host, and their experience personalised before they sit down.
- Visit History Log Dates, dishes ordered, spend, server assigned — full history per guest, searchable by name or phone. Churn classifier identifies guests who haven't returned within their normal frequency — triggers a win-back campaign.
- Preference & Allergy Profile Dietary restrictions, seating preferences, disliked ingredients — recorded once, available always.
- Occasion Tracking
- Recognition at Arrival When a returning guest checks in, their profile is surfaced — host can greet by name with context.
- Post-Visit Feedback Satisfaction rating logged after each visit — negative experiences flagged for manager follow-up.
“Which returning guests haven't visited in over 60 days?”
High-AOV Upsell AI
An AI recommendation engine calibrated for high-average-order-value fine dining — suggesting premium dishes, supplements, and experiences at the optimal moment.
- Supplement Suggestions When a main is ordered, AI suggests truffle supplement, wagyu upgrade, or premium preparation. Collaborative filtering model trained on historical upsell acceptance by dish, guest profile, and table spend trajectory.
- Occasion-Based Upsell Birthday table? AI suggests a dessert cake or champagne toast — prompted to the server.
- Wine by the Bottle vs Glass
- Experience Add-Ons Chef's table experience, kitchen tour, sommelier table — upsell prompts configured for relevant tables.
- Upsell Conversion Tracking Every suggestion logged with accepted/declined outcome — model retrains on actual conversion data.
“What is our supplement attach rate on the wagyu main this month?”
Per-Cover Revenue Tracker
Revenue per cover is the key metric in fine dining. This module tracks it live — by service, by day, by server — and benchmarks it against historical averages and targets.
- Real-Time Revenue per Cover Live metric updated with every order placed — current service RevPAC visible on manager dashboard. Driver attribution model decomposes RevPAC variance — how much came from price, mix, volume, or supplement attach rate.
- Server-Level Breakdown RevPAC by server — identifies upselling stars and underperformers without subjectivity.
- Daypart Comparison
- Target vs Actual RevPAC target set per service — dashboard shows live gap with time remaining in service.
- Menu Item Contribution Which dishes drive the highest RevPAC — informs menu engineering decisions.
“What drove the increase in revenue per cover last Saturday vs the Saturday before?”
Chef's Table Pre-Order Manager
Chef's table and tasting menu experiences require advance pre-orders for dietary restrictions, allergies, and preferences. This module captures, organises, and surfaces all pre-orders before service.
- Pre-Order at Booking Guests submit dietary requirements and preferences at reservation time — before they arrive. No ML model — pure data management and surfacing. Chef satisfaction and repeat booking rate tracked as outcome KPIs.
- Pre-Order Summary for Kitchen Chef receives a single pre-order sheet per service — all allergies and special requests in one view.
- Allergen Conflict Check
- Menu Substitution Planning Pre-order details linked to specific courses — chef can plan substitutions in advance.
- Guest Brief for Front of House Host and server receive a one-page guest brief before arrival — name, occasion, preferences, notes.
“How many guests had dietary requirements pre-submitted for this weekend's chef's table?”
Table Turn Predictor AI
AI predicts when each occupied table will become available — so the host can manage the waitlist and seat walk-in guests with precision, not guesswork.
- Current Dwell Time Display How long each table has been occupied — shown live on the floor plan. Regression model trained on historical dwell time by party size, day, service, and course progress — predicts table availability ±8 minutes.
- Predicted Free Time AI estimates when each table will pay and leave — based on course stage, historical dwell, and pacing.
- Waitlist ETA Communication
- Overstay Alert Tables significantly beyond predicted dwell time flagged — gentle intervention prompted if needed.
- Turnover Efficiency Score Average gap between one party leaving and next being seated — tracks hosting team efficiency.
“What is our average gap between a table clearing and the next guests being seated?”
Fast Bar Workflow & Barista Label Printing
Orders from the counter flow directly to the bar as a printed label — cup size, milk type, temperature, syrup, and customer name. Barista never has to call back to the till.
- Instant Label Print at Order The moment a drink is ordered, a label prints at the bar — barista begins immediately, no verbal relay. Queue wait estimator predicts current wait time — displayed at entry so customers decide whether to join.
- Full Modifier on Label Size, milk, shots, syrups, temperature, and special notes all printed — no ambiguity.
- Customer Name on Label
- Queue Display Screen Digital board shows orders in queue — customers see their position and estimated wait.
- Bar Throughput Tracking Drinks per hour by barista tracked — identifies bottlenecks in bar workflow.
“What is the average time from order to drink collection on Monday mornings?”
Modifier Engine — Size, Milk, Syrups, Temperature
Full drink configurability — every combination of size, milk type, syrup, shots, temperature, and special instruction is captured accurately and printed on the barista label.
- Size Variants Small, medium, large — with automatic price adjustment per size for every drink. Most common modifier combinations surface as one-tap presets — reducing order time by 30% for frequent drinks.
- Milk Type Selection Full fat, semi, oat, almond, soy, coconut — each with individual pricing if applicable.
- Syrup & Flavour Add-Ons
- Shot Strength & Temperature Single, double, triple — or decaf. Iced vs hot. Temperature preferences recorded.
- Free-Text Special Instructions For unusual requests — typed at till and printed verbatim on the barista label.
“What is the most popular milk type ordered with flat whites this month?”
Weight-Based Selling — Per Gram / Kg
Items priced by weight are sold accurately without pre-packaging — customer chooses quantity, scale integrates with POS, price is calculated and added to the bill instantly.
- Scale Integration Connected scale sends weight directly to POS — no manual entry, no rounding errors. Daily production recommender uses weight-based sales velocity to forecast how much to produce per item tomorrow.
- Price per Gram / Kg Configuration Each weight-priced item has its own price per unit weight — multiple items, multiple rates.
- Partial Purchase Logging
- Tare Weight Handling Container or packaging weight automatically deducted — customer pays only for the product.
- Weight Audit Trail Every weight transaction logged — supports HACCP and portion control audits.
“What was the average daily weight sold of sourdough this week?”
Daily Production Limits — Auto-Cut on Sell-Out
Each product has a daily production quantity. When it sells out, it is automatically removed from the menu on all channels — customers cannot order what hasn't been made.
- Production Quantity per Item Morning entry of how many units were produced — counter, kiosk, and app all start with that inventory. Production recommender predicts tomorrow's optimal quantity per item based on day-of-week, weather, and seasonal trends.
- Auto-Removal at Zero When the last unit is sold, the item is automatically removed from the menu across all channels.
- Low Stock Alert
- Daily Production Log What was made, what sold, what was wasted — tracked daily for each item.
- Carry-Over Prevention Items set to non-carry-over are zeroed at close of day — no yesterday's stock misrepresented as fresh.
“Which products sell out before noon most frequently?”
Display Case Inventory — Depletes on Sale
Items on display are tracked as inventory units. Each sale automatically depletes the display case count — staff always know what remains without counting physically.
- Unit Tracking per Display Item Each pastry, slice, or display item tracked as a unit — sold units deducted in real time. Sell-through rate forecaster predicts which display items will have excess stock by close — triggers markdown prompt 2 hours before closing.
- Display Replenishment Alerts When a display item drops below a threshold, back-of-house is alerted to replenish from production.
- Multi-Display Section Tracking
- End-of-Day Remaining Count System shows expected remaining units at close — cross-checked against physical count for waste recording.
- Sell-Through Rate per Item What % of each item produced was sold each day — identifies over or under-production patterns.
“Which display items have the lowest sell-through rate on Tuesdays?”
Personalised Order Suggestions AI
Returning customers are recognised at the counter or app. Based on their order history, the POS surfaces their most likely order before they have to say anything.
- Returning Customer Recognition Customer identified by phone number, loyalty card, or app profile — order history retrieved instantly. Customer memory model builds preference vectors per individual. Recommendation engine surfaces top-3 likely orders at recognition.
- 'Your Usual?' Prompt If a customer orders the same drink 3+ times, it becomes their 'usual' — one-tap reorder available.
- Time-of-Day Preference
- Seasonal Menu Adaptation When a seasonal item replaces a favourite, system suggests the closest equivalent.
- New Item Suggestion Customer hasn't tried the new croissant — suggestion offered based on past pastry purchases.
“What is the most commonly reordered drink by loyalty members?”
End-of-Day Markdown AI
Two hours before closing, AI identifies which items have excess stock and recommends a markdown price — maximising recovery and minimising waste without manual judgement.
- Excess Stock Detection System identifies items with more units remaining than historically sold in the final 2 hours. Dynamic pricing model trained on historical sell-through rates and markdown outcomes. Optimal discount predicted per item per day.
- Markdown Recommendation AI recommends the optimal discount % — deep enough to clear stock, not deeper than necessary.
- One-Click Markdown Apply
- Waste vs Markdown Comparison Running total of revenue recovered through markdown vs cost of unsold wastage — ROI tracked daily.
- Markdown Performance Log Did the markdown clear the stock? Tracked per item — model improves over time.
“How much revenue did we recover through end-of-day markdowns last month?”
Loyalty & Rewards Engine
Points, tiers, stamp cards and streak rewards on one wallet — earned and redeemed identically at the counter, on the web store and in the app. Iris decides which reward actually moves a guest.
- One Wallet, Every Channel A guest earns on a counter flat white and redeems on an app cake order — the balance is a single ledger, never per-channel silos that confuse staff and guest alike.
- Tiers, Stamps & Streaks Stamp cards for the daily coffee habit, tiers for annual value, streaks for consecutive-week visits. Each mechanic is a rule you switch on, not a rebuild.
- Iris Picks the Reward Rather than blanket 10% off, Iris selects the offer with the highest modelled redemption for that specific guest — a free pastry for the coffee-only regular, a bulk discount for the office buyer.
- Breakage & Liability Tracked Outstanding points are a liability. The engine values it live and posts it to finance, so loyalty never becomes an unbudgeted surprise.
“Which loyalty tier generates the highest repeat spend per member this quarter?”
RFM Customer Intelligence
Every guest scored on Recency, Frequency and Monetary value, then auto-segmented — champions, loyal regulars, slipping, at-risk and lost. Iris tells you who is quietly leaving before they are gone.
- Three Scores, One Picture Recency, frequency and monetary value are computed per guest and combined into a single RFM cell — the fastest honest read of who matters and who is fading.
- Automatic Segment Movement A guest sliding from weekly to monthly moves from Champion to Slipping without anyone running a report. The segment is the alert.
- Win-Back Before Loss Iris flags the slipping cohort while the habit is still recoverable and proposes the specific offer and channel most likely to bring each one back.
- Value Concentration Warning When too much revenue concentrates in too few guests, the platform says so — a quiet risk most cafés only discover after a churn event.
“Which regulars have not visited in 21 days but used to come weekly?”
Voucher & Promo Redemption
Digital vouchers, gift cards and campaign codes issued to the guest’s phone and redeemed by QR scan, NFC tap or code entry — validated against every rule in under a second, at the counter or online.
- Issued Straight to the Phone A voucher is created once and delivered to wherever the guest already is — the app wallet, an email, or a WhatsApp message. It carries a QR code, an NFC token and a short human-readable code, so it can be redeemed by whichever method is quickest at the counter.
- Scan to Redeem — QR or NFC The barista scans the QR from the guest’s screen or the guest taps their phone on the terminal. If the phone is dead or the screen cracked, the six-character code can be typed. Three routes to the same redemption, because the counter is not the place to debug a phone.
- Validated Before It Applies In the moment between scan and confirmation the platform checks expiry, minimum spend, eligible items, per-guest limit, stacking rules and whether the campaign still has budget. The guest sees accepted or declined with a reason, not a shrug.
- Applied and Settled Correctly The discount lands on the bill as its own line, attributed to its campaign. Gift-card value is drawn down from the liability rather than treated as revenue, so the books stay honest and the campaign carries its own cost.
- Budget Cap and Fraud Guard Each campaign has a spend ceiling; when it is reached redemption stops rather than quietly eroding margin. Repeat redemption from one device, shared codes and staff-assisted abuse are flagged by anomaly detection the day the pattern appears.
“How much promo value was redeemed this month, through which channel, and what did it return in incremental sales?”
Omni-Channel Ordering
Every order source on one book — walk-in counter, self-service kiosk, drive-through, web store, mobile app, call centre, WhatsApp and the delivery aggregators. Same menu, same live stock, same ledger, one promise time.
- Eight Sources, One Order Book Walk-in, kiosk, drive-through, web, app, call centre, WhatsApp and aggregators all write into the same queue. The kitchen sees one sequenced list rather than four screens and a phone, and no channel can be quietly forgotten at the rush.
- One Menu, One Live Stock A sell-out at the counter removes the item from web, app, kiosk and aggregator in the same second. Availability is a single fact, not eight copies of it drifting apart until someone gets an apologetic phone call.
- Channel-Aware Promise Times Each source is quoted a different collection or delivery time, because they deserve different ones. The app accounts for the live counter queue; drive-through is quoted on lane depth; a rider is quoted on prep plus travel — not one optimistic fixed number for everybody.
- Fulfilment Routing by Type Eat-in goes to the pass, take-away to the bagging station, drive-through to the lane window, delivery to the rider shelf with a seal, and advance cake orders to the production plan. The order knows where it is going before anyone touches it.
- Advance, Catering and Occasion Orders Celebration cakes, catering trays and event orders are captured with lead time, deposit, decoration notes and a firm slot — then enter the bake plan rather than living on a notepad by the till.
- One Ledger, One Channel Report Every channel settles into the same ledger with its own commission, packaging and delivery cost attached, so you can see true margin by channel rather than a blended number that hides which one is losing money.
