Menu Engineering: Out of 90 Dishes, 7 Make the Money

01Two numbers: profit and popularity

Every dish needs two numbers. The first is contribution margin: selling price minus food cost. The second is unit sales: how many times it sold last month. A third number is not needed and only distracts.

Most people think in percentages: "this dish has a 70% margin". That is a mistake. 70% of a 12 000 tea is 8 400 UZS; 35% of a 90 000 meat dish is 31 500 UZS. What lands in the till is money, not percent.

In the laCafe admin panel this report is already built: Dashboard → Products → menu engineering. Unit sales come from the till (or from iiko); you enter the food cost on the dish card, an internal field a guest never sees.

02Four quadrants: star, workhorse, puzzle, dead weight

Cross the two axes and four groups appear. Star: sells a lot, earns a lot. Workhorse: sells a lot, earns little. Puzzle: earns a lot, sells little. Dead weight: low on both.

Each group is handled differently. Leave the star alone: hold the quality and put it in the most visible slot. Raise the workhorse's price by 5–8% or cut its cost; it will still sell, because it has become a habit.

The puzzle is the interesting group. The dish is profitable but the guest doesn't see it or doesn't understand it. The fix: add a photo, rewrite the description, move it higher, tell the waiters to recommend it. If sales double in two weeks, it has become a star.

Cut the dead weight. Every surplus dish is stock sitting in the fridge, prep time in the kitchen and a distracting line on the menu. Removing 5–7 of 90 items does not cut revenue: the guest orders something else.

03Order works differently in a digital menu

On paper the eye goes to the top-right corner. On a phone the guest scrolls top to bottom and rarely goes past the first screen. So in a QR menu the category order and the first three rows of each category are the most valuable real estate.

A practical rule: put stars and puzzles in the first two rows of every category. Push dead weight down or remove it. Reorder the categories too: if the evening sells kebabs, "Salads" does not have to come first.

Category hours work even more precisely. Give "Breakfast" an 08:00–11:00 window and after noon it disappears entirely while "Business lunch" moves up. The till still sees everything; the limit applies only to guest-facing screens.

• The first 2 rows of a category: stars and puzzles
• Modifiers (extra sauce, larger portion): the cheapest lift to the average check
• The cross-sell report: which dishes are bought together — put that pair side by side
• Calorie and allergen data speeds up the choice for fitness-minded and family guests

04A one-month experiment: what to measure

Menu engineering is not a one-off. The cycle is simple: open the report at the start of the month, make three changes (one price, one ordering, one removal), open it again at month end. Never make more than three: you won't be able to tell what worked.

Three metrics to watch: average check, share within the category and cancelled orders. If a dish you repriced drops more than 15% in units, you crossed the price threshold — roll it back.

Read it together with the stop-list history. If plov ran out at 14:30 on 22 days this month, that is not a menu question but a production one: you need a second kazan. The report answers both.