Hokoku, a restaurant platform that knows what every menu item costs this week

The engagement

He runs two quick-service restaurants in Mississippi, and he priced the menu the way the founder taught him: case cost, per-ounce cost, recipe cost, food-cost percentage, worked out by hand whenever he found the time. The software he could buy didn’t do that math. I visited the store, photographed the forms he was working from, and built for about a week. The next time we met I opened a working version running on his own numbers, rather than a plan for one.

It starts from the vendor invoices. Every ingredient cost rolls up through the recipe book, batch recipes like cooked taco meat included, by yield, into what each menu item costs to make this week against what it sells for. Around that core sit the rest of the pieces of his week that used to live on paper: daily sales from the POS, the weekly inventory walk, and the owner’s weekly scorecard.

What’s running

On Monday he opens one brief and sees which three or four menu items drifted out of target, and what to charge instead. Changing a price is still his call. I go back to the store and sit with the people using it, which is where most of the next round of work comes from.

Try it

There’s a public demo with three fictional restaurants and a full operating week of data; it resets every night. To open it, enter an email address and type in the six-digit code that arrives (type the code rather than tapping the link in the email, which some mail filters use up first). The email you enter is the only thing I see.

Screens

Public demo, fictional data. Three made-up restaurants and a full operating week. Click any screen.

Hokoku public demo, owner view: the completed-week decision brief for a fictional taqueria — data readiness checks, food cost against target, and ranked price actions — Owner view: the completed-week brief for a fictional taqueria. What to fix and what to price next.Public demo · fictional data

The Monday owner brief: top price actions, food cost versus target, menu items outside the target band, and ingredient price moves, written in sentences — Every Monday this writes itself. The AI writes the sentences; the numbers are checked against the week's figures first.Public demo · fictional data

Price review: menu items out of band this week with cost, price, food-cost percentage, target, and a suggested in-band price; ingredient price movers; longest-untouched prices — The weekly price check, running continuously. The owner decides which suggestions to take.Public demo · fictional data

Invoices screen: vendor invoices that arrived by email, staged for review before any price applies — Vendor invoices land here from email. Small moves apply themselves; the big ones wait for the owner.Public demo · fictional data

Inventory review: what should have left the shelf according to sales versus what the count says, with the variance shown per item — What the sales say should have left the shelf, against what the count says did. The gap is shrink.Public demo · fictional data

Trends: food cost and shrink over several weeks, as charts — The same gap over time, without extra counting.Public demo · fictional data

What's running

Monday food-cost review

Cadence: Every Monday morning
What it does: Lists menu items whose food cost drifted out of the owner's target, with suggested price changes.
Guardrail: Suggestions only. Nothing applies a price change on its own.

Invoice pull

Cadence: Every day
What it does: Picks up vendor invoices from email and the distributor's portal, reads the line items, and updates ingredient costs.
Guardrail: Small price moves apply themselves. Large ones are staged for the owner to approve.

POS sales pull

Cadence: Every morning
What it does: Pulls the previous day's sales export from the POS so the week's numbers are already in when the brief runs.
Guardrail: Read-only against the POS. Nothing is written back.

Hokoku, a restaurant platform that knows what every menu item costs this week.