“What share of orders came through each channel last week, and which channel has the highest average basket?”
Social Selling Agent
An Iris-powered agent that sells wherever the guest already is — WhatsApp, Instagram, Facebook and TikTok. It answers, recommends, quotes a real price against live stock, takes payment and tracks collection, all inside the conversation.
- Sells in Every Conversation One agent across WhatsApp, Instagram, Facebook and TikTok. Most café discovery now happens in a feed or a thread, so the agent replies there rather than pushing the guest to a website they may never open.
- Reorder My Usual A returning guest types a few words and the agent reconstructs their exact regular order — size, milk, sugar, pastry — and confirms it in one tap. New guests get today’s bake pushed to them each morning.
- Menu-Aware and Price-Aware It quotes only what is genuinely available, at today’s price, with today’s lead time, because it reads the same live catalogue and stock as the counter. A sell-out removes the item from the conversation instantly.
- Closes Inside the Thread Item, quantity, collection slot, payment link, order number and a ready-for-collection ping all happen without leaving the chat — and land in the same order book as every other channel.
- Escalates Honestly Complaints, allergy questions and custom cake briefs are handed to a person with the full thread attached. The agent never bluffs its way through something it should not answer.
“How many orders came from social and messaging conversations this month, and what was the average value?”
Customer Feedback & NPS
Feedback captured right after collection, scored for sentiment, tracked as NPS by outlet and day-part — with automatic recovery flows when a guest has a poor experience.
- One Question, High Response A single tap rating sent on the guest's own channel earns far more responses than a long survey emailed the next day.
- Sentiment on Free Text Written comments are scored and themed automatically — queue, temperature, staff, freshness, price — so patterns surface without anyone reading every line.
- Recovery Before Review A low score triggers an immediate apology and make-good offer, usually before the guest reaches a public review site.
- Tied to the Shift Every score is linked to outlet, day-part and shift, so a service problem is traced to a specific window rather than argued about in the abstract.
“Which outlet had the sharpest NPS drop this month and what did guests complain about?”
Recipe & BOM Intelligence
Recipe cards, nested sub-recipes and bills of material with real yields, batch scaling, live cost per sellable unit and allergens that propagate automatically into the printed label.
- Recipe Card and Method Every product has one card: ingredients with quantities and units, the method in steps, oven and proofing parameters, and the equipment it needs. It is the same card the baker follows and the platform costs from — not two documents that quietly diverge.
- Sub-Recipes That Nest Laminated dough is its own recipe consumed by six products. Change the dough once and every product built on it recosts, re-labels and re-plans automatically — rather than six separate edits and one you forget.
- Yield, Loss and True Cost Proofing loss, baking loss and trim are recorded as real yields, so cost is expressed per sellable unit rather than per theoretical unit. A 96% yield on croissant is the difference between the cost you think you have and the one you actually have.
- Live Recosting as Prices Move When butter moves 6%, every laminated product recosts in the same second and any that crosses its margin floor is flagged with the size of the gap — before the month-end report tells you it already happened.
- Allergen Propagation and Label Allergens travel up through every sub-recipe automatically, so the printed label and the allergy answer at the counter come from the same source as the cost. Publishing is blocked if any ingredient is untagged.
“Which products lost the most gross margin as ingredient prices moved this quarter?”
Live Ingredient Ledger & Auto-Depletion
Raw-material stock that depletes itself against every sale through the bill of material — with par levels, batch and expiry tracking, and low-stock alerts before a bake is at risk.
- Depletes Through the Recipe Selling a croissant removes flour, butter, yeast and salt in recipe proportion. Stock is a consequence of trading, not a separate weekly chore.
- Batch and Expiry Tracked Each delivery carries a batch code and use-by date, rotated first-expiry-first-out, with expiry alerts before product is written off.
- Par Levels That Learn Reorder points are set from forecast demand and supplier lead time rather than habit, and adjust as the season shifts.
- Variance Surfaced Early When actual consumption drifts from theoretical, the anomaly is raised while the cause is still findable — not at month-end stocktake.
“How much butter did we actually consume last week versus what the recipes say we should have?”
Waste & Production Variance
Theoretical versus actual usage per recipe and per batch — surfacing shrinkage, over-production, portion drift and spoilage with a cost attached to each.
- Theoretical Versus Actual What the recipes say you should have used, against what actually left the store. The gap is the number that quietly decides food-cost percentage.
- Waste With a Cause Unsold, spoiled, damaged, staff meal and trial bake are separated — because the fix for over-production is nothing like the fix for spoilage.
- Portion Drift Detected Slow, invisible over-portioning is caught by variance patterns long before it shows up as a margin problem in the P&L.
- Feeds Tomorrow's Plan Waste by item and day-part flows back into production planning, so a repeatedly over-baked product is corrected rather than repeated.
“Which products generated the most waste cost last month and at which outlet?”
Integration Hub — Payments, Delivery & Tax
Every system a café actually depends on, connected and watched: card and wallet payments, the delivery aggregators, local tax rules, and the accounting ledger. Configured per market rather than coded per customer.
- Payments — Every Method, One Settlement Stripe, Adyen, Square and PayPal as processors; Apple Pay, Google Pay, cards, and local wallets as methods. The guest picks whatever they carry; you get one settlement file and one reconciliation, not four.
- Delivery Apps — One Menu, Every Aggregator DoorDash, Uber Eats, Grubhub, Deliveroo and Just Eat connected through Deliverect. One menu published everywhere, one live stock, and their orders land in the same order book as the counter — no tablet farm behind the till.
- Tax & Compliance — Local Rules, Configured US sales tax through Avalara, GST/HST for Canada, VAT for the EU and UK, and eat-in versus take-away rates where they differ. Rules are configured per market, not coded per customer — which is why a new country takes days rather than a release.
- Accounting — Sales Post Themselves Sales, tax, tips, refunds and settlement post to QuickBooks, Xero or the group ERP on their own schedule, already coded to the right accounts. Month-end becomes a review rather than a re-keying exercise.
- Connection Health — Watched, Not Assumed Every connection is health-checked continuously with last-sync times, retry counts and error rates. When an aggregator API degrades, Iris raises it and quantifies the orders at risk rather than letting a branch discover it at the rush.
“Which integration failed most often last month, and how many orders did it cost us?”
Weight & Scoop-Based Billing
Sold by scoop, by weight, or by container size — the scale is the till, tare is deducted automatically, and stock depletes by grams rather than by guess.
- Scale & Container Setup The bench scale pairs once and is remembered. Every container — small cup, large cup, 500 ml tub, cone — carries its own tare weight, so packaging is never sold as product.
- Price by Gram, Scoop or Size Set the rate once and it applies everywhere: per 100 g for weighed, a fixed price per scoop, or a size-based price for tubs. Changing it repricing every branch in the same second.
- Serving the Cup The server builds the cup, the scale reads net weight, and the price appears before the guest has finished choosing toppings. No mental arithmetic at the counter in a queue.
- Over-Scoop Detection Average grams per scoop is tracked by server and by flavour. A server running 12% heavy is not disciplined — they are shown the number, because most over-scooping is a habit nobody measured.
- Weight Audit Trail Every weighed sale is written with the container, the tare, the net weight and the operator. A margin question three weeks later is answered from the record rather than from memory.
“What is the average weight per cup sold, and which server is consistently over-scooping?”
Topping & Mix-In Modifier Engine
Toppings, sauces, mix-ins and inclusions priced independently, with dietary swaps and allergen flags carried through to the label and the ticket.
- Topping Groups & Limits Free toppings, premium toppings, and how many of each are included before a charge applies. The rules live in one place rather than in the head of whoever trained the last server.
- Sauce, Drizzle & Inclusion Pricing Each add-on has its own price and its own stock line, so the cost of a sauce is visible instead of hidden inside the cup price.
- Dietary Swaps Vegan bases, sugar-free sauces and nut-free alternatives are offered as swaps rather than exceptions — with the allergen matrix updating on the ticket automatically.
- Attach Rate Learning Iris learns which combinations actually sell together by day-part, then suggests the one with the highest modelled acceptance rather than the most expensive.
- Visual Order Summary The built cup is shown back to the guest before payment — every scoop, topping and swap listed — which is where most remakes get prevented.
“Which topping combinations have the highest attach rate on weekends?”
Display Freezer Inventory
Every tub tracked in its well by batch, fill level and best-before, depleting as you serve — with hardening times and rotation enforced rather than remembered.
- Well Map & Tub Assignment The display is modelled as it physically is: sixteen wells, each with an assigned flavour, batch code and fill level. What the screen shows is what the server is looking at.
- Depletes as You Serve Each weighed scoop removes grams from that specific tub. Fill level falls in real time, so a tub about to run out is visible before a guest is told no.
- Batch, Best-Before & Hardening Tubs carry batch codes and best-before dates, and newly filled tubs are held for their hardening window before they can be served — enforced, not trusted.
- Rotation Enforced First-expiry-first-out is applied at the well, so the older tub is served down before the fresh one is opened, rather than the other way round.
- Run-Out Forecast Iris projects which flavours will run out before close at the current rate, and how much of each to bring from the back — before the queue makes the decision for you.
“Which flavours are below a third of a tub and will not last the evening?”
Gifting, Packaging & Take-Home
Gift boxes, take-home tubs, cakes and hampers sold alongside the counter — with dry-ice handling, travel-time guidance and packaging costed into the margin.
- Take-Home Formats Pints, litres, gift boxes, cake formats and hampers each carry their own packaging, price and stock line — they are products, not an afterthought at the till.
- Dry Ice & Travel Time The guest is asked how far they are travelling, and the right amount of dry ice is added and charged. Ice cream that arrives soft is a refund and a bad review.
- Packaging Costed In Boxes, sleeves, spoons and dry ice are consumed from stock and costed into the margin, so a gift box that looks profitable actually is.
- Occasion Prompting Iris offers the take-home format when the basket and the moment suggest it — a large mixed order on a Sunday afternoon, not a single cone on a weekday commute.
- Pre-Order & Collection Cakes and hampers are booked with lead time, deposit and a firm collection slot, entering the production plan rather than a notebook by the till.
“What share of revenue comes from take-home versus eat-in, and which format has the better margin?”
Seasonal & Limited-Time Menu
Seasonal flavours, limited runs and collaborations launched with a start date, an end date and a planned quantity — then measured honestly against the core range.
- Season & Wave Calendar Flavours are planned in waves with live-from and ends-on dates. The menu changes itself on the day rather than depending on someone remembering.
- Launch Quantity from Analogue A new flavour has no history, so Iris forecasts it from the closest analogue — and caps confidence honestly rather than pretending it knows.
- Cannibalisation Measured A limited flavour that sells 400 tubs while taking 300 from pistachio has added 100, not 400. The platform reports the second number.
- Sell-Out or Extend Mid-run, Iris says whether to extend, reorder or let it end — with the margin and the displacement effect both on the table.
- Retire Cleanly At the end date the flavour leaves the till, the app and the aggregators together, and its remaining stock is routed to markdown rather than stranded.
“How did last summer’s limited flavour perform against the core range it displaced?”
End-of-Day Markdown AI
Soft-serve mix, prepared desserts and short-life items marked down on remaining life rather than the clock — the shallowest discount that still clears the case.
- Shelf-Life Watchlist Prepared desserts, waffle batter and opened mix enter a watchlist by remaining life, not by a fixed hour — because Tuesday and Saturday do not run down at the same rate.
- The Curve, Not the Cliff A first markdown early beats a deep one late. Iris models how many units clear at each depth and proposes the shallowest one that empties the case.
- Applied Across Channels The markdown lands at the counter, in the app and on the aggregators together, so the guest is never quoted two prices for the same tub.
- Redistribute Before Waste What will not sell at any depth is offered to the redistribution partner before close, rather than weighed into the bin at midnight.
- Recovery Reported Honestly Markdown revenue is reported against the margin given away and the waste avoided, so the net is visible rather than the flattering half.
“How much did markdown recover last month, and how much would we have written off without it?”
Customer Favourite Recall AI
The guest’s usual build — base, scoops, toppings, container — recognised at the counter, in the app and in chat, and offered in one tap.
- Building the Usual After three consistent visits a guest has a usual. It is inferred from what they actually reorder, not from what they once told a form.
- Recognised at the Counter A loyalty scan brings the usual up on the till before the guest has finished saying hello — with one tap to accept and one to change.
- Offered in the Right Moment Iris offers the usual when it fits and stays quiet when it does not. A guest with three children in tow is not ordering their usual single scoop.
- Seasonal Substitution When a usual flavour is out or retired, the closest match is offered with the reason stated, rather than a silent swap the guest discovers on the first spoonful.
- Privacy by Default The guest can see what is remembered and clear it. A preference held without consent is a liability, not an asset.
“Which guests have a strong usual order, and how much more do they spend than one-off visitors?”
Waste Minimiser AI
Batch production sized to forecast, mix usage tracked against theoretical, and the flavours that quietly cost you money identified before the month closes.
- Batch Sized to Forecast How much mix to run tomorrow, by flavour, from the demand forecast — weather, weekend and school calendar included, because ice cream is more weather-sensitive than almost anything else you sell.
- Theoretical vs Actual Mix What the recipes say you should have used against what actually left the store. The gap is where margin goes without ever appearing as an obvious loss.
- Waste With a Cause Over-production, spoilage, machine purge, trial batch and staff sampling are separated — because the fix for each is completely different.
- The Quiet Losers Some flavours never waste much but never sell either, occupying a well that a stronger flavour could use. Iris names them with the opportunity cost attached.
- Feeds Tomorrow’s Batch Yesterday’s waste by flavour and day-part flows back into tomorrow’s production plan, so a repeatedly over-run flavour is corrected rather than repeated.
