The kitchen sold eighty biryanis and used rice for ninety-six

Say the POS shows eighty plates of biryani on Sunday. The recipe card says a plate takes 180 grams of rice, so that is 14.4 kilos. Then somebody opens the store register and finds the kitchen was issued enough rice for about ninety-six plates. Nobody stole anything you can prove. Nobody will admit to anything. And you will not ask the question until the second week of the following month, when your accountant hands you a food cost percentage three points worse than last month and offers no explanation beyond rates having gone up.

That is a timing problem before it is an honesty problem. Sales are watched hourly; every owner knows roughly what the day did before the shutter comes down. Consumption is watched monthly, in aggregate, from purchase bills. One number is fresh, the other stale, and the gap between them surfaces long after anyone can act on it.

That gap has a name. Theoretical food cost is what your recipes say the day's sales should have consumed. Actual food cost is what the kitchen used. The difference is variance. Rising rates are a purchasing problem you negotiate once a quarter. Variance is an operating problem you can attack this week, for free, if you can see it at all.

Two numbers: what you sold, and what you consumed

Everything here rests on producing two figures for the same outlet over the same period. Most restaurants produce one cleanly and the other not at all.

• Theoretical consumption = item-wise sales quantity from the POS times the recipe quantity for each item, ingredient by ingredient.
• Actual consumption = opening stock, plus everything received into the store during the period, minus closing stock. Counted, not estimated.
• Variance = actual minus theoretical, in quantity and in rupees at your current purchase rate. Rupees are what you act on; quantity tells you the cause.

Note what is not on that list: purchase spend. Buying a fifty-kilo bag of rice on the 3rd does not mean you ate fifty kilos that month, and a month where you stocked up before a festival will look catastrophic. Almost every restaurant that says it tracks food cost is really tracking spend divided by sales, which moves for reasons that have nothing to do with the kitchen.

The recipe is a measurement tool, not a chef's document

Ask a cook who has run your kitchen for nine years to write down grams and you will get resistance, and the resistance is not unreasonable. Andaaz is real; a good cook adjusts for the day's tomatoes. But a recipe card in a system is not an instruction to the cook and not a challenge to his judgement. It is the only instrument you own for converting a sale into an expected consumption. It is also his defence, because he is the one who gets blamed when the food cost moves and nobody can say why.

• Per-portion quantities in grams and millilitres for every ingredient that matters in rupees, not for every ingredient that exists.
• Sub-recipes costed once and reused: the base gravy, the biryani masala, the marinade. A gravy that feeds nine dishes should be one recipe, not nine copies that drift apart.
• Yield stated honestly. A kilo of raw chicken is not a kilo of usable meat. A card written on purchased weight while the cook works from cleaned weight is wrong forever, and wrong in the same direction.
• Half-plates, family packs and combos as their own recipes, because they sell as their own items.
• Modifiers accounted for. Extra cheese, extra butter, no onion. If the guest can order it, the recipe has to know about it.
• A re-cost trigger, so that when the rate of oil or paneer moves, the affected dishes get repriced rather than quietly losing money.

The modifier point catches people out. If one menu runs across dine-in, takeaway and delivery, with modifiers and combos rolling into a single consolidated order, the order data already carries the extra cheese. The recipe has to carry it too, or you are comparing a full-plate theoretical against a week that sold two hundred half-plates.

Where the variance comes from: portion, waste, theft, yield

Variance is not a single thing, and the mistake most owners make is jumping to the last item on this list first. There are five real sources, and they show up differently in the numbers.

• Portion drift. The serving spoon is not the spoon the recipe was written for, and the regular who always gets a little extra gets it four hundred times a year. Usually the biggest contributor and the easiest to fix, because a scoop and a weighing scale fix it, not a confrontation.
• Yield loss. Trim, bone, peel, evaporation, the last of the gravy that will not leave the pot. Unavoidable, and the reason an unstated yield makes a recipe a lie.
• Waste. Burnt, dropped, over-prepped for a rush that did not come, gone off in the chiller during a slow week. The most recoverable variance, because it points at a prep decision somebody makes every morning at nine.
• Unrecorded consumption. Staff meals from the same sack, the owner's family on Sunday, tasting portions for a new dish. None of it is theft and all of it looks exactly like theft in the numbers.
• Actual pilferage. Last on the list, not first. It exists, it is usually smaller than the other four, and it is the only one a process change will not solve.

Keep that order as an investigation discipline: recipe first, yield second, prep and waste third, unrecorded consumption fourth. Accusing a cook of theft when the answer was a card written on uncleaned weight costs you the cook and does not fix the number.

A food cost percentage tells you that you are bleeding. Item-wise variance tells you which finger to look at.

Daily issues from the store, not a weekly guess

The structural change that makes all of this possible is separating the store from the kitchen. In most small outlets they are the same shelf and the same person, so consumption is never a recorded event, only an inference drawn at month end from what has vanished. A daily issue does not require a warehouse. It requires a rule.

• The kitchen asks for what it needs for the shift, by item and quantity, before service.
• The store issues it and records it. One line per item, thirty seconds a line.
• Anything drawn mid-service gets a second issue, not a mental note. Mid-service top-ups are where the register usually breaks.
• Returns to the store at close are recorded too, or every top-up looks like consumption.
• Supplier deliveries are received against the item, so tomorrow's opening stock is a number rather than an argument.

