How to Sort Your Menu into Stars, Plowhorses, Puzzles and Dogs in One Afternoon

The four-box menu matrix — Stars, Plowhorses, Puzzles, Dogs — comes from Menu Engineering: A Practical Guide to Menu Analysis, published in 1982 by Michael Kasavana and Donald Smith. It has survived forty years because it needs only two things you already have: how many of each item sold, and what each one makes you.

Most explanations of it stop at the four labels. The arithmetic underneath is the part that matters, and it is arithmetic you can do on paper. Here is the whole thing on a real-shaped seven-item entree category.

What you need before you start

• Item sales counts for a consistent period — 90 days is the usual choice. Every POS will export this; look for "product mix", "item sales" or "PMIX".
• Plate cost for each item, costed with yield and a Q factor rather than straight off the invoice.
• Current menu price.

One rule that decides whether this works: run it inside one category at a time. Entrees against entrees, desserts against desserts. Comparing a $34 ribeye to a $6 side is arithmetic without meaning — the thresholds below are relative to the set you put in.

The data

Item	Price	Plate cost	Contribution	Sold (90 days)
Roast chicken	$18.00	$3.44	$14.56	420
Burger	$16.00	$4.80	$11.20	910
Ribeye	$34.00	$14.60	$19.40	260
Pasta primavera	$17.00	$3.10	$13.90	180
Fish special	$24.00	$8.90	$15.10	150
Veggie bowl	$15.00	$3.60	$11.40	330
Lamb shank	$26.00	$13.40	$12.60	60

Step 1: The contribution line

Contribution margin is price minus plate cost. Not price minus cost minus labor minus rent — those are fixed whether or not this dish sells, so they do not belong in the comparison.

Total units sold: 2,310. (420 + 910 + 260 + 180 + 150 + 330 + 60)

Total contribution: 420 x 14.56 = 6,115.20; 910 x 11.20 = 10,192.00; 260 x 19.40 = 5,044.00; 180 x 13.90 = 2,502.00; 150 x 15.10 = 2,265.00; 330 x 11.40 = 3,762.00; 60 x 12.60 = 756.00. That adds up to $30,636.20.

Step 2: The contribution threshold (this is the step people get wrong)

The threshold is the weighted average — total contribution divided by total units — not the average of the contribution column.

$30,636.20 / 2,310 = $13.26 per plate

The plain average of the seven contribution figures is $14.02. Use that instead and the pasta primavera drops below the line, turning from a Puzzle into a Dog — and you go off deleting a dish that actually makes $0.64 more than your average plate. The weighting matters.

Anything above $13.26 is a high-contribution item. Below it, low.

Step 3: The popularity threshold

If all seven items sold equally, each would take 1/7 of the volume — 14.29%. Kasavana and Smith set the bar at 70% of that equal share, on the reasoning that an item pulling nearly its fair share is doing its job even if it is not the top seller.

14.29% x 0.70 = 10.00%

So an item needs 10% of the 2,310 units — 231 plates — to count as popular. The shares:

Item	Share of units	Contribution	Quadrant
Burger	39.4%	$11.20	Plowhorse
Roast chicken	18.2%	$14.56	Star
Veggie bowl	14.3%	$11.40	Plowhorse
Ribeye	11.3%	$19.40	Star
Pasta primavera	7.8%	$13.90	Puzzle
Fish special	6.5%	$15.10	Puzzle
Lamb shank	2.6%	$12.60	Dog

Check any row you like: 260 / 2,310 = 11.3%, which clears the 10% bar, and $19.40 clears $13.26 — so the ribeye is a Star, not the luxury item you tolerate. That result surprises people, and it is the single most common finding when a menu gets run through this properly for the first time.

What to do with each quadrant

Stars (chicken, ribeye) — protect them. Do not touch the recipe, do not cheapen the garnish, do not move them off the top right of the page. If you must test a price increase anywhere, this is where a small one does least damage, because these customers are choosing the dish, not the price.

Plowhorses (burger, veggie bowl) — popular, thin. The burger is 39.4% of your covers at $2.06 below the average contribution. You have three levers, in order of least risk: cut the plate cost without the guest noticing (portion, supplier, trim), add a high-margin attachment the server can offer every time, or raise the price. Do one at a time, four weeks apart, so you know which one moved the number.

Puzzles (pasta, fish) — good money, nobody orders them. This is a menu design problem, not a recipe problem. Rewrite the description, move it out of the bottom-left of the page, put it in a box, get the servers naming it. If it is still under the line after eight weeks of real effort, it is not a Puzzle, it is a Dog with good margins.

Dogs (lamb shank) — low margin, low volume. Sixty plates in ninety days, at $12.60, is $756 of contribution for a SKU that carries its own prep, its own waste and its own line on the order guide. Cut it, or reprice it into a Puzzle and give it one honest try. The usual objection is "but a regular always orders it" — check the count. A dish with one loyal table is a special, not a menu item.

Cadence and traps

• Rerun it quarterly, and after any price change. Thresholds move when the mix moves.
• Do not run it across fewer than five items — the 70% rule stops discriminating.
• Watch seasonal items. Ninety days that include a holiday will label your seasonal dish a Star it cannot repeat.
• Do not cut every Dog at once. Removing four items changes both thresholds and reshuffles the board.
• Labor is not in here. A Star that ties up a station during the rush costs you in a way this matrix cannot see. Use it as the first filter, not the verdict.

Everything above is a spreadsheet and an afternoon. If you would rather see it done on your own menu first, MenuGenie does this classification from a photo or PDF of your menu, free — upload yours at menu-genie.com.