“Which flavour produced the most waste last month, and was it over-production or spoilage?”
Queue & Peak-Hour Flow
Saturday at eight is a different business from Tuesday at three. Order capture, staffing and the serving line all re-shape around the queue rather than fighting it.
- Queue Measured, Not Guessed Wait time is derived from real order and collection timestamps rather than a manager's impression. You cannot fix a queue you are estimating.
- Where It Actually Stalls The stall is rarely scooping. It is usually toppings, or payment, or one server handling both. The platform shows which station is the constraint tonight.
- Line-Busting Capture At peak, orders are taken down the queue on a handheld and land at the serving station ahead of the guest, so the counter becomes assembly rather than conversation.
- Peak Menu Simplification Iris can narrow the build options during the busiest forty minutes — the same trade, fewer decisions per guest, measurably shorter queue.
- Staffing to the Curve Tomorrow's peak is forecast to the half-hour, so the roster covers the rush and does not carry the lull. This is where most dessert labour cost is won or lost.
“What is our average serve time during Saturday peak, and where does the queue actually stall?”
Machine & Batch Freezer Monitoring
Soft-serve machines, batch freezers and blast chillers monitored live — mix levels, temperatures, pasteurisation cycles and the faults that spoil a night's trade.
- Live Machine Status Every soft-serve barrel, batch freezer and blast chiller reports its state: mix level, temperature, cycle stage and hours since last clean.
- Mix Low Before It Runs Dry A barrel running low is flagged with enough warning to refill between guests rather than mid-cone, which is the difference between service and apology.
- Pasteurisation & Clean Cycles Heat-treatment and clean cycles are logged automatically to a standard an inspector will accept, rather than recorded on a sheet after the fact.
- Fault Before Failure A barrel drawing more power or holding temperature poorly is flagged as degrading. Three such signals in a fortnight is a failing compressor, not bad luck.
- Downtime Costed Every fault carries an estimated cost in lost serves at that hour, which is what turns a maintenance request into a decision someone will approve.
“Which machine has failed most often this quarter, and what did the downtime cost us?”
Loyalty, Streaks & Family Wallets
Stamp cards, streaks and a shared family wallet — built for a business where one adult pays for four people and the child chooses the flavour.
- Stamps, Points and Streaks Three mechanics for three habits: stamps for the regular cone, points for the bigger spend, streaks for consecutive weekends. Each is a switch, not a rebuild.
- The Family Wallet One wallet, several members. The parent pays, the children earn, and the reward can be redeemed by any of them — which is how dessert is actually bought.
- Reward Chosen by Iris Rather than a blanket discount, Iris picks the reward with the highest modelled redemption for that guest: a free topping for the frequent small spender, a family bundle for the Sunday visitor.
- Weather-Aware Offers A cold, wet Tuesday is when a loyalty nudge earns its keep. Offers can be released on weather conditions rather than on a fixed calendar.
- Liability Tracked Outstanding stamps and points are valued live and posted to finance, so loyalty never becomes a surprise sitting off the books.
“Which loyalty mechanic drives the most repeat visits — stamps, streaks or the family wallet?”
Delivery & Cold-Chain Integrity
Ice cream is the hardest thing on any menu to deliver. Packaging, dry ice, rider time and a temperature guarantee that either holds or refunds itself.
- Delivery Radius by Product Not everything travels. Radius is set per product from real melt data, so a soft-serve sundae is not offered to an address forty minutes away.
- Packaging & Dry Ice Rules Insulated packaging and dry-ice quantity are calculated from order size and estimated travel time, then consumed from stock and costed to the order.
- Rider Window Held The order is not made until a rider is genuinely close. Ice cream sitting on a shelf waiting for collection is the single biggest cause of a bad delivery.
- Temperature Guarantee Where the aggregator supports it, arrival condition is captured. Where it does not, the platform still models risk and flags the orders most likely to arrive soft.
- True Delivery Margin Commission, packaging, dry ice and refund rate are attributed to each channel, so you can see whether delivery is a growth channel or an expensive habit.
“What share of delivery orders arrived outside the temperature window last month?”
Drink Builder — Size, Base, Add-Ins, Sweetness
Complex drink customisation captured accurately every time — boba tea sweetness level, fruit base, bubble type, ice level, and serving size all configured at the POS in seconds.
- Size & Base Selection Small, medium, large — paired with base type: black tea, green tea, taro, fruit. Price adjusts automatically. Most popular customisation combinations surface as preset options — one tap for the most ordered builds.
- Sweetness & Ice Level 0%, 25%, 50%, 75%, 100% sweetness — and ice level: no ice, less, normal, extra. Both mandatory selections.
- Topping / Bubble Type
- Add-In Modifier Pricing Aloe vera, basil seeds, extra fruit — mixed into drink with individual pricing per add-in.
- Allergy & Dairy-Free Flags Oat milk swap, nut-free toppings — flagged on the drink label for the bar team.
“What is the most popular sweetness level for taro milk tea this week?”
Fast Bar Workflow — Cup Label Printing
Every drink order prints as a label at the bar — no verbal relay, no order confusion, no wrong drinks. Barista or bartender builds exactly what the POS says.
- Instant Label at Order Label prints at bar station the moment the order is placed at counter or via app. Queue ML estimates wait time per customer — displayed at counter so customers can plan.
- Full Spec on Label Size, base, sweetness, ice, toppings, add-ins, dietary flags, customer name — all printed.
- Multi-Drink Order Sequencing
- Queue Visibility Bar team sees all pending drinks on a display — can batch similar drinks for efficiency.
- Drink Completion Rate Time from order to collection tracked per barista — identifies workflow bottlenecks.
“What is the average wait time from order to collection during the Saturday afternoon rush?”
Age Verification Gate — Alcohol Items
When an alcohol item is added to an order, the POS displays a mandatory age verification prompt — staff must confirm the customer has been ID-checked before the item is added to the bill.
- Item-Level Alcohol Flag Alcohol items flagged in the menu — trigger the age check prompt automatically when selected. No ML model — deterministic rule engine. Rule triggers on every alcohol SKU. Compliance log is immutable.
- Mandatory Staff Confirmation Staff must confirm 'ID checked' before the item is added — cannot be bypassed.
- Refusal Logging
- Minor Alert If a companion appears underage while an alcohol item is on the table — alert prompts re-check.
- Compliance Report All age verifications logged — full audit trail for licensing compliance.
“How many age verification checks were performed last Friday night?”
Open Tab Management — Bar & Pub
Tab opened at arrival, drinks added throughout the session, settled at departure. Staff focus on service — not on re-opening orders for every round.
- Tab Opens at Arrival Tab created when a customer is seated or presents a card — all subsequent orders added without re-opening. High-spend tab alert fires when a tab reaches a configurable threshold — manager awareness without interrupting service.
- Add Rounds Without New Order Each round added to the running tab — no new POS transaction for each drink.
- Live Tab Total Visibility
- Tab Transfer Between Staff Tab assigned to a staff member — can be transferred when shift changes.
- Pre-Auth for Card Tabs Card pre-authorised at tab opening — reduces end-of-night payment friction.
“What is the average tab value on Friday nights vs Saturday nights?”
Shisha / Hookah Add-On Module
Shisha is sold as a time-based or fixed service alongside food and drinks. This module manages shisha orders, coal top-ups, flavour tracking, and per-session billing.
- Shisha as a Billable Item Shisha session added to the table's tab — flavour and pipe type recorded at order. Flavour demand forecaster predicts weekly consumption per flavour — guides purchasing to avoid stockouts.
- Coal Top-Up Logging Each coal refill logged against the session — tracks service frequency per table.
- Flavour Inventory Tracking
- Session Timer Optional timer per session — alert fires when the session is nearing its pre-paid duration.
- Flavour Popularity Report Which flavours are ordered most — guides purchasing and seasonal flavour rotation.
“Which shisha flavour was ordered most this month and which time of day?”
Loyalty Churn Predictor AI
Beverage loyalty programmes are highly retention-dependent. This module identifies members who are going cold and triggers automated win-back offers before they lapse.
- Visit Frequency Baseline Each loyalty member's normal visit frequency established over their first 8 weeks. Binary churn classifier trained on visit frequency, recency, and spend patterns. Win-back timing optimised by reinforcement learning on offer acceptance.
- Churn Risk Scoring Members who have fallen below their normal frequency receive a churn risk score — 0 to 100.
- Automated Win-Back Trigger
- Offer Performance Tracking Did the win-back offer bring them back? Tracked per member per campaign.
- Churn Rate Dashboard Overall loyalty churn rate tracked weekly — alongside active member count and visit frequency.
“How many loyalty members are at churn risk this week and what is the recommended offer?”
Weather-Demand Correlator AI
Drink demand changes dramatically with weather — cold drinks surge on hot days, hot drinks on cold days. This module anticipates demand before the weather arrives.
- Live Weather Feed Integration Real-time weather data — temperature, humidity, rain forecast — pulled into the demand model daily. Time-series demand forecast augmented with weather covariates — trained on historical orders and corresponding weather conditions.
- Drink Category Demand Prediction Hot drink vs cold drink vs blend demand predicted per hour based on forecast temperature.
- Preparation Recommendations
- Seasonal Pattern Learning Model learns which weather conditions drive which drinks in your specific location over time.
- Revenue Forecast Adjustment Weather-adjusted revenue forecast — more accurate than day-of-week average alone.
“How much more sugarcane juice did we sell on 38°C days vs 28°C days?”
Seasonal Drink Menu Auto-Switch
Summer menu, winter menu, Ramzan specials, Eid drinks — all pre-configured and switched automatically on the right date. No manual menu changes on the day.
- Date-Triggered Menu Activation Seasonal menu launches automatically on its start date — replaces or supplements the regular menu. Seasonal demand forecaster uses prior year data and current trends to predict volumes for new seasonal items.
- Graceful Item Handover Items being retired are tracked for sell-through before removal — no waste on last day.
- Seasonal Availability Badges
- Cross-Channel Simultaneous Update Counter, app, kiosk, and aggregators all switch at the same moment — consistent experience.
- Seasonal Performance Archive How did last year's summer menu perform? Accessed for planning this year's.
“How did our Ramzan special drinks perform vs the regular menu during the same period last year?”
Happy Hour & Dynamic Pricing Windows
Timed price windows that switch themselves — by hour, day, weather and live footfall — with the margin floor enforced so a promotion can never quietly sell below cost.
- Windows That Switch Themselves Prices, menus and combos change at the window boundary across till, kiosk, app and table QR simultaneously. No one has to remember to flip a board at 17:00, and no one can forget to flip it back.
- Priced Against a Floor Every discounted line carries its own margin floor from the recipe cost. A two-for-one on a cocktail with an expensive base spirit is refused rather than silently sold at a loss.
- Weather and Fixture Aware A warm Friday and a home fixture move drinks demand far more than the menu does. Iris widens or narrows the window against forecast footfall rather than a fixed calendar.
- True Incrementality, Not Reported Uplift Measured against a matched control group of comparable outlets: the volume that would have come anyway is separated from what the window actually created.
- Staffing Moves With It A window that doubles volume without adding a bartender produces a queue, not a profit. The forecast feeds the roster so the two decisions are made together.
“What did happy hour actually earn last month once the discount and the extra staffing are both counted?”
Draught Line & Keg Yield Control
Every keg tracked from tapping to float — poured volume against expected volume, line-cleaning schedules, gas usage and the yield loss that never reaches a till.
- Metered, Not Estimated Flow meters on each line record what actually poured. Expected yield from keg volume is compared with poured volume and till sales — three numbers that should agree and usually do not.
- Loss With a Cause Foam and first-pour waste, line-cleaning volume, over-pour and unrecorded staff drinks are separated. The fix for a badly balanced gas line is nothing like the fix for a generous bartender.
- Line Cleaning on Schedule Cleaning cycles are scheduled, logged and costed, with the volume sacrificed recorded rather than disappearing into shrinkage. A skipped clean shows as a quality risk before it shows as a complaint.
- Keg Float Predicted Iris projects when each keg will float from the current pour rate, so the change happens between rounds rather than mid-service with a queue waiting.
- Yield by Line and by Shift Variance is attributed to the line and the shift, which turns a vague sense that “we lose a lot on draught” into a specific conversation about a specific tap.
“Which line lost the most volume to waste last month, and was it foam, cleaning or over-pour?”
Cocktail Pour Cost & Recipe Control
Cocktail specs with real pour cost per serve, free-pour variance measured against jigger standard, and margin recomputed live as spirit prices move.
- A Spec, Not a Guess Each cocktail carries an exact spec — spirit, modifier, mixer, garnish, ice and dilution — with a cost per serve that reflects the actual bottle price this week rather than the one from last year.
- Free-Pour Drift Measured Free-pouring is faster and it drifts. Comparing theoretical spirit usage against actual reveals the drift per bartender and per shift — typically 6–12% before anyone notices.
- Garnish and Ice Counted Garnish, ice and dilution are real costs that most bars never attribute. On a high-volume signature serve they are frequently the difference between the margin you plan and the one you get.
- Recosted as Spirits Move When a base spirit rises, every cocktail built on it recosts in the same second and any that crosses its floor is flagged with the size of the gap — before the quarter ends.
- Menu Engineering for the Bar Cocktails are graded on contribution and velocity, so the signature serve that is loved and unprofitable is repriced or re-specced rather than defended out of habit.
“Which cocktails fell below margin floor this quarter, and was it price or pour?”
Bottle, Wine & Cellar Inventory
Spirits, wine and bottled stock counted by weight or by unit, with wine-by-the-glass yield, open-bottle life and cellar value tracked continuously.