Once raw-material stock and recipe consumption sit in one place, the issue and the expected depletion can be read against each other, and low-stock alerts fire before you are short rather than during the Saturday rush. Most weeks that is worth more than the variance report itself.

Count the ten things that matter, not the four hundred

Owners abandon inventory control because they start with a full physical count of every SKU in the building. That count takes two hours, gets done badly by a tired steward at midnight, and stops within a fortnight. It is not needed. Almost all the rupee value of your variance sits in a handful of lines.

• The proteins: chicken, mutton, fish, prawn, paneer. Highest rate per kilo, highest portion sensitivity, highest theft appeal.
• Cooking oil, the item most likely to be over-drawn and least likely to be missed.
• Rice, atta and the base staples, because volume makes small percentage errors expensive.
• Dairy: milk, butter, cream, cheese. Short shelf life plus high rate is the worst combination you own.
• Anything imported or seasonal costing several hundred rupees a kilo, however small the quantity.
• Delivery packaging, which nobody counts and everybody over-consumes.
• Liquor, if you serve it, counted every single night without exception.

Ten to twelve lines, counted at close, on a sheet printed in the order the racks are arranged so nobody walks back and forth. Ten minutes once the habit sets. Widen weekly, count fully monthly. That cadence survives a busy season; a daily full count does not.

Reading variance by item to find the real leak

An aggregate food cost percentage is a smoke alarm with no address. Item-wise variance is the address. Put a week of sold quantities beside a week of issues and counts, and the shape of the numbers names the cause before anyone has to interrogate anybody.

• One ingredient consistently over, everything else clean: the recipe is wrong or the portion tool is wrong. Start with the card, then the scoop.
• Everything over by a similar small percentage: a yield assumption in the base recipes, or how stock is received and weighed.
• A clean week with one violent spike: a specific event. A spoiled batch, a large comp, a party order cooked off-recipe, a delivery received short and never disputed.
• Over on delivery but not dine-in: delivery portions are heavier, or packaging and the extra gravy sachet were never costed in.
• Over on weekends only: a shift, a person, or bulk prep whose surplus gets thrown at closing.
• Consistently under theoretical: not good news. Usually an overstated recipe or a kitchen quietly under-portioning, which arrives later as guest complaints instead of cost.

The last one deserves more attention than it gets. Owners celebrate a negative variance. Guests notice it first, and they do not write a report; they stop coming.

Staff meals, comps and wastage as recorded categories

The most common self-inflicted wound here is the staff meal. Twelve people eat twice a day out of the same rice sack, the same oil tin and the same vegetables, and none of it is recorded. That consumption is entirely legitimate and entirely invisible, and it lands in your variance as an unexplained loss that makes an honest kitchen look crooked.

• Staff meal, issued and recorded like any other issue, with a rough headcount attached.
• Wastage, recorded with a reason: spoiled, burnt, over-prepped, dropped, returned by guest. The reason is the whole value of the entry.
• Complimentary items, tied to the bill they were given against and to the person who approved them.
• Tasting and recipe development, kept separate so a month of menu work does not read as leakage.
• Transfers between outlets, recorded on both sides on the same day.

The approval line on comps matters more than the record. If any manager can write off a table with no name attached, comps become a drain nobody will ever trace. Make the approver visible on the bill and the volume corrects itself within a month, without a single conversation.

The weekly routine that keeps food cost honest

None of this works as a project. It works as a fifteen-minute habit on a fixed day. This is the version that survives contact with a real outlet.

• Every night at close: count the ten high-value lines, record wastage and staff meal, settle the store register.
• Monday morning: pull item-wise sales for the previous week and compute theoretical consumption for those same ten lines.
• Put actual against theoretical in quantity and in rupees. Rupees to prioritise, quantity to diagnose.
• Set a tolerance you are willing to defend, tighter on your top three items, and treat a breach as an event needing a written cause rather than a shrug.
• Fix exactly one cause per week. Chasing five at once means you learn nothing about which fix worked.
• Re-cost affected dishes whenever a purchase rate moves enough to matter, and check the menu price against the new cost before the month ends, not after.

How the workspace makes the two numbers meet

The reason this stays undone in most kitchens is not laziness. The sales number lives in a billing machine, the consumption number lives in a notebook, and reconciling them by hand every week is a job nobody has. BizRevolt's restaurant workspace puts the two in the same place. One-tap KOT fires each order to the right station and tracks every item from fired to served, so the sales side is item-level rather than a lump daily total. One menu runs across dine-in, takeaway and delivery, with modifiers, combos and item-level discounts consolidated into a single order, so the theoretical is computed against what the guest actually ordered.

Inventory tracks raw-material stock and recipe consumption with low-stock alerts, which is the store register and the expected depletion in one view instead of two books that never meet. The live dashboard carries today's sales, order count, open orders and top-selling items, so the lines worth counting tonight are obvious. Sales, item-wise and day-end reports export to CSV when you want to work the numbers yourself or hand them to an accountant.

You did not open a restaurant to run a store register. But the difference between a kitchen that makes money and one that merely turns over is rarely the menu, and usually the ten items nobody counted. Pick your ten tonight.