Know what every plate costs you. Food cost is about a third of your revenue and most independent kitchens are guessing at it — we build the systems that tell you the real number, per dish, updated every time an invoice lands.
Three things that come up in a lot of kitchens.
Which was a reasonable guess two years ago. Since then your protein, your oil and your packaging have all moved, and nothing on the menu moved with them. We cost every dish off your real invoices so the price is a decision, not a hand-me-down.
You find out at year-end, or when the accountant asks why food cost is up four points. Every invoice you upload re-costs your whole menu, and anything that drops below target margin gets flagged the same day.
So you over-order what you can see and forget what's behind it. We turn your actual sales into a draft order twice a week — you review it, approve it, and it goes to the vendor.
Built on the data you already generate — your POS sales, your supplier invoices and your deliveries — plus the weather and what’s on in town. Built for your kitchen, no per-seat pricing, no lock-in.
The number almost nobody has: what each dish on your menu actually costs to make, today, at today's prices. Most kitchens never work it out because entering every recipe and every ingredient is a week of evenings. We do that part:
$499 setup, then $99/month Setup includes building your recipes from whatever you've got — binder, spreadsheet or the back of a napkin.
We start from whatever you have — a binder, a spreadsheet, photos of recipe cards, or just the menu. Each dish is broken into its ingredients with quantities and yields, and every ingredient is matched to what you actually buy from your suppliers, including pack sizes and unit conversions. That mapping is the fiddly part and it's the part we do for you. From then on, each invoice you upload updates the ingredient prices and every dish that uses them, so the costing never goes stale. Everything runs on a server we operate for you, and the numbers are yours to export whenever you want.
Illustrative example of the per-dish costing we build.
Once every dish is costed, your POS sales tell you the rest. We rank the whole menu by margin and popularity, so you can see which dishes are carrying the restaurant and which ones are quietly costing you money every time they're ordered:
Included with Plate Cost
We pull your per-item sales from your POS — Toast, Square, Clover or a scheduled export — and join it to the plate costing. That gives every dish two numbers that matter: how much money it makes per order, and how often it's ordered. Sorting on those two together is what surfaces the dishes worth acting on. A cheap dish with a good percentage but four orders a week isn't your problem; a popular one at 44% food cost is. We show it as a ranked list rather than a chart you have to interpret, and you can filter by menu section or daypart.
Illustrative example of the menu ranking we produce.
Your sales say you should have used 40 kilos of chicken this week. Your invoices and counts say you used 47. That gap is waste, over-portioning, or something walking out the back door — and most independent kitchens have never seen the number:
From $1,500 setup, then $249/month Needs a weekly count. We keep it to the twenty items that actually move the number.
Every dish you sell depletes its ingredients — that's the theoretical usage, calculated straight from your POS sales and the recipes we already built. Actual usage comes from your opening count, your deliveries, and your closing count. The difference between the two is your variance, and we cost it so it reads in dollars. We deliberately keep the counted list short: twenty or so high-value, high-movement items produce almost all of the signal, and asking a kitchen to count four hundred SKUs every week is how these systems get abandoned by the second month.
Illustrative example of the variance report we produce.
Ordering is a forecasting problem, and forecasting is something software is genuinely good at. We turn your sales history into a draft order twice a week — you look at it, change what you want, and approve:
From $2,500 setup, then $349/month Includes the vendor setup and a pilot period where we check every order before it sends.
We forecast demand per menu item from your sales history, allowing for day of week and seasonality, then run it back through your recipes to get ingredient-level demand for the ordering period. Subtract what you have on hand and what's already inbound, round to the pack sizes your vendor actually sells, and that's your draft order. A manager reviews it in the browser twice a week and approves, which emails the purchase order to the vendor. We also gate it on forecast accuracy — if the model isn't predicting a category well enough to trust, it says so and asks a human rather than quietly ordering the wrong amount. The forecast also takes in local weather and the events calendar for your area, because a patio in a heat wave and a Friday with a home game are not an average Friday. Each vendor keeps its own catalogue and its own lead time, so the draft splits into one order per supplier and the buying window is measured from when that supplier actually delivers rather than from a single assumed lead time.
Illustrative example of the draft orders we generate.
A weekly count is a photograph. This is the video. Every delivery puts stock in and every sale pulls its recipe back out, so the number on the screen is what is actually in the building right now:
Included with Theoretical vs Actual Runs on any tablet or phone the kitchen already owns — there is no hardware to buy.