- Counted by Weight High-value spirits are weighed rather than eyeballed, so a part-bottle is a number instead of an estimate. Variance stops being a monthly argument and becomes a daily fact.
- Wine by the Glass, Honestly Expected glasses per bottle are set by pour size and compared with sales. Over-pouring wine by the glass is one of the quietest margin leaks in the trade.
- Open-Bottle Life Preservation method and open date drive a shelf-life clock. A bottle past its window is flagged for markdown or staff training use rather than served to a guest who will notice.
- Cellar Value Live Stock value is current rather than a quarterly snapshot, so working capital tied up in slow-moving wine is visible while there is still time to do something about it.
- Reorder Against Real Velocity Par levels come from actual sell-through and lead time by supplier, which matters more for allocated and vintage stock than for anything else behind the bar.
“What is the current cellar value, and how much of it has been open longer than it should be?”
Subscription Drink Pass & Loyalty
A monthly coffee or drinks pass alongside points and rewards — with the unit economics watched so a heavy user stays profitable rather than becoming a loss.
- Pass, Points and Stamps Together A subscription pass, a points balance and a stamp card can all sit on the same wallet without conflicting. The guest sees one balance; you see three mechanics doing different jobs.
- Redemption Rules at the Till One drink per visit, eligible sizes, blackout windows and fair-use limits are enforced at the moment of redemption rather than discovered in a monthly reconciliation.
- Unit Economics Watched Iris tracks contribution per subscriber. A pass priced for two visits a week that gets used ten times is not loyalty, it is a subsidy — and the platform says so before the quarter closes.
- Churn Predicted Early A subscriber whose visit frequency has halved is flagged while the habit is recoverable, with the specific offer most likely to bring them back rather than a blanket discount.
- Deferred Revenue Handled Properly Subscription income is recognised as it is consumed rather than when it is collected, so the books reflect the obligation you actually carry.
“Which subscribers are unprofitable at their current redemption rate, and what would fix it?”
Omni-Channel Ordering — Bar, App, Table QR
Counter, kiosk, table QR, app, web and the delivery aggregators on one order book — with bar-specific routing so a round arrives together rather than in pieces.
- Table QR Without the Chaos Guests order from the table while staff keep serving. Orders arrive sequenced by promise time rather than by who reached the bar first, which is the fairest queue a busy bar can offer.
- Rounds Kept Together A round of six drinks is built and released as one, so nobody carries two pints back and returns for the rest. Bar service is judged on the round, not the drink.
- Live Availability Everywhere A keg floats and that beer disappears from table QR, app and aggregator in the same second, with the nearest equivalent offered in its place.
- Promise Times That Reflect the Bar Quotes account for live queue depth and station load. On a Saturday at 21:00 an honest twelve minutes beats an optimistic four that everyone knows is wrong.
- One Ledger, True Channel Margin Aggregator commission, packaging and delivery cost are attributed per channel, so the channel that looks busiest is judged on what it actually contributes.
“What share of drinks orders came through table QR versus the bar last Saturday, and which was faster?”
Responsible Service & Compliance Log
Age verification, refusal logging, licensed-hours enforcement and incident records — the evidence trail a licensing review will ask for, captured as it happens.
- Age Check at the Point of Sale Age-restricted lines cannot be rung through without a verification step. The check is recorded with staff member and time — which is precisely what a licensing officer asks to see.
- Refusals Logged, Not Remembered A refusal to serve is captured in seconds with reason and witness. A bar that can produce a refusal log is in a fundamentally stronger position than one relying on recollection.
- Licensed Hours Enforced The till refuses alcohol lines outside licensed hours rather than trusting everyone to watch the clock during a busy close. Extensions are configured, dated and logged.
- Intoxication and Incident Records Incidents are recorded with time, staff, location and action taken, building the evidence trail that protects the licence and the people working under it.
- Training Currency Tracked Personal licence and responsible-service certification are tracked against expiry, and the roster will not place an uncertified person on a licensed station.
“How many refusals were logged last month, and were licensed hours breached at any outlet?”
Waste, Spillage & Over-Pour Variance
Theoretical against actual across draught, spirits, wine and mixers — with spillage, breakage, comps and staff drinks separated so each has an owner.
- Four Causes, Four Fixes Spillage, breakage, comps and over-pour are recorded separately. Treating them as one number called “wastage” guarantees none of them ever improves.
- Comps With a Reason Every comp carries a reason and an authoriser. Generosity that recovers an unhappy guest is worth paying for; generosity that is habit is worth knowing about.
- Variance by Shift and Bartender Attribution to shift and person turns an abstract percentage into a specific coaching conversation — or occasionally a specific control problem.
- Costed at Real Prices Variance is valued at current cost, not last year’s, so the number reflects what the loss is actually worth today.
- Fed Back Into Ordering Persistent variance on a line adjusts its par level, so you stop ordering to cover a leak you have not yet fixed.
“Which category lost the most value to variance last month, and at which outlet and shift?”
Integration Hub — Payments, Delivery & Tax
Card and wallet payments, the delivery aggregators, licensed-trade tax rules and the accounting ledger — connected, health-checked and configured per market.
- Payments — Every Method, One Settlement Stripe, Adyen, Square and PayPal as processors; Apple Pay, Google Pay, cards and local wallets as methods. Tabs, split bills and tips all settle into one reconciliation.
- Delivery Apps — One Menu Everywhere DoorDash, Uber Eats, Deliveroo and Just Eat through Deliverect, with alcohol lines correctly restricted per market rather than published where they cannot legally be sold.
- Tax — Licensed Trade Rules, Configured Alcohol duty, VAT and eat-in versus take-away rates differ by market and by product. Rules are configured per market rather than coded per customer.
- Accounting — Sales Post Themselves Sales, duty, tips, comps and settlement post to QuickBooks, Xero or the group ERP already coded to the right accounts.
- Connection Health — Watched, Not Assumed Every connection is health-checked with last-sync times and error rates, and Iris quantifies the orders at risk when an API degrades.
“Which integration failed most often last month, and how many orders did it cost us?”
Allergen & Dietary Flag per Item
Every menu item is tagged with its allergen profile — gluten, dairy, nuts, soy, eggs, shellfish. Tags are visible to staff at the POS and to customers on the app and kiosk.
- 14 Allergen Tagging All 14 major allergens tagged per item — staff see flags on the POS item card before confirming. Order-level allergen conflict checker — if a customer's saved profile clashes with any item in their basket, alert fires before confirmation.
- Customer-Facing Allergen Filter App and kiosk allow customers to filter the menu by their allergen profile — only safe items shown.
- Modifier Allergen Awareness
- Cross-Contamination Warning Items prepared in shared equipment or areas flagged with cross-contamination risk.
- Allergen Audit Trail If a complaint occurs — order record shows exactly what was ordered and what allergens were present.
“How many orders this month included a nut allergen flag?”
Calorie & Macro Display per Dish
Every item's calorie count and macronutrient breakdown is displayed on the POS, kiosk, and app — enabling customers to make informed choices and meeting regulatory requirements.
- Calorie Display at POS Calories shown next to every item on the counter POS — staff can confirm for customers. Nutritional recommendation model suggests items that meet a customer's stated calorie or macro target.
- Macro Breakdown on App / Kiosk Protein, carbohydrates, fat, and fibre displayed per item — full nutritional transparency.
- Build-Your-Own Nutrition Calculator
- Daily Intake Guidance % of recommended daily intake shown — contextualises the number for health-conscious customers.
- Regulatory Compliance Export Nutritional data exportable in the format required for menu labelling compliance.
“What is the average calorie count of orders placed on weekday lunchtimes?”
Build-Your-Own Bowl / Plate Engine
Customers configure their meal from a base, protein, toppings, and dressing — each choice tracked as an ingredient, priced accurately, and depleted from inventory.
- Step-by-Step Build Flow Base → protein → toppings → dressing — each step presented in sequence with available options. Popular build clustering identifies the top 5 configurations — these become suggested 'signature bowls' on the menu.
- Per-Component Pricing Each component priced individually — total updates as customer builds their bowl.
- Ingredient-Level Inventory Depletion
- Portion Size Control Standard portion per component defined — staff prompted if a portion looks oversized.
- Popular Build Tracking Most common bowl configurations tracked — identifies potential preset menu items.
“What are the 5 most commonly built bowl combinations this month?”
Subscription Meal Plan Billing
Customers pre-pay for a weekly or monthly meal plan — fixed number of meals, specific days, or a calorie target. Billing, delivery scheduling, and plan management all in one system.
- Plan Configuration 5-day lunch plan, 3-meal daily plan, weekly calorie target — each plan defined with inclusions and rules. Plan utilisation classifier identifies customers under-using their plan — churn risk high. Proactive check-in triggered.
- Advance Billing Plan billing collected weekly or monthly in advance — card on file charged automatically.
- Delivery / Collection Schedule
- Plan Pause & Resume Customer can pause their plan for a week — system suspends billing and delivery automatically.
- Plan Utilisation Report Which customers are consistently using their plan vs lapsing — retention risk flagged.
“How many meal plan subscribers have used less than 50% of their plan this month?”
Organic / Batch Inventory Tracking
Organic and farm-to-table produce arrives in batches with specific provenance. Each batch is tracked separately — expiry, supplier, certification — and depleted in FIFO order.
- Batch Receipt Logging Each produce delivery logged as a batch — supplier, date, quantity, certification number. Expiry risk scorer predicts which batches will expire before sell-through — triggers menu prioritisation or markdown.
- FIFO Depletion Oldest batch depleted first automatically — reduces waste and ensures freshness.
- Expiry Alert
- Certification Tracking Organic, halal, gluten-free certifications attached to each batch — auditable at any time.
- Supplier Traceability Report Any ingredient can be traced back to its source batch, supplier, and delivery date.
“Which ingredients are approaching expiry in the next 3 days?”
Nutritional AI Recommender
Based on a customer's stated health goal — weight loss, muscle gain, energy boost — the AI recommends the most suitable items from the current menu, with full nutritional justification.
- Health Goal Capture Customer selects their goal — weight loss, muscle gain, low carb, high protein, balanced. Recommendation model scores items against customer nutritional profile. Trained on accepted vs rejected suggestions over time.
- Menu Scoring Against Goal AI scores every current menu item against the stated goal — ranks by suitability.
- Top 3 Recommendation with Rationale
- Macro Gap Identification If a customer is low on protein this week, AI suggests protein-dense options.
- Recommendation Acceptance Tracking Which AI suggestions are accepted — model improves over time based on real choices.
“What is the most recommended item for customers with a weight-loss goal?”
Meal Prep Demand Forecaster AI
Protein meal prep is produced in advance — wrong forecasts mean waste or stockouts. This model predicts daily demand per meal type and guides production the evening before.
- Per-Meal Demand Forecast Predicted orders per meal type for tomorrow — broken down by delivery window and collection time. Time-series demand forecast with subscription base as an anchor. MAPE under 9% on 30-day test sets.
- Plan-Based Demand Base Active subscription plans provide a known order floor — forecast adds walk-in and app orders on top.
- Day-of-Week Seasonality
- New Item Demand Estimation Launching a new meal? Model estimates demand based on similar items' historical performance.
- Production Plan Output Forecast output is a ready-to-use kitchen production plan — how many of each meal to prepare.
“What is tomorrow's predicted demand for the high-protein chicken bowl?”
Ingredient Traceability — Farm to Table
Every ingredient's journey from farm to plate is recorded. Supplier, farm, certification, delivery batch, and recipe usage — all linked and auditable in one system.
- Supplier-to-Batch Linking Each delivery batch linked to its supplier, farm name, and certification — one-click traceability. No ML model — deterministic traceability chain. Recall response is automated rule-based: batch → dishes → orders → customers notified.
- Recipe-to-Ingredient Linking Each dish links to its ingredients — trace any dish back to the exact batch used on any given day.
- Recall Response
- Customer-Facing Provenance Menu app can display where each key ingredient comes from — supports farm-to-table brand positioning.
- Certification Expiry Tracking Organic, halal, gluten-free certifications have expiry dates — system alerts before a certificate lapses.
“Which orders used the chicken batch from Supplier X that was recalled last Tuesday?”
Macro & Calorie Engine — Calculated, Not Estimated
Every build computes its own calories, protein, carbs, fat, fibre and sugar from the actual ingredients and weights used — so the number on the label is the number in the bowl.
- Computed From the Build A bowl with double protein and extra dressing is not the bowl on the menu board. Macros are calculated from what was actually added, at the weights actually used, rather than from a menu average.
- Weights, Not Scoops Scoop sizes drift by staff member and by how busy it is. Scale-linked portions mean the protein figure is a measurement rather than an assumption — which matters when a guest is counting.
- Every Ingredient Carries Its Data Calories, macros, fibre, sugar, sodium and micronutrients sit on the ingredient, so adding a new topping updates every build that uses it automatically.
- Label Printed From the Same Source The sticker, the app and the receipt all read the same computed values. Three different numbers for the same bowl is the fastest way to lose a health-conscious guest.
- Ranges Stated Honestly Natural variation in produce is real. The platform states a range where a range is truthful rather than implying a precision that a kitchen cannot deliver.
“Which builds most often exceed their stated calorie range, and which ingredient causes it?”
Allergen Enforcement & Cross-Contact Control
Allergens propagated from every ingredient and sub-recipe, enforced at the point of order, with cross-contact risk flagged by station and prep sequence.
- Blocked, Not Warned A guest-declared allergy blocks the incompatible build at the till rather than showing a message someone can tap past. On allergens, a warning is not a control.
