I build systems for restaurants, cafés and cloud kitchens: kitchen order tickets and billing, recipe-costed inventory, aggregator reconciliation, purchase control and multi-outlet dashboards.
Where does a restaurant actually lose money?
Not at the till. The till is the one thing everyone watches. The losses sit in four places nobody has time to look at:
Recipes are uncosted, so menu prices are informed guesses and the dish everybody orders may be the one earning least. Stock consumed is never reconciled against stock sold, which means wastage, over-portioning and theft are indistinguishable — they arrive as one unexplained monthly gap. Aggregator commissions and refunds are accepted as billed, because verifying nine hundred line items is nobody’s evening. And purchases are negotiated fresh each time from memory, so the rate you paid last month is not the rate you are paying now.
Food cost creeping two points is a serious monthly number at any real volume. Chhota leak, bada bill.
What does the system cover?
- KOT and billing. Table, takeaway and delivery in one flow, GST-correct, fast enough for a Friday rush.
- Recipe costing. Every dish a bill of materials, so margin per dish is known rather than assumed — and menu engineering becomes possible.
- Stock reconciliation. Theoretical consumption from sales against actual counts, with variance named daily. Wastage and portioning drift become separable problems.
- Aggregator reconciliation. Payouts against orders against commissions against refunds, per platform, with mismatches flagged. Talabat and Deliveroo for Gulf outlets follow the same discipline.
- Purchase control. Rate history per supplier, standing quantities, invoice-versus-goods-received matching — the buying discipline that survives the buyer’s day off.
- Outlet dashboards. Sales, food cost and variance per branch, comparable at a glance.
Why does daily variance matter more than monthly?
Because a monthly variance number tells you something went wrong and nothing about what. Thirty days is enough time for four different causes to overlap, and by the time you see the total, the evidence has been eaten, thrown away or forgotten.
Daily variance on your top items is a different instrument entirely. A gap that appears on Tuesdays points at a shift. A gap on one item points at portioning. A gap that starts the week a new supplier began points at the supplier. Same measurement, taken often enough to be diagnostic rather than merely upsetting — which is exactly the reconciliation argument I make about money, applied to paneer.
The same logic explains why purchase rate history matters more than a good negotiation. A rate agreed in March and drifted by June costs you every single day in between, silently, on an item you order weekly — and nobody notices because no single invoice looks wrong.
Related reading
/industries/catering — the event-driven sibling · /industries/hospitality — hotels with F&B attached · /products/retail-erp — the multi-outlet stock core · /blog/reconciliation-at-scale — why daily beats monthly
Tonight’s test
Take your top five items. Compare what the sales say you should have used against what the store says you actually used. If you cannot do that comparison tonight, that gap has a rupee value and it is running every single day. /contact.
Questions I actually get
Cloud kitchen running three brands from one kitchen?
That is the design case. Brand-wise profit and loss on shared inventory, with consumption apportioned by recipe rather than guessed by revenue share. Most multi-brand operators genuinely do not know which brand is subsidising which, and the answer is often the opposite of what everyone assumes.
Does it reconcile aggregator payouts?
Yes — orders against commissions against refunds against the actual payout, per platform, per cycle, with mismatches queued. Nobody has time to verify nine hundred line items by hand, which is exactly why the deductions go unchecked and why an automated match pays for itself quickly.
Do you integrate with Zomato, Swiggy or Talabat APIs?
Where the platform exposes them, yes; otherwise file-based reconciliation on their settlement reports. The discipline matters more than the pipe — a matched settlement file catches the same errors as an API, just a day later.
My chef guards his recipes.
Costing needs quantities, not secrets. Grammage for the costed items goes in; the masala ratios and the technique stay with the chef. I have never needed a recipe to cost a dish, and no system should require one.
How does stock variance actually get measured?
Theoretical consumption from sales, compared against actual counts. The gap has a rupee value and a direction, and it separates wastage from portioning drift from theft — three very different problems that look identical in a monthly total.
Multiple outlets with different menus?
Per-outlet menus and pricing with consolidated reporting: food cost, variance and margin per branch, comparable side by side. The comparison is usually the most useful screen an owner gets, because it makes an outlier obvious.