Menu Engineering: The Four-Quadrant Matrix

By Marji · Published August 10, 2026 · Updated September 30, 2026 · 5 min read

Export four weeks of unit sales, put a current plate cost beside each dish, then draw two lines: 70% of an equal share of units, and the weighted average contribution margin. Here they land at 117.5 units and $13.96, and five of the eight dishes above the popularity line — the best-selling burger among them — fall below the margin line. Stale costs name the wrong dish.

What you need before you start

A product mix report covering at least four weeks — avoid a holiday week, it distorts everything. Square exports one directly.
A current plate cost for each dish. Not last year's. See recipe costing.
Nothing else. No software required for a first pass — though the menu engineering worksheet will draw both lines and sort the quadrants for you.

Step 1 — Build the table

Here is a sample 12-item entrée section over four weeks; the prices, costs and sales are illustrative. Contribution margin is price minus plate cost.

Four weeks, 12 entrées, 2,014 units
Dish	Price	Cost	CM	Units	Total CM
Burger	$15.00	$4.20	$10.80	412	$4,450
Chicken sandwich	$14.00	$5.10	$8.90	310	$2,759
Margherita pizza	$16.00	$3.20	$12.80	268	$3,430
Pasta pomodoro	$19.00	$2.90	$16.10	180	$2,898
Fish & chips	$21.00	$7.40	$13.60	166	$2,258
Caesar + chicken	$17.00	$4.05	$12.95	158	$2,046
Salmon	$29.00	$8.29	$20.71	142	$2,941
Steak frites	$34.00	$13.10	$20.90	118	$2,466
Mushroom risotto	$22.00	$3.60	$18.40	94	$1,730
Duck breast	$36.00	$12.80	$23.20	71	$1,647
Lamb shank	$32.00	$14.60	$17.40	58	$1,009
Vegan bowl	$18.00	$4.90	$13.10	37	$485

Step 2 — Find the two dividing lines

Popularity line

(1 ÷ number of items) × 70% × total units

The 70% factor is the convention from Kasavana and Smith's menu engineering model: a dish counts as popular once it sells at least 70% of an equal share, so a long menu does not push most dishes below the line.

Worked example · The two lines for this menu

Total units: 2,014
Items: 12
Fair share per item — 2,014 ÷ 12: 167.8
Popularity line — 167.8 × 70%: 117.5 units
Total contribution: $28,119
Average CM per unit sold: $13.96
Lines: 117.5 units · $13.96 CM

Use the weighted average CM (total contribution ÷ total units), not the simple average of the CM column. The simple average lets a rarely-sold luxury item drag the whole line up.

Step 3 — Sort into quadrants

The same 12 dishes, classified
Quadrant	Dishes	What it means
Stars — popular, high CM	Pasta, salmon, steak frites	Above 117.5 units and above $13.96 CM
Plowhorses — popular, low CM	Burger, chicken sandwich, pizza, fish & chips, Caesar + chicken	Above 117.5 units, below $13.96 CM
Puzzles — unpopular, high CM	Risotto, duck, lamb shank	Below 117.5 units, above $13.96 CM
Dogs — unpopular, low CM	Vegan bowl	Below both lines

Step 4 — Act, by quadrant

Plowhorses: reprice or rework — start here

Demand is proven, so a modest increase is the lowest-risk test on the menu — but it is still a test, so count the units afterwards. The burger at $16.00 instead of $15.00, with an assumed 5% volume loss:

Worked example · The burger, repriced

Current — $15.00, CM $10.80, 412 units: $4,450
After — $16.00, CM $11.80
Assume 5% volume loss → 391 units: $4,614
Change over 4 weeks: +$164
Annualised: +$2,132

And that assumes you lose volume. The break-even volume loss is 8.5% — you'd have to lose more than one burger in twelve before the increase stopped paying.

Stars: protect them, and recost them first

Don't discount them, don't shrink them, don't bury them on the page. The trap is complacency: a star whose protein has crept up 20% since you costed it is quietly your most expensive problem, because it sells the most. Recost stars monthly.

Puzzles: reposition before you cut

Three dishes here contribute $17.40 to $23.20 each and sell fewer than 100 in four weeks. Before concluding they don't work, try the cheap interventions: move them up the section, rewrite flat descriptions, have servers mention one for two weeks. Duck breast at 71 units contributing $23.20 is worth ten minutes of effort — every additional 20 units a month is $464. Steak frites, at 118 units against a 117.5 line, is a star by half a unit; a dish that close to a line can change quadrant from one period to the next, so read it as a close call rather than a verdict. The menu psychology evidence review covers what has been tested for two of those moves: position and descriptions.

Dogs: cut, with exceptions

The vegan bowl is a textbook dog: 37 units, below-average CM, consuming prep and inventory SKUs. But cutting it may cost you tables — a group of six doesn't book a restaurant where one person can't eat. This is exactly the case where the matrix informs a decision it shouldn't make for you.

How often to redo it

Quarterly for the full matrix — sales mix shifts slowly.
Monthly for recosting your stars and plowhorses — costs shift fast.
After any menu change, because adding an item changes the popularity line for everything.

Should I use contribution margin or food cost percentage?

Contribution margin, in dollars. A percentage tells you a dish drifted from where you priced it; it can't tell you how much money the dish makes. The burger in this example is on-target by percentage and the biggest opportunity on the menu by dollars.

Yes, and it's easier — with twelve items you can do it in a spreadsheet in an hour. The popularity line adjustment matters more on large menus, where without it everything looks unpopular.

What about dishes that sell mostly as specials?

Exclude them from the matrix, or the availability difference will misclassify them. A dish available two nights a week isn't unpopular; it's rationed.

Keep the contribution column current.

Marji keeps each dish's plate cost tied to your supplier invoice prices, so the contribution margins you plot are today's rather than the ones you costed at launch. Unit sales still come from your own POS report.

Try it on my menu →