- Propagated Through Sub-Recipes A sauce containing sesame carries sesame into every bowl that uses it, automatically. Manual allergen matrices go stale the first week someone changes a dressing.
- Cross-Contact by Station Shared utensils, shared fryers and prep sequence are modelled, so ‘may contain’ reflects your actual kitchen rather than a legal template copied from elsewhere.
- The Guest’s Allergy Travels A recorded allergy follows the profile across counter, app and delivery, so the guest states it once rather than every visit and hopes it is remembered.
- Evidence for an Inspection Every allergen check, override and refusal is logged with staff and time — the record an environmental health officer asks for, and the one you need if a claim is ever made.
“Which allergens are declared most often by guests, and where do our cross-contact risks sit?”
Build-Your-Own Pricing & Portion Logic
Base, protein, toppings, dressing and extras priced by tier or by weight, with portion guidance at the line so margin survives a generous shift.
- Tiered or Weighed Some items price cleanly by tier; premium proteins and avocado do not. The engine supports both on the same build without confusing the guest at the counter.
- Margin Floor per Build Because a build is assembled by the guest, its cost is variable. The floor is checked as the bowl is composed, and an unprofitable combination is repriced rather than discovered at month-end.
- Portion Guidance at the Line Scale prompts show target weight per component. Over-portioning premium protein by 12 grams a bowl is invisible per bowl and substantial across 486 a day.
- Upsell That Fits the Goal Iris suggests the addition that matches the guest’s stated goal — protein for a training day, fibre for a light lunch — rather than the most expensive topping on the line.
- Waste Tracked by Component Component-level waste shows which toppings are prepped optimistically. Prep quantities then follow the forecast rather than habit.
“Which topping combinations sell below target margin, and how often?”
Meal Plan Subscriptions & Unit Economics
Weekly meal plans with delivery windows, dietary profiles and pauses — with contribution tracked per subscriber so growth does not quietly become a loss.
- Plans Built Around a Diet Macro targets, exclusions and dislikes generate the week’s menu automatically, so a subscriber is not sent the same six meals until they cancel.
- Pause, Skip and Swap Without Support Guests pause for holidays, skip a day, swap a meal. Every one of those handled in the app is a support call you do not take and a cancellation you do not receive.
- Contribution per Subscriber Iris tracks margin per subscriber including packaging and delivery. A plan priced for five meals that is consumed as seven with two swaps is not loyalty, it is a subsidy.
- Churn Predicted From Behaviour Skipped deliveries, unopened emails and declining swaps predict cancellation earlier than a stated complaint does. Intervention is cheapest while the plan is still active.
- Deferred Revenue Handled Correctly Subscription income is recognised as meals are delivered rather than when payment clears, so the books reflect the obligation actually outstanding.
“Which meal plan tiers are unprofitable at current redemption, and what would fix them?”
Ingredient Sourcing, Batch & Traceability
Every batch traced from supplier to bowl — organic certification, harvest date, cold-chain history and the ability to answer a recall in minutes rather than days.
- Batch Captured at Receipt Supplier lot, harvest or production date, certification and temperature on arrival are recorded at goods-in rather than reconstructed later from paperwork.
- Forward Trace to the Guest A recalled batch resolves to the preps that used it, the bowls that contained them and — where the guest is identified — the orders to contact. Minutes, not days.
- Certification That Is Current Organic, halal, kosher and fair-trade certificates carry expiry dates. A lapsed certificate blocks the claim on the menu before a guest reads something untrue.
- Cold Chain Unbroken Temperature history from delivery through storage to prep, so a claim about freshness rests on a record rather than on confidence.
- Provenance the Guest Can See Where it makes sense, the guest can see the farm and the harvest week. In wellness the story is part of the product — but only if it is verifiable.
“If a supplier recalls a batch, which orders and which guests were affected?”
Loyalty, Goals & Guest Wellness Profile
A guest profile built around goals rather than points — dietary preference, macro targets, training pattern — driving recommendations that actually fit.
- A Goal, Not Just a Balance Cutting, maintaining, training, recovering, managing a condition — the goal changes what a good recommendation looks like far more than purchase history does.
- Recommendations That Respect It Iris suggests within the guest’s stated constraints. Recommending a 900-calorie bowl to someone tracking a deficit is worse than recommending nothing at all.
- Progress the Guest Owns Weekly macro summaries belong to the guest, in their app, deletable. Wellness data demands a higher standard of handling than a coffee stamp card.
- Rewards That Fit the Category A free dessert is the wrong reward here. Extra protein, a premium topping or a nutrition consultation land better and cost less.
- Lapsed Guests Recovered Honestly A guest whose visits halved is flagged with what they used to order — and if their goal has changed, the offer changes with it.
“Which guests changed their goal recently, and did their order pattern follow?”
Omni-Channel — Counter, App, Delivery & Meal Prep
Counter, kiosk, app, delivery aggregators and scheduled meal-prep collection on one order book — with builds, macros and allergens intact across every channel.
- The Build Survives the Channel A custom bowl ordered in the app arrives at the line with every component, substitution and allergy note intact. Rebuilding it from a free-text note is how errors happen.
- Macros Shown Before Purchase The app computes and shows macros as the guest builds, which changes what they order — usually toward the outcome they came for.
- Delivery That Does Not Wilt Dressings separated, hot and cold components split, and items that travel badly hidden from delivery. A soggy bowl costs more in reputation than the order was worth.
- Meal Prep on a Schedule Weekly prep collections are planned production rather than same-day demand, which is the most predictable and highest-margin volume in this sector.
- True Margin per Channel Commission, packaging and the extra prep time each channel consumes are attributed, so app growth is judged on contribution rather than order count.
“What share of bowls came through app versus counter, and which channel has the higher average build value?”
Nutrition Compliance & Menu Labelling
Calorie labelling, nutrition claims, allergen statements and regional regulation handled per market — with every claim on the menu backed by the data that produced it.
- Claims Substantiated or Blocked ‘High protein’, ‘low sugar’ and ‘source of fibre’ have legal thresholds in most markets. The platform checks the computed value against the threshold and refuses to publish a claim it cannot support.
- Calorie Labelling per Market Mandatory calorie display applies in some markets and not others, with different rules for outlet size and channel. Configured per market rather than coded per customer.
- Allergen Statements Generated The statement is generated from the recipe rather than typed by a person, so it cannot drift out of date when a supplier changes a formulation.
- Regulation Changes Tracked When a threshold or labelling rule changes, affected items are listed with the specific claim at risk rather than leaving someone to re-read the menu.
- An Honest Limit Nutrition values for fresh produce carry natural variation, and the platform states this rather than implying laboratory precision. Overclaiming accuracy is its own compliance risk.
“Which menu items carry a nutrition claim, and can each one be substantiated today?”
Integration Hub — Payments, Delivery, Nutrition & Tax
Payments and wallets, delivery aggregators, nutrition databases, subscription billing, local tax rules and the accounting ledger — connected, health-checked and configured per market.
- Payments and Subscription Billing Stripe, Adyen, Square and PayPal for one-off orders; recurring billing, dunning and card-update handling for meal plans — with failed renewals recovered rather than silently churned.
- Nutrition Data In, Not Typed Ingredient nutrition sourced from recognised databases and supplier specifications, versioned so a change is visible rather than silent.
- Delivery Apps — Build-Aware Uber Eats, Deliveroo, Just Eat and DoorDash through Deliverect, with custom builds and allergen notes passed through intact rather than flattened into a comment field.
- Tax — Eat-In, Take-Away and Cold Food Rates differ by market and frequently by temperature and where the food is eaten. Configured per market, applied before the guest is asked.
- Connection Health — Watched, Not Assumed Every connection health-checked with last-sync times and error rates, and Iris quantifies the orders at risk when an API degrades.
“Which integration failed most often last month, and how many orders did it cost us?”
Multi-Brand Order Routing
A single kitchen operates multiple virtual brands — each with its own menu, packaging, and identity. Orders from all brands arrive in one queue and routed to the correct kitchen station.
- Brand-Specific Order Streams Each brand's orders displayed separately on the KDS — kitchen never confuses Brand A's packaging with Brand B. Kitchen load predictor models concurrent demand across all brands — alerts when a station is approaching capacity.
- Shared Station Routing Where brands share a kitchen station, orders are queued and prioritised by target time.
- Brand-Level KDS Display
- Simultaneous Multi-Brand Orders Kitchen handles concurrent orders from all active brands without conflict — throughput tracked per brand.
- Brand Order Volume Dashboard Live count of orders per brand — manager sees demand split in real time.
“Which brand had the highest order volume between 7pm and 9pm this week?”
Per-Brand Revenue Settlement
Revenue, COGS, and shared costs are tracked and settled per brand — owner can see exactly how each virtual brand is performing and how shared kitchen costs are allocated.
- Brand-Level Revenue Tracking Sales, refunds, and aggregator commissions tracked per brand — total and net revenue separated. Brand performance ranker scores each brand by revenue, margin, and order volume — identifies which brands to grow and which to retire.
- Shared Cost Allocation Kitchen rent, staff, utilities — allocated to brands by order volume, revenue %, or fixed split.
- Brand P&L Report
- Aggregator Commission Breakdown Foodpanda and Uber Eats commissions deducted per brand — net payout vs gross sale tracked.
- Multi-Owner Settlement If different brands are owned by different partners — settlement report per owner with audit trail.
“Which of our virtual brands has the highest net margin after aggregator commission?”
Virtual Brand Management
Create, configure, and manage delivery-only virtual brands without a physical presence. Menu, pricing, branding, and platform listings all managed from one dashboard.
- Brand Creation New virtual brand created in the dashboard — name, logo, cuisine type, and menu configured. Virtual brand launch advisor analyses local delivery demand data — recommends which cuisine gap to fill with a new virtual brand.
- Independent Menu per Brand Each brand has its own menu — different items, prices, and descriptions even if made in the same kitchen.
- Platform Listing Management
- Brand Performance Comparison Side-by-side comparison of all active brands — revenue, orders, rating, and margin.
- Brand Sunset Protocol Underperforming brand deactivated from all platforms simultaneously — no manual delisting per platform.
“Which cuisine type has the highest demand but lowest supply on Foodpanda in our delivery zone?”
Delivery Aggregator Sync — All Platforms
Orders from every delivery platform arrive in one system. Menu updates, pricing changes, and item availability propagate to all platforms simultaneously — from a single dashboard.
- Unified Order Queue Foodpanda, Uber Eats, Talabat, Deliveroo — all orders in one queue. No separate tablets. AI Aggregator Demand Surge Predictor — forecasts order spikes by platform and hour, alerts kitchen to pre-stage before the surge arrives.
- Real-Time Menu Sync Price change, item deactivation, or new item added in Zentallio — syncs to all platforms in under 60 seconds.
- Platform-Level Order Acceptance
- Aggregator Revenue Dashboard Gross sales, commission, and net revenue per platform — side by side, updated in real time.
- Platform Rating Tracking Customer rating per platform tracked alongside order volume — correlates rating with operational events.
“Which platform had the most orders between 8pm and 10pm last Friday?”
Dark Store FIFO Inventory
Dark stores combine grocery and F&B fulfilment. Inventory is managed on strict FIFO rotation — oldest stock picked first, expiry tracked per item, and replenishment automated.
- FIFO Bin Assignment Every SKU assigned a bin location — new stock placed behind existing. Picker always selects from the front. Demand forecast per SKU drives reorder point calculation — adjusts dynamically as sales patterns change.
- Expiry Date per Batch Expiry date logged at each goods receipt — FIFO enforces oldest-first picking automatically.
- Expiry Alert
- Real-Time Stock Level Live stock level per SKU updated at every order pick — no manual counting during shifts.
- Replenishment Auto-Order When a SKU hits its reorder point, a purchase order is generated automatically to the preferred supplier.
“Which SKUs are at risk of expiry in the next 48 hours?”
Brand Performance Ranker AI
Each virtual brand is scored and ranked across revenue, margin, order volume, customer rating, and repeat rate — enabling data-driven decisions about which brands to invest in.
- Multi-Metric Ranking Brands ranked by composite score — weighted across revenue, margin, orders, rating, and repeat rate. Multi-metric ranking model (composite scoring) with trend weighting — brands on an upward trajectory scored more favourably.
- Per-Metric Breakdown See exactly where each brand is strong and weak — revenue leader may be margin laggard.
- Trend Direction
- Benchmarking Against Category Average How does a brand compare to others in its cuisine type on the same platform?
- Action Recommendations AI recommends: invest more in Brand A, review pricing on Brand B, consider retiring Brand C.
“Which of our brands is declining in margin despite growing in order volume?”
Delivery Demand Surge Predictor AI
Predicts delivery order spikes by brand, platform, and hour — so the kitchen can pre-stage ingredients and staff before the rush arrives, not after it overwhelms them.
- Hourly Order Volume Forecast Predicted orders per brand per hour — for the next 6 hours, updated every 30 minutes. Time-series demand forecast with platform, day, weather, and event covariates. Updated in real time as the day's actual orders develop.
- Surge Alert When forecast exceeds normal capacity, kitchen receives an alert 30 minutes in advance.
- Platform-Specific Patterns
- Day / Event Adjustment Sports events, public holidays, rain — all correlated with historical surge data and incorporated into the forecast.
- Pre-Stage Recommendations 'Prepare 40 extra portions of Brand A's top item — surge expected between 7:30pm and 9pm.'
“What time do we typically see our biggest delivery surge on Fridays?”
Rider Handover & Pickup Gate
Prep timed backwards from rider arrival, a shelf that knows which bag belongs to which rider, and a handover scan that ends the disputed-order argument.