Receiving runs in the browser on whatever tablet or phone is already in the kitchen, laid out for the cooler rather than for a desk — targets big enough to hit with a glove on, and nothing that needs typing. When a delivery lands, the order we sent is already on screen, so receiving is confirming quantities rather than transcribing them; anything short, substituted or repriced is flagged there and then, and the new price flows straight into the plate costing. Between deliveries the count moves on its own, because each sale depletes its recipe from live stock — the same theoretical usage the variance report is built on. That turns your weekly count into a check on a number that already exists, rather than the only moment anyone knows what is in the room. Use-by dates are captured at receiving, and anything inside its window appears on a short list the kitchen can build specials from.
Illustrative example of the receiving screen we build.
Everything above tells you what a plate costs. This is the other half of the room — taking the order, getting it to the right station, and getting paid. It runs on the same recipes and the same stock as the costing, so the floor and the back of house are never working from two different sets of numbers.
The guest scans the code on the table and orders from the phone in their hand. Nothing to download, nothing to sign up for:
Quoted to your business Built around your menu and your floor plan, so we scope it per room rather than per seat.
The guest menu is generated from the same recipe and pricing data the costing runs on, which is the part most setups get wrong — a QR menu bought separately from your costing tool means two menus to maintain and one of them is always out of date. Each table gets its own code tied to a table record, so an order is attached to a table rather than to a session that can be lost when a phone locks. Modifiers and allergen flags are part of the item definition, so they are carried to the kitchen ticket rather than retyped. Service state is authoritative on our side, not the guest's: if a server closes or locks the table, the guest's next tap is refused with an explanation rather than silently accepted. It is a mobile web page, so it works on whatever the guest is carrying and there is nothing to install on your side either.
Illustrative example of the guest ordering screen we build.
Orders arrive at the station that actually cooks them, in the order they ought to be fired — rather than on one printer by the pass that everybody has to crowd around:
Quoted to your business Runs on tablets or screens you already own. Station layout is scoped to your line.
Every menu item is mapped to the station that produces it when we build the recipes, so routing is a property of the dish rather than something a server chooses under pressure. A ticket is split across stations and each screen sees its own portion, with the table and course kept visible so the line knows what it is part of. Pacing works off course numbers on the order: starters fire on receipt, mains hold until the starter course is bumped or a manager fires them early. Bumping is the event that depletes stock, which keeps the live count honest — food is deducted when it is actually made, not when it is ordered and possibly voided. The printer fallback is deliberate: screens and networks fail during exactly the services you can least afford it, so a station that loses its connection starts printing rather than going quiet.
Illustrative example of the station screen we build.
One screen that answers the only question a busy floor actually asks: what needs doing right now, and by whom:
Quoted to your business Scoped to your room — sections, patios and bar are set up to match how you actually run service.
The board is a view over the same order and ticket records the kitchen works from, not a second system kept in sync with the first — which is the usual reason a floor screen and a kitchen screen end up telling two different stories. Each table carries its seated time, cover count, course state and outstanding balance, and the alerts are thresholds on those values rather than anything anyone has to remember to raise. We set the thresholds with you during the pilot, because "too long" means something different at a forty-minute lunch than it does on a tasting menu. Sections are assigned per shift so the same board serves the whole team without anyone reading a room they aren't working. It is a browser page, so it runs on a terminal by the pass, a tablet behind the bar, or a phone in an apron.
Illustrative example of the floor board we build.
The guest pays from the same phone they ordered on, whenever they're ready, without anyone walking a terminal across the room twice:
Quoted to your business Stripe's processing fees are charged by Stripe at their standard rates, separately from anything we charge. You hold the Stripe account and the money lands in it directly.
Payments run through Stripe against an account in your name, so settlement goes directly to your bank and we are never in the flow of your money. Card entry happens inside Stripe's own hosted fields, which means card data never reaches our systems and your PCI obligations stay at the lightest tier rather than expanding to cover a server we run. A payment is tied to the table's open bill, so a split settles against the same record the floor board and the kitchen are reading, and the table closes the moment the last portion clears. Tips are captured at checkout and reported per server for payout. Refunds and voids are manager-gated and written to the audit trail with who did it and why, because that is the record you will want the week somebody queries a chargeback.
Illustrative example of the guest checkout we build.
Smaller pieces that bolt onto any of the above.
Your supplier raises the price of chicken and you get a message the same day telling you which dishes just dropped below target margin, and what to reprice them to.
A text every morning with yesterday's sales, food cost and labour percentage. Thirty seconds with your coffee instead of waiting for month-end.
Running more than one room? The same numbers side by side, so you can see which location is running three points hotter and go find out why.
An AI phone agent for takeout, reservations and catering enquiries, automatic review requests after a visit, and win-back messages to regulars who've stopped coming in.