Menu engineering from your own bills: what to promote, what to reprice, what to cut.

Most owners can name their top five sellers from memory. Almost nobody can name the five dishes that put the most money in the drawer, because that list is invisible from the floor. A dish that sells 400 plates a month at ₹40 of margin beats a dish that sells 30 at ₹200, and nothing on the pass tells you which is which.

FireAI, an Indian menu tool, states the point more clearly than most of the literature: "A ₹100 dish at 25% food cost contributes ₹75. A ₹500 dish at 40% food cost contributes ₹300" (fireai.in). The dish with the worse percentage is the better dish to sell. Food cost percentage is a pricing check, which is what the menu pricing formula post is for. What you rank a menu by is contribution per sale multiplied by how often it sells.

So the exercise needs two lists side by side: how many of each dish went out, and what each one left behind. Your till has the first list already. The second one needs a number nobody has written down yet.

Open a dish in Menu & stock and there is now a field beside the price: "Cost to make (₹, optional)". That is the whole data-entry job. Not a recipe tree, not a purchase ledger, one honest number per dish that you update when a vendor's price actually moves. A dish with no cost still appears in the table under the plot, labelled cost unknown, and is left off the plot itself; the report counts how many are missing so you can work through them.

The honest limit. EasyKOT has no inventory module and no recipe module. It knows what you sold because it rang every bill; it does not know what is in your store room, what spoiled, or what your true food cost was last month, because that needs purchase records and recipes. The cost you type is your estimate, and the plot is only as good as it. Petpooja and Restroworks both ship inventory and recipe modules, including multi-stage recipes that deduct stock as it is used, and deserve the credit for it.

Do it roughly and do it now. Weigh one portion, price the ingredients off this week's vendor bills, add a rough share for oil, masala, gas and packing. A number that is roughly right still separates a ₹40 margin from a ₹200 one, which is the only question this exercise asks; a number that is exactly right changes none of the decisions below. For a sanity check against the whole kitchen, Petpooja's free food cost calculator puts Indian full-service food cost at 28% to 32% and says anything above 35% means waste, over-purchase, theft or wrong pricing (11 May 2026), while DineOpen's margins guide gives a slightly wider 28% to 35% including packaging and wastage, with a warning above 38% (dineopen.com, 12 Mar 2026). Both are vendor benchmarks rather than survey data, so treat them as a band and not a target.

The matrix comes from Michael L. Kasavana and Donald I. Smith, Menu Engineering: A Practical Guide to Menu Analysis (Hospitality Publications, Okemos MI, 1982), catalogued at WorldCat. They adapted the BCG portfolio matrix, plotting contribution margin, meaning selling price minus that item's own food cost, against sales volume or menu mix (FIU Hospitality Review). Toast's menu engineering matrix guide is the clearest English explainer if you want the long version.

Two rules make the plot honest. Compare like with like: mains against mains, starters against starters, never the whole card at once, or every dessert is a Dog and every biryani a Star (NetSuite). And the line between popular and unpopular is conventionally drawn at (1 ÷ number of items) × 0.70, so on a group of ten dishes anything at 7% of that group's sales or above counts as popular (Northern Arizona University, meez). That is what the chart draws: the vertical line sits at 70% of the average units sold, and the horizontal one at the average profit per unit. The 0.70 is a convention repeated across the teaching literature rather than a tested finding, so treat a dish sitting on the line as neither, and look again next month.

The instinct with a popular low-margin dish is to raise its price, and it is usually the wrong first move. The dish is popular. It brings people through the door and it is what they compare when they decide whether you got expensive. Three cheaper moves come first.

• Check the portion against what comes back. If plates return with food on them, the portion is bigger than the appetite, and trimming it costs you nothing in satisfaction and takes the cost down directly.
• Check the cost you typed. A Plowhorse is often a Star with a stale number behind it. Re-price the ingredients from this week's bills before you touch the menu card.
• Sell it as a Half. Half and Full are two prices on the same dish in the register, and half the food is not half the cost: the gas, the cook's time and the plate do not halve. Pricing the Half above half the Full lifts the margin per plate while the number on the card stays where guests remember it. One limit: the report reads a dish whole, so a Half and a Full share one dot and one cost. Judge portions off the bills rather than off the plot, and the pricing formula post works that sum.

Only then move the price, by a little, and watch the units for two weeks. The dashboard's Top sellers list ranks units over the last seven days, so a drop shows up fast enough to reverse. A Plowhorse that stays popular at the new price was underpriced. One that falls off a cliff tells you the old price was doing work.

A Dog earns little and sells little, and every one of them costs you prep, a spot in the fridge, a line on the card and one more dish a new cook can get wrong. Cutting five of them usually makes the kitchen faster before it makes anything more profitable. Take them off the card at the end of a month, not mid-week, and tell the kitchen why.

Two Dogs are worth keeping deliberately. The first is the one that anchors a decision: the single Jain or plain dish that one person in a group of eight needs before the group picks your restaurant. Its own sales look terrible and the seven bills beside it do not. The second is the one that shares its prep completely with a Star, so carrying it costs almost nothing. A paneer dish that uses the same gravy as your best-selling one is not really a separate item.

Everything else should go, and a dish you keep for sentiment is a decision rather than an accident, as long as you made it once. If a dish is out of stock often, the register's stock-out toggle marks the tile Out of stock on every register until somebody switches it back on, instead of leaving the captain to apologise; a dish that is out most nights is answering the question for you.

A month is the right window. It gives most dishes enough sales to mean something, and it is short enough to catch a vendor price rise in the same quarter it happens. Yearly is too slow to act on and weekly turns ordinary variation into a decision. Put it on the same date each month, next to the weekly accounts close, and keep the ambition small: promote one Puzzle, fix one Plowhorse, remove one Dog. Twelve of each in a year changes a menu completely.

Know what the model does not see. It treats contribution as profit, so it ignores labour and prep time entirely, and it quietly favours expensive dishes, a criticism repeated across the teaching literature. A biryani that takes a cook forty minutes and a raita that takes two can land in the same box. Read the plot, then apply what you know about your own kitchen.

If you would rather type numbers into a form than record costs on your dishes, both DineOpen's menu engineering tool and Petpooja's food cost calculator are free and good, and they need nothing from your till. The difference here is that the popularity axis is not typed by anyone: it is the dishes you actually sold. Setting the costs up is covered in the menu setup guide. EasyKOT is ₹299 a month or ₹2,990 a year plus 18% GST, with three months free and no card, if you want next month's plot drawn from this month's bills.