- Cooked to the Rider, Not the Clock Starting prep the moment an order lands guarantees food waits. Prep is scheduled backwards from predicted rider arrival, so the bag is sealed as the rider walks in.
- A Shelf That Knows Numbered, lit collection slots matched to order and rider. A rider taking the wrong bag costs two orders, two refunds and two ratings — the cheapest thing in this sector to fix.
- Handover Scanned One scan at pickup timestamps the handover and closes the order. When an aggregator claims a late delivery, you have the minute the rider took it rather than an argument.
- Shelf Time Is the Quality Metric Delivery ratings correlate with shelf time far more than with cook quality. Iris tracks minutes-on-shelf per brand and flags the pattern rather than the incident.
- Rider Waiting Time Tracked A kitchen that keeps riders waiting is deprioritised by the platform's own algorithm. Rider wait is measured because it decides how much volume you are offered tomorrow.
“How long does food sit on the shelf before pickup, by brand and by aggregator?”
Brand-Level P&L & Contribution
Each virtual brand carrying its own P&L — revenue, food cost, packaging, commission, labour share and marketing — so a brand that looks busy is judged on what it actually contributes.
- Revenue Is Not Contribution A brand doing 40% of orders at 26% commission on low-ticket items can contribute less than one doing 15% at higher ticket. Order count is the most misleading number in this sector.
- Shared Cost Allocated Honestly Kitchen labour, rent, energy and packaging shared across brands by an agreed driver — orders, prep minutes or covers — set once and applied consistently rather than negotiated every month.
- Commission Attributed Per Order Each order carries its actual platform commission, promotion cost and delivery fee, so brand margin is real rather than an average applied after the fact.
- Marketing Attributed to Its Brand Platform-funded promotions, sponsored listings and discounts land against the brand that ran them, which is the only way to judge whether the spend worked.
- Close a Brand on Evidence A virtual brand is cheap to launch and easy to keep alive out of sentiment. The P&L makes the decision to close it a number rather than a debate.
“Which virtual brand has the highest revenue and the lowest contribution, and why?”
Aggregator Commission & Price Parity
Commission, promotion cost and channel margin tracked per order — with menu price uplift set per platform so a discount never quietly sells below cost.
- Uplift Set Per Platform A 26% commission requires a different shelf price from a 14% direct order. Uplift is configured per platform and applied automatically rather than maintained by hand across five portals.
- Promotions Costed Before They Run Platform promotions look free because the platform proposes them. The cost lands on you. Every offer is modelled against contribution before it is accepted.
- Margin Floor by Channel An order that would settle below its floor after commission and promotion is refused or repriced rather than accepted and regretted at month-end.
- Parity Managed Deliberately Where price parity clauses apply, the platform tracks them so you stay compliant — and where they do not, direct ordering is priced to be genuinely worth choosing.
- Direct Channel Given a Reason Iris quantifies what moving a repeat guest from aggregator to direct is worth over a year. That number is what should fund the incentive to switch them.
“What is our true margin per aggregator after commission and promotions, by brand?”
Virtual Brand Launch & Menu Cloning
Spin up a new virtual brand from an existing kitchen in days — cloned recipes, its own menu, packaging and storefronts, with a defined kill criterion before launch.
- Built From What You Already Cook A new brand mostly recombines existing preps. Cloning recipes rather than inventing them keeps the kitchen's complexity flat while the storefront count grows.
- Storefronts Published Together Menu, photography, descriptions and pricing published to every platform from one place. Maintaining five storefronts by hand is how prices and availability drift apart.
- Complexity Counted Honestly Each new brand adds SKUs, packaging lines and pass complexity. Iris shows the marginal cost so a launch is judged on net contribution rather than incremental revenue.
- A Kill Criterion Set First Threshold, review date and decision-maker agreed before launch. Brands that quietly persist below breakeven are the most common waste in delivery-only operations.
- Cannibalisation Measured A new brand that takes orders from your existing one has not grown anything. Measured against a matched control rather than assumed to be incremental.
“Which brands launched this year cleared their contribution threshold by week 12?”
Delivery Demand & Prep Forecasting
Order volume predicted by brand, hour and platform — with weather, promotions and local events feeding prep quantities and staffing rather than a guess at 10am.
- Forecast Per Brand and Platform Six brands on four platforms is twenty-four demand curves, not one. They peak at different times and respond differently to weather and promotion.
- Weather Moves Delivery Hardest Rain lifts delivery volume more than almost any other signal in food service. A kitchen that staffs for it earns the surge; one that does not, gets bad ratings during it.
- Prep Quantities, Not Just Totals The forecast resolves to component preps — how much of each base, protein and sauce — because that is the decision the kitchen actually makes at 10am.
- Surge Staffed in Advance Predicted surges drive the roster. Discovering a surge as it happens means either riders waiting or orders declined, and both are penalised by the platform.
- Accuracy Reported Per Brand MAPE tracked by brand so trust is earned with a number. A new virtual brand with six weeks of history is forecast with appropriate humility.
“What volume should we prep for Friday 19:00–21:00 across all six brands?”
Packaging, Cost & Order Assembly
Packaging specified per dish, costed per order, and assembly checked before sealing — because a missing item costs the order, the rating and the refund.
- Packaging Is a Real Cost Line Between 4% and 9% of order value in most delivery-only operations, and frequently absent from brand P&L entirely. Specified per dish and costed per order.
- Assembly Checked Before Sealing A generated pack list checked at the bench. Missing items are the single largest driver of refunds in delivery, and almost all of them happen at assembly rather than in the kitchen.
- Packed for the Journey Hot and cold separated, sauces sealed, items that steam vented. The dish is judged where it is eaten, not where it left.
- Refunds Traced to Cause Missing item, wrong item, spillage, temperature, lateness. Only the first two are yours to fix at assembly, and separating them stops you fixing the wrong thing.
- Sustainable Where It Holds Up Recyclable and compostable options tracked with honest cost and honest performance — packaging that fails in transit is not a sustainability win.
“What is our packaging cost per order by brand, and how often are orders assembled incomplete?”
Dark Store Inventory & FIFO Depletion
Ingredient stock across brands depleting through shared recipes, with FIFO rotation, expiry alerts and shared-component visibility across every brand that uses them.
- One Store, Many Brands Shared ingredients deplete across every brand that uses them. Six brands each keeping their own mental stock count is how a kitchen runs out mid-service.
- FIFO Enforced at the Bench Rotation is prompted rather than trusted. In a high-throughput dark kitchen, first-in-first-out survives about a week on goodwill alone.
- Expiry Ahead of Loss Items approaching expiry are surfaced with the brands and dishes that could consume them, so the stock is used rather than written off.
- Availability Propagated Instantly An ingredient runs out and every dish that needs it goes unavailable across all brands and all platforms in the same second — before an order is accepted that cannot be made.
- Variance Attributed to a Brand Theoretical against actual usage per brand, so a portioning problem is traced to the brand and station that caused it rather than absorbed into a shared number.
“Which shared ingredients are closest to expiry, and which brands consume them?”
Ratings, Refunds & Platform Standing
Ratings, refund rate, cancellation rate and preparation-time accuracy watched per brand and per platform — because platform standing decides how much volume you are offered.
- Standing Decides Volume Aggregator algorithms rank on rating, acceptance, prep accuracy and cancellations. Slipping standing quietly reduces the orders you are shown, long before anyone notices a revenue drop.
- Refunds by Cause, Per Brand Missing item, cold food, lateness, quality. Attribution by brand and cause is the difference between fixing assembly and blaming riders.
- Prep Time Accuracy Tracked Quoting eighteen minutes and taking twenty-nine is penalised twice — by the guest and by the platform. Quoted times come from live kitchen load rather than an optimistic default.
- Reviews Scored and Themed Free text scored for sentiment and themed automatically, so a pattern across three hundred reviews surfaces without anyone reading them all.
- Recovery Before the Rating Sticks A refunded order triggers immediate outreach where the platform allows it. Recovering a guest before they rate is worth more than any response posted afterwards.
“Which brand's rating is falling on which platform, and what is the stated reason?”
Integration Hub — Aggregators, Payments & Tax
Every aggregator, payment processor, dispatch platform, tax regime and the accounting ledger — connected, health-checked and configured per market and per brand.
- Every Aggregator, One Order Book Uber Eats, Deliveroo, Just Eat, DoorDash and Glovo through Deliverect — six brands across five platforms without a wall of tablets and a person to watch them.
- Own Channel Alongside Them A direct web and app storefront running on the same menu and stock, so the cheapest channel is not the worst-maintained one.
- Payments and Settlement Reconciled Aggregator payouts arrive net of commission, promotions, refunds and adjustments. Automatic reconciliation against orders is the only practical way to know you were paid correctly.
- Tax by Market and Channel Delivery, collection and dine-in rates differ, as do rules on service and delivery fees. Configured per market rather than coded per customer.
- Connection Health — Watched, Not Assumed A degraded aggregator API means orders accepted that cannot be made, or a menu that will not update. Iris quantifies the orders at risk rather than waiting for a complaint.
“Which integration failed most often last month, and how many orders did it cost us?”
Headcount Billing — Per Person / Package
Institutional food service bills by headcount — not by individual item. This module manages per-person rates, package tiers, and group billing across all institutional formats.
- Per-Head Rate Configuration Set a daily per-person rate per meal type — breakfast, lunch, dinner — across multiple institutional clients. Headcount demand forecaster predicts attendance by day — guides kitchen on how much to prepare per meal service.
- Headcount Capture Daily headcount entered by location manager or synced from access control system.
- Tiered Package Pricing
- Group Billing per Client All meals for Client A consolidated into a single monthly invoice — with meal-type breakdown.
- Variance Reporting Budgeted headcount vs actual — client can track spend against their catering budget.
“What is the average daily headcount for lunch at our corporate client this month?”
Subsidised Meal & Voucher Engine
Corporate clients subsidise employee meals — the employee pays a portion, the company pays the rest. Vouchers, meal credits, and subsidy rules are managed automatically.
- Subsidy Rule Configuration Company pays 60%, employee pays 40% — or a fixed PKR/GBP subsidy per meal. Configured per client. Subsidy utilisation classifier identifies employees consistently under-using their meal credit — HR can use this for benefit optimisation.
- Employee Meal Credits Monthly meal credits allocated per employee — balance deducted at each meal, topped up monthly.
- Voucher Redemption
- Employee vs Company Billing Split Each meal generates two billing lines — employee portion and company subsidy — settled separately.
- Subsidy Utilisation Report Which employees are using their full benefit vs leaving credits unused — reported monthly.
“What percentage of employee meal credits were unused last month?”
Corporate Account Invoicing
Meals served to a corporate client are consolidated and invoiced on a pre-agreed cycle — weekly, fortnightly, or monthly — with full meal-by-meal detail for finance reconciliation.
- Account-Level Meal Tracking Every meal served to a client's employees logged against their account — in real time. Invoice anomaly detector flags invoices that deviate from the client's historical billing pattern — catches errors before they are sent.
- Cycle-Based Invoice Generation Invoice auto-generated at the end of each billing cycle — no manual compilation.
- Itemised Invoice Detail
- Purchase Order Matching Client's PO number attached to each invoice — facilitates finance team reconciliation.
- Credit Note Handling Disputed meals or shortfalls resolved via credit note — tracked against the account.
“What is the total invoice value raised for corporate clients this month?”
Event Catering Order Sheet
For catering events, a pre-event production sheet is generated from the confirmed order — exactly how many of each dish, when to prepare, and how to package and deliver.
- Event Order Entry Event date, headcount, menu selection, delivery time, and dietary requirements captured at booking. Event demand forecaster predicts quantities based on headcount, event type, time of day, and historical event data.
- Auto-Generated Production Sheet Kitchen production sheet generated from order — how many portions of each dish, preparation sequence.
- Packaging & Labelling Instructions
- Delivery Manifest Delivery checklist generated per event — driver can confirm every item before departure.
- Post-Event Waste Record Returned or unconsumed food logged — feeds the demand forecaster for future similar events.
“How much food was returned unused at corporate lunch events last quarter?”
Multi-Location Consolidated Reporting
Institutional operators often serve multiple sites. This module consolidates all site data into one management view — enabling central oversight without losing site-level detail.
- Site-Level Revenue & Cost Revenue, COGS, headcount, and waste tracked per site — and rolled up to a consolidated total. Multi-site anomaly detector flags sites with unusual cost or waste patterns — prompts investigation before the month-end report.
- Cross-Site Comparison Which sites are performing above target on cost per cover? Which are over-wasting?
- Centralised Purchasing
- Staff Deployment Overview Head counts and labour cost per site tracked centrally — identifies over-staffed or under-staffed sites.
- Client Billing Consolidation For clients with multiple sites, a single consolidated invoice generated across all locations.
“Which site had the highest cost per cover this week and what drove it?”
Demand Forecast per Institution AI
Each institution has unique demand patterns — school attendance varies by term, hospital census fluctuates, corporate offices have WFH days. AI models each institution separately.
- Institution-Specific Model Separate demand model per institution — school term dates, hospital census trends, corporate WFH patterns. Time-series forecaster with institution-specific covariates (term calendar, census data, WFH schedule). MAPE target under 7%.
- Daily Headcount Prediction Predicted headcount per meal service per institution — delivered the evening before.
- Event & Holiday Adjustment
- Production Plan Output Forecast drives the kitchen production plan — exactly how many portions per dish per service.
- Forecast vs Actual Tracking Each day's predicted vs actual headcount logged — model accuracy tracked and improved over time.
“What is the predicted headcount for Tuesday lunch at the hospital site next week?”
No-Show & Headcount Adjuster AI
Pre-orders and booking estimates never match actual attendance exactly. This model learns the gap between booked and actual headcount per institution — and adjusts production accordingly.
