Six steps, from export to measured lift.
This is the classic menu-engineering analysis — the Kasavana-Smith matrix, popularity against profitability — run on your own sales data and returned as a short list of moves with dollar figures. No inventory integration, no recipe database, no three-month onboarding.
Every plate of that steak puts $20.40 toward rent, labor and profit. That number — not the food-cost percentage — is what the analysis runs on.
What the product actually does
One file in
A CSV or Excel upload of your menu-item sales export: item name, units sold, gross revenue, per-item food cost where available. Ninety days is the right window — long enough to smooth out a slow week, short enough to reflect the current menu.
Margin, not percentage
We compute contribution margin per dish — menu price minus plate cost — not food-cost percentage. The percentage misleads, and the full example is below. Where your POS lacks plate costs, you fill a one-page cost sheet for your top 20 items.
The matrix
Every dish lands in one of four quadrants: Stars (high margin, high volume), Plowhorses (low margin, high volume), Puzzles (high margin, low volume), Dogs (low margin, low volume).
Five moves, ranked
Never more than five specific moves — reprice, reposition, replace, remove — each with an annual dollar estimate and the arithmetic shown. An operator will act on three things, not thirty.
Where each dish sits
Where each item should sit on the physical menu, with the reasoning: stars featured top-right, puzzles repositioned and renamed, dogs removed unless they serve a purpose like dietary coverage.
Thirty days later
After the new menu goes live, we compare the next 30 days against the baseline and report actual contribution-margin lift — not a projection, the measured number.
A 40% food-cost steak can out-earn a 20% food-cost salad
Food-cost percentage tells you how much of the price went to ingredients. It says nothing about how much money the plate actually made. Run both dishes with real numbers:
| Steak | Salad | |
|---|---|---|
| Menu price | $34.00 | $14.00 |
| Food cost | 40% · $13.60 | 20% · $2.80 |
| Contribution margin | $20.40 | $11.20 |
| Per 100 plates sold | $2,040 | $1,120 |
The steak "fails" the percentage test and wins by $920 on every 100 plates. An operator who prices to a percentage target pushes the steak off the menu, keeps the salad, and banks less money with a better-looking food cost.
This is the most common error operators make when they engineer a menu by gut. Correcting it is the whole job: contribution margin pays the rent; the percentage is an accounting ratio.
None of this excuses a bad plate cost. At a 32.4% industry-average food cost, a steak priced so its margin collapses is still a problem — the matrix catches that too, as a plowhorse or a dog. The point is that the percentage alone cannot tell you which problem you have.
Reprice the ricotta hotcakes from $15 to $16. The brunch staple, last touched two menus ago.
The matrix, plainly labelled
Popularity runs left to right, margin runs bottom to top. Each point is a dish. No jargon on the axes — you read it the way you read the dining room.
| Dish | Units / 90 days | Margin / plate | Quadrant |
|---|---|---|---|
| Cast-iron ribeye | 980 | $27.60 | Star |
| Buttermilk fried chicken | 1,580 | $18.00 | Star |
| House burger, aged cheddar | 2,100 | $12.73 | Plowhorse |
| Lamb chops, mint chimichurri | 290 | $22.80 | Puzzle |
| Duck confit | 310 | $20.15 | Puzzle |
| Seasonal sorbet trio | 460 | $7.20 | Dog |
Quarterly by default, because menus change seasonally.
Each quarter you send a fresh 90-day export and the report re-runs against the new mix. On the subscription, weekly prime-cost tracking runs alongside — weekly food-cost tracking catches problems roughly three times faster than monthly, and prime cost above 65% of revenue is structural distress in every segment.
The deliverable, page by page
The full sample report for a fictional Portland restaurant is on this site — every table, the matrix, the five moves, the menu layout, the 30-day tracking plan — and downloadable as a PDF.