- Booking vs Actual Tracking Each event: booked headcount vs actual headcount logged — delta calculated and stored. Bayesian regression model trained on booked vs actual pairs per institution. Learns adjustment factor and updates it as new events occur.
- Institution-Specific Gap Model Gap varies by institution — hospital consistently 8% lower, corporate 12% lower.
- Adjusted Production Recommendation
- Seasonal Variation Learning Friday attendance at corporate offices lower in summer — model accounts for seasonal patterns.
- Gap Reduction Recommendations If no-show rate is consistently high — AI recommends revised booking confirmation or deposit.
“What is our average headcount no-show rate for corporate lunch events?”
Contract Management & Client Billing
Every contract modelled as it was actually signed — cost-plus, fixed-price, management fee or nil-subsidy — with the client invoice generated from live service data rather than assembled by hand.
- The Contract Model Is the System Cost-plus, fixed-price per head, management fee and nil-subsidy behave completely differently. The billing model is configured once from the signed agreement rather than reinterpreted each month by whoever prepares the invoice.
- Invoice Built From Service Data Meals served, headcount, waste, agreed deductions and pass-through costs assemble themselves. Most invoice disputes in this sector are arithmetic and evidence problems, not disagreements about price.
- Open-Book Where Open-Book Applies Cost-plus contracts require visible cost lines. The client sees the same figures you do, which removes the single largest source of friction in institutional catering relationships.
- Indexation and Rise-and-Fall Food and labour indexation clauses applied automatically at the agreed dates rather than remembered late and negotiated awkwardly six months afterwards.
- Margin Against the Model Actual contribution compared with the margin assumed at tender. A contract drifting below its bid model is visible in month two rather than at annual review.
“Which contracts are running below their modelled margin this quarter, and by how much?”
Subsidy, Entitlement & Free Meal Rules
Employer subsidies, student entitlements, free-school-meal eligibility and clinical diets enforced at the till — with the funding split recorded on every transaction.
- Eligibility Checked, Not Assumed Entitlement is verified against the roll or the HR feed at the point of sale. Manual lists go out of date within days and create exactly the errors an audit finds.
- The Split Recorded Per Meal Employer share, employee share, grant funding and top-up recorded on every transaction, so the funder report and the till total reconcile without a spreadsheet in between.
- Dignity by Design Free-meal entitlement is invisible at the till — the same card, the same queue, the same experience. In school and welfare settings this is not a feature detail, it is the whole point.
- Caps and Carry-Over Daily caps, weekly allowances and carry-over rules enforced automatically, including the awkward cases where a shift pattern crosses a calendar boundary.
- Reconciled to the Funder Claims produced from transaction data with the evidence attached. Under-claiming grant funding is as common in this sector as over-claiming, and both are avoidable.
“What proportion of meals were subsidised last month, and what did each funder actually pay?”
Menu Cycles, Nutrition Standards & Compliance
Multi-week menu cycles built against mandated nutrition standards — school, hospital or workplace — with compliance checked before publication rather than at inspection.
- Cycles, Not Menus Institutional catering runs three, four or six-week cycles with variety rules across them. Building a compliant single day is easy; building a compliant cycle that people do not tire of is the actual problem.
- Standards Enforced at Build Energy, protein, saturated fat, salt, sugar and portion standards checked per day and averaged across the cycle to the rule that actually applies in that setting.
- Clinical and Texture Diets Hospital and care settings require texture-modified and therapeutic diets. These are modelled as first-class menu items rather than annotations someone might miss.
- Allergen Matrix Always Current Allergens propagate from ingredients through sub-recipes into the published cycle, so the matrix given to a parent or a ward is generated rather than maintained.
- Evidence Ready for Inspection The compliance record, the recipes and the substitutions that were made on the day are retained together — because a menu that was compliant as designed is not the question an inspector asks.
“Does next term's menu cycle meet the standard on every day, and where is it marginal?”
Headcount Forecasting & Waste Reduction
Attendance predicted per site, per service and per day from roster, term calendar, occupancy and weather — so production matches the people who actually turn up.
- Attendance Is Not Headcount Payroll headcount, site occupancy and canteen attendance are three different numbers. Producing against the first is the most common cause of institutional food waste.
- Signals That Actually Matter Here Term dates, shift patterns, hybrid-working rates, ward occupancy, exam periods and weather. In a staff canteen a rainy Wednesday can move covers by fifteen per cent.
- Production Matched to the Forecast The forecast resolves to batch quantities per dish, which is the decision the kitchen makes at 07:00 — not a daily total someone has to interpret.
- Waste Costed and Attributed Over-production, plate waste and trim separated by site and service. Institutional plate waste is a nutrition problem as much as a cost one, and the two arguments reinforce each other.
- Confidence Stated Honestly A new contract with six weeks of history is forecast with visible uncertainty rather than false precision, so the chef knows when to trust the number and when to hold stock back.
“How many covers should we produce at each site on Friday, and what is the confidence?”
Multi-Site Standards & Central Kitchen Supply
One operating standard across every site, with central production, transport and site-level compliance measured the same way everywhere.
- Same Standard, Comparable Sites Cost per meal, waste percentage, compliance score and satisfaction measured identically at every site, normalised for size and format so comparison means something.
- Central Production Where It Pays Make-or-buy modelled per component. Central production reduces unit cost and raises transport cost and risk, and the answer differs by site distance and volume.
- Transport as Part of the Chain Cold-chain temperature, transfer accuracy and delivery windows tracked between central kitchen and site with the same rigour as external supply.
- Drift Detected Early A site drifting on portion, waste or compliance is flagged with the specific measure and the trend, rather than surfacing as a surprise at a quarterly review.
- Rollout Without Chaos New menus, price changes and procedures published to every site with confirmation of receipt and adoption, so ‘we never got that’ stops being a valid answer.
“Which sites are drifting from standard, and is it a training issue or a supply issue?”
Procurement, Tender & Supplier Compliance
Framework agreements, tender compliance, approved-supplier enforcement and price-list adherence — with off-contract purchasing visible rather than discovered at audit.
- Framework Prices Enforced Contracted prices applied at receipt and any variance flagged immediately. In public-sector catering, paying above framework price is a compliance failure as well as a cost.
- Approved Suppliers Only Purchases outside the approved list are blocked or escalated. Food safety, insurance and audit obligations flow through supplier approval, and a well-meant local purchase can break all three.
- Tender Evidence Assembled Volume, spend, performance and compliance data assembled for re-tender from the operating record rather than reconstructed under deadline pressure.
- Ethical and Sustainability Criteria Tracked Where a contract commits to local sourcing, welfare standards or carbon targets, adherence is measured continuously rather than asserted annually.
- Off-Contract Spend Visible Maverick purchasing is quantified by site and category, which turns a vague suspicion into a specific conversation with a specific manager.
“How much did we spend off-contract last quarter, and with which suppliers?”
Client Reporting, SLAs & Satisfaction
The reporting pack the client contract actually requires — service levels, cost per meal, participation, waste, compliance and satisfaction — produced from live data.
- SLAs Measured, Not Estimated Service times, availability, temperature compliance and complaint resolution measured against the thresholds in the contract rather than reported approximately.
- Participation Is the Real Health Metric In staff and student catering, participation rate predicts contract renewal better than any satisfaction score. It is tracked per site and per service.
- The Pack Assembles Itself Monthly and quarterly client reporting generated from operating data. Preparing these by hand consumes days and introduces exactly the inconsistencies clients notice.
- Satisfaction Captured on Site Quick feedback at the point of service, themed automatically — far more representative than an annual survey with a self-selecting sample.
- Breach Predicted, Not Reported Iris flags an SLA trending toward breach while there is still time to act. Reporting a breach accurately after the fact is not a service.
“Are we meeting every SLA this month, and which one is closest to breach?”
Integration Hub — HR, Payments, ERP & Compliance
HR and student systems for entitlement, payroll deduction, card and wallet payments, client ERP invoicing, and the accounting ledger — connected and health-checked.
- HR and Roll Feeds for Entitlement Workday, SAP SuccessFactors, school MIS and student records drive eligibility. A stale feed creates refusals at the till, which in a school setting is a serious failure.
- Payroll Deduction and Cashless Employee meal cost deducted at source, or held on a cashless card or wallet, with limits and balances enforced live at the till.
- Client ERP Invoicing Invoices delivered into the client's Oracle, SAP or Coupa in their required format, because a correctly calculated invoice in the wrong format is still an unpaid invoice.
- Tax and Public-Sector Rules VAT treatment differs for staff canteens, education and healthcare, and by whether food is subsidised. Configured per market and per contract type.
- Connection Health — Watched, Not Assumed Every connection health-checked with last-sync times, and Iris quantifies the meals at risk when an entitlement feed degrades before service starts.
“Which integration failed most often last month, and how many meals did it affect?”
Food Safety, HACCP & Audit Evidence
Temperature records, cleaning schedules, allergen controls and corrective actions captured as they happen — the evidence pack an environmental health or client audit asks for.
- Records Captured, Not Reconstructed Temperatures, cleaning, deliveries and corrective actions logged with time, person and photo at the moment they happen. Reconstructed records are the finding auditors look for hardest.
- Critical Control Points Enforced Cook, chill, hot-hold and reheat limits checked at each step, with an out-of-limit reading blocking service of that batch rather than generating a note.
- Corrective Action Closed Out Every breach carries an action, an owner and evidence of closure. An open finding at audit is worse than the breach that caused it.
- Vulnerable Settings Held Higher Hospital, care and school settings carry stricter controls and tighter tolerances, configured per contract type rather than applied uniformly.
- Any Site, Any Date, On Request The full record for a site and date produced in seconds. That capability changes the character of an audit from an investigation into a verification.
“Can we produce a complete HACCP record for any site and any date on request?”
Daypart-Only Menu Lockout
Brunch-only venues and daypart-specific concepts need menus that are strictly available only in the right window. Customers cannot order brunch items at 3pm.
- Time-Locked Menu Activation Menu items only visible and orderable during their configured time window — no manual management. Demand pattern analyser identifies optimal daypart boundary times based on when sales naturally shift — data-driven cutoff recommendations.
- Graceful Cutoff Message Customer attempting to order after cutoff sees a friendly 'Brunch service ends at 2pm' message.
- Transition Menu
- Staff Override with Authorisation Manager can override the lockout for exceptional circumstances — logged with reason.
- Multi-Daypart Configuration Morning, midday, afternoon, evening — up to 6 separate daypart menus per day configurable.
“At what time does brunch demand naturally drop off on Sundays?”
Extended Hours & 24-Hour Shift Close
Late night and 24-hour operations span multiple calendar days and require shift management, cash reconciliation, and reporting across non-standard windows.
- 24-Hour Shift Definition Shifts defined by business logic, not calendar day — a shift from 10pm to 6am is one shift, not two days. Late-night demand forecaster identifies which nights expect high traffic — guides staffing decisions for overnight shifts.
- Multi-Staff Handover Shift handover transfers the open till, any open tabs, and pending orders to the incoming staff member.
- Overnight Cash Reconciliation
- Late Night Sales Report Sales by hour through the night — identifies peak windows in the 1am–4am period.
- Fatigue-Aware Scheduling Integration HR module ensures staff are not scheduled for back-to-back overnight and morning shifts.
“What is our busiest hour between midnight and 4am and which nights of the week?”
Event & Ticket-Based Billing
Supper clubs, themed dinners, and pop-up experiences sell tickets rather than individual items. This module manages ticket sales, seat allocation, and event-night billing in one system.
- Ticket Sale & Inventory Event tickets sold online or at the counter — total capacity tracked, sold out enforced. Event demand forecaster predicts add-on spend per guest based on ticket tier and event type — staffing and stock planned accordingly.
- Seat / Table Allocation Each ticket linked to a seat or table — floor plan shows booking status per event.
- Ticket Price Tiers
- Add-On Sales at Event Drinks, experiences, or upgrades sold on the night linked to the guest's ticket.
- Post-Event Revenue Report Total ticket revenue, add-on revenue, and total per-head spend — per event.
“What was the average per-head spend at our last supper club event?”
Pop-Up Location & Event Tagging
Pop-up operators move between locations and events — each session tagged with its location, event, and date so revenue and performance can be compared across all trading locations.
- Session Location Tagging Before each session, operator selects or creates the location — all sales tagged to that session. Location scoring model ranks venues by revenue per trading hour — adjusted for day of week and time of year.
- Event Type Classification Farmers market, food festival, corporate event, private party — each session classified.
- Location Performance Comparison
- Best Location Recommender AI ranks locations by historical revenue per hour of trading — guides booking decisions.
- Session Summary Report End-of-session summary: total revenue, top items, covers, and comparison to the same location last time.
“Which market location generates the highest revenue per hour of trading?”
Offline Mode — Full POS Without Internet
Food halls, markets, and pop-up events often have unreliable connectivity. Zentallio operates fully offline — takes orders, processes payments, and syncs everything when connectivity returns.
- Full Offline Functionality Orders, payments, and inventory depletion all work without internet — no functionality degraded. No ML model for offline mode — deterministic sync logic. Post-sync anomaly detection checks for unusual patterns in offline transactions.
- Card Payment in Offline Mode Card transactions queued offline and processed when connectivity is restored — within accepted risk parameters.
- Sync on Reconnect
- Sync Conflict Resolution If a stock item sold out in offline mode, sync resolves the conflict and flags the item as 86'd.
- Offline Session Log Duration of offline period and transactions completed are logged — full audit trail maintained.
“How many transactions were processed in offline mode at the last food festival?”
Concept Performance Scoring AI
Themed restaurants and niche concepts need to know if the concept is working — not just if the food is selling. This module scores concept performance across experience, revenue, and retention metrics.
- Experience vs Revenue Balance Revenue per cover balanced against repeat visit rate — a concept that earns but never retains guests is failing. Multi-metric concept scorer trained on revenue, repeat rate, and sentiment signals. Trend model identifies concept fatigue before revenue falls sharply.
- Theme Night vs Regular Night Comparison Does the themed menu outperform the regular menu? Revenue, covers, and margin compared.
- Social Sentiment Correlation
- Concept Fatigue Detection If themed night revenue declines over consecutive events — concept fatigue flagged, refresh recommended.
- Concept Scorecard Monthly scorecard: revenue, retention, social sentiment, margin — is the concept viable?
“Has our Friday themed dinner concept shown declining attendance over the last 8 weeks?”
Theme Night Revenue Analyser AI
Themed events — Eid specials, Valentine's menus, cultural nights — need post-event analysis to determine if the investment in theming paid off in revenue and margin.
- Theme Night vs Baseline Comparison Theme night revenue vs the same day of week with a regular menu — net uplift calculated. Attribution model decomposes theme night revenue into base demand, theme uplift, and special item contribution.
- Cost of Theming Decorations, special ingredients, entertainment costs — entered and deducted from theme night margin.
- Menu Item Performance
- Advance Booking Conversion Did theming drive more advance reservations? Booking rate on theme nights vs regular nights.
- Theme ROI Report Revenue uplift minus theming cost = theme ROI. Reported per event.
“What was the revenue uplift from our Valentine's dinner versus a regular Friday?”
Ticketed Seatings & Prepaid Covers
When the meal is the ticket — fixed seatings, tiered pricing, prepayment and a guest list that becomes the service plan.
- The Ticket Is the Cover Prepaid seatings turn a restaurant into an event business. Covers are known days ahead, which makes production, staffing and ordering near-exact rather than forecast.
- Tiers Priced Deliberately Early release, standard, premium seat, chef's counter and paired-drinks upgrade. Tiering is where experience concepts earn their margin, and it has to be designed rather than improvised.
- Release Timing Modelled Iris models how a seating sells across the window and flags one selling too slowly while there is still time to promote, or too fast — which usually means it was underpriced.
- The Guest List Is the Brief Allergies, occasions, dietary requirements and returning guests are compiled into the service plan the kitchen and floor actually work from.
- Refunds and Transfers Handled Clear transfer and resale rules reduce refund pressure and keep seats filled. A rigid no-refund policy on a $90 ticket costs more in reputation than it protects in revenue.
“What is our realised yield per seat by seating and tier, and where did we leave money on the table?”
Pop-Up, Residency & Site P&L
Every temporary location, residency and festival pitch with its own P&L, stock, staffing and payback — opened in days and closed cleanly.
- A Site That Exists for Six Weeks Temporary locations still need stock, staffing, compliance and a P&L. Running them out of a notebook is why so many pop-ups feel busy and finish flat.
- Payback Modelled Before Opening Fit-out, pitch fee, transport and staffing against forecast covers gives a break-even day. A residency that needs to be exceptional to break even should be renegotiated rather than attempted.
- Stock Moves With the Concept Transfers between the main kitchen and the pop-up tracked like external supply, so what came back is reconciled with what went out.
- Close-Down Without Loose Ends Final stock count, equipment return, staff reconciliation and site P&L closed as one process rather than trailing for weeks afterwards.
- What the Successful Ones Shared Iris compares completed sites on footfall, ticket price, format and weather to identify the pattern behind the ones that worked — which is the only reliable guide to picking the next.
“Which pop-up sites paid back, and what did the ones that failed have in common?”
Offline-First Trading & Sync
Full trading with no connectivity — festival fields, basements, moving trucks — with everything reconciled exactly when the connection returns.
- Trading Continues Regardless Orders, payments, discounts and stock all work offline. A concept that trades in a field or a basement cannot have connectivity as a dependency.
- Card Payments Held Safely Offline card handling within agreed floor limits and risk rules, so a lost signal does not become a lost sale or an unrecoverable transaction.
- Conflict Resolution That Is Honest When two tills traded the last portion offline, the platform surfaces the conflict rather than silently picking one. Quiet resolution is how stock records lose credibility.
- Sync Verified, Not Assumed Every offline transaction is confirmed as synced with a count reconciled both ways. ‘It should have uploaded’ is not an acceptable position on revenue.
- Duration and Cause Logged Offline periods recorded with duration and cause, which turns an intermittent connectivity problem into evidence for the venue or the provider.
“Which sites traded offline last month, for how long, and did every transaction sync?”
Experience Design & Run Sheet
The show itself — course timings, service beats, lighting, sound and staff cues — sequenced so the room lands the same way every night.
- The Run Sheet Is the Product In experience dining the sequence is the product. Course five landing eleven minutes late does not just delay dinner, it breaks the story the room was told.
- Beats, Not Just Courses Welcome, first pour, opening beat, courses, interludes, finale. Each carries a target and a live variance so drift is visible at beat three rather than discovered at the finale.
- Cues Where People Need Them Kitchen, floor, bar and technical cues delivered to the right person at the right moment. A shared paper run sheet works until one person is not holding it.
- Repeatable Without Being Rigid The sheet adapts to a slower table without collapsing the sequence for everyone. Rigidity in an experience concept reads as impatience to the guest.
- Reviewed Against the Night Planned against actual for every beat, so the sequence improves across a run rather than depending on who was on.
“Which seatings ran behind the run sheet, and at which beat did the drift start?”
Theme Nights, Collaborations & Limited Runs
One-off nights, guest chefs and limited runs planned, costed and measured properly — including what they did for the weeks either side.
- Costed Before Committing Guest fee, ingredients, marketing, extra staffing and the covers displaced from the normal offer. Collaborations frequently sell out and lose money, which is the worst combination.
- The Halo Is Real and Measurable A strong collaboration lifts the following weeks. Measured against a matched baseline rather than claimed, because the halo is often the entire return.
- Cannibalisation Counted Too If a theme night draws the same guests who would have come anyway on Friday, it moved revenue rather than created it.
- New Guests Versus Regulars The split matters more than the covers. A sold-out night of existing regulars is a lovely evening and a poor acquisition channel.
- What to Repeat Iris ranks completed runs on contribution, new-guest acquisition and halo, so the decision to repeat is evidence rather than enthusiasm.
“Did the guest-chef collaboration pay for itself once the halo and cannibalisation are both counted?”
Merchandise, Retail & Bookable Extras
Aprons, cookbooks, sauces, gift vouchers, masterclasses and add-ons sold alongside the experience — with stock and margin tracked properly rather than as an afterthought.
- A Second Revenue Line Merchandise and bookable extras often carry higher margin than the food and require no kitchen capacity. In experience concepts they are frequently under-built.
- Sold Where the Feeling Is The best moment to sell the cookbook is at the end of the meal, not on a website three days later. Offered at the table, added to the same bill.
- Gift Vouchers as Acquisition Vouchers bring new guests who were chosen by someone who already trusts you. Tracked as an acquisition channel with its own redemption and conversion rate.
- Stock Without a Stockroom Merchandise stock, reorder points and supplier lead times handled in the same ledger as food, so a sold-out cookbook is visible before the guest asks.
- Masterclasses and Add-Ons Classes, tastings and behind-the-pass extras booked and capacity-managed alongside seatings rather than in a separate spreadsheet.
“What is our attachment rate on merchandise per seating, and which item carries the best margin?”
Community, Waitlist & Membership
A following rather than a customer base — memberships, priority access, waitlists and the drop mechanics that make a release feel worth showing up for.
- A Following, Not a Database Experience concepts live on people who want to be there. Membership, priority windows and early access reward that rather than treating every booker as a stranger.
- Waitlist That Converts When a seat frees, the waitlist is worked in order with a real hold window. A waitlist nobody actions is just a list of people you disappointed.
- Drops Without the Chaos Timed releases, queue handling and per-member limits, so a release feels exciting rather than broken — and the bots do not take the room.
- Lapsed Members Noticed A member who has missed three releases is a relationship worth a message. Iris flags them with what they used to book rather than a generic re-engagement blast.
- The Room Composed Deliberately Where the concept depends on the mix in the room, seating can balance regulars and newcomers rather than filling first-come and hoping.
“Which members have not attended in three releases, and what did they used to book?”
Mobile, Festival & Multi-Pitch Operations
Trucks, stalls and festival pitches with route planning, pitch fees, generator and water dependencies, and daily cash reconciliation per site.
- Every Pitch Is a Site Pitch fee, travel, fuel, staffing and expected footfall give a break-even before the booking is accepted. Many event pitches lose money in a way that feels busy.
- Weather Decides the Day Outdoor trading is weather-bound more than any other format. Prep and staffing follow the forecast rather than the booking, because unsold prep at a festival is total loss.
- Dependencies Tracked Power, water, waste and licence conditions per pitch, checked before departure. Arriving without the right connection is a whole day gone.
- Cash Reconciled Daily, Per Site High-cash environments with temporary staff need same-day reconciliation per pitch rather than a shoebox emptied at the end of a festival.
- Event History That Compounds Every event scored on contribution and footfall accuracy, so next season's calendar is built from evidence rather than from which organisers called first.
“Which pitches and events actually paid, once fees, travel and staffing are counted?”
Integration Hub — Ticketing, Payments & Tax
Ticketing platforms, payment processors, offline card handling, event and pitch fees, local tax rules and the accounting ledger — connected and health-checked.
- Ticketing Platforms Connected Eventbrite, Tock, Resy and direct sales writing into one guest list, so a seating is never oversold across two systems on release night.
- Payments Including Offline Stripe, Adyen, Square and SumUp, with offline card handling for field and basement trading and reconciliation when the connection returns.
- Deposits, Balances and Vouchers Prepayment, deposits, balance-on-arrival and gift vouchers handled as distinct liabilities rather than blended into revenue on the day they are taken.
- Tax on Tickets and Food Ticketed dining, events and merchandise carry different treatment in most markets. Configured per market and per revenue type.
- Health-Checked Before a Release A payment or ticketing outage during a timed release costs an entire seating. Connections are verified ahead of a release rather than discovered mid-drop.
“Which integration failed most often last month, and did any seating lose bookings because of it?”
Solutions
Balanced Scorecard
Financial, customer, operational and AI-learning KPIs on one live board.
- Financial · Customer · Ops · AI
- Threshold beats & recovery plays
- Per-outlet & group roll-up
- Iris narration
- Drill to the source metric
Order Taking Application
A point of sale built for how restaurants actually run.
- Orders, modifiers & every channel
- Split, tip & card·cash·wallet·QR
- Iris cross-sell at the till
- Offline-first with NexNode
- Posts every sale to the CFO ledger
Drive-Through
Purpose-built for the lane: dual-order points, car-ahead sequencing and live timers on every stage.
- Dual-lane order & pay
- Stage-by-stage lane timers
- Car-ahead sequencing
- Predicted-demand pre-stage
- Headset & display integration
Loyalty & CRM
Enroll members at the till, earn and redeem points, and let Iris target the offer that actually returns a guest.
- Tap-to-enroll at the till
- Points, tiers & redemptions
- RFM segmentation
- Targeted win-back offers
- Best-of discount logic
Promotions & Pricing
Run promotions, bank-card offers and vouchers with a hard margin floor Iris enforces at the till.
- Promotions & voucher engine
- Bank-card offers by BIN
- Manager-approval thresholds
- 45% margin-floor guardrail
- What-if price modelling
Supply Chain
Demand-planned replenishment across every outlet and distribution centre.
- Demand forecasting by outlet
- Auto-replenishment proposals
- DC & outlet transfers
- Lead-time & fill-rate tracking
- Weather-aware planning
Procurement
Supplier catalogues, purchase orders and multi-level approvals in one flow, with Iris benchmarking every price line against what you paid last and what others pay.
- Supplier catalogues & contracts
- PO raise & approval flow
- Price-line benchmarking
- 3-way match to receipts
- Spend by category & vendor
Production / Kitchen
Prep planning, kitchen-display routing and batch/yield control.
- Forecast-driven prep lists
- KDS routing & bump
- Batch & yield control
- Recipe & sub-recipe costing
- Waste at source capture
Stock & Inventory
Live stock on hand with theoretical-vs-actual variance and expiry tracking.
- Live on-hand by location
- Theoretical vs actual variance
- Expiry & markdown alerts
- Cycle counts & recounts
- Shrink root-cause
CFO Workbench · Numerus
The finance cockpit: trial balance, P&L, balance sheet, cash flow and sub-ledgers that tie out to the penny.
- Trial balance & GL
- P&L, balance sheet, cash flow
- AR / AP / inventory sub-ledgers
- Live posting from the till
- Ask-Zen over the ledger
Operation Management
Digital checklists, shift routines and standard operating procedures across every outlet.
- Opening / closing checklists
- SOP routines & tasks
- Live completion tracking
- Exception escalation
- Multi-outlet roll-up
Ask Iris
The natural-language brain over your whole operation.
- Plain-language questions
- Answers from live data
- Works over every module
- Voice & text
- Suggested next questions
Zen AI · Computer Vision
Computer vision that reads what a camera sees.
- Queue & wait detection
- Portion & plate consistency
- Cleanliness & compliance
- Footfall & dwell heatmaps
- Anonymised, on-prem capable
Reporting & Analytics
Every number, every outlet — scheduled reports, ad-hoc exploration and boardroom roll-ups.
- Self-serve dashboards
- Scheduled report delivery
- Group & outlet roll-ups
- Export & e-invoice packs
- Natural-language queries