Star-plowhorse-dog-puzzle matrix: what it COSTS before, what it leaves after

What does the star-plowhorse-dog-puzzle matrix cost in 2026

As of August 2026, the star-plowhorse-dog-puzzle matrix costs between 0 and 380 USD if you build it in-house with standard recipes already written, and between 1,400 and 4,800 USD if you hire a consultant who also rebuilds your per-portion costing and reorders the menu. The Medellín steakhouse that opened this piece paid for the full version and recovered the investment in six weeks: it swapped the charcoal loin, running 41% food cost and 9,400 pesos of contribution per plate, for the crusted trout, which returned 16,800 at 24% cost. That 7,400-peso gap per unit, multiplied by roughly 340 monthly covers, explains why the consulting fee becomes irrelevant against the first quarter of execution. The low range is not worse because it is cheap; it is worse when the input data is dirty. The three investment tiers buy different things, and it pays to know which.

What each price range includes?

Between 0 and 120 USD you are paying for a spreadsheet template and your own time: export the sales mix from the POS, paste in the costs you already have, and out come the four quadrants.

From 120 to 380 you get subscription menu engineering software, with a POS connection and automatic mix updates, roughly 30 USD a month over a year. The real jump happens in the 1,400 to 2,600 band: there a consultant builds per-portion costing with actual waste and cleaning yield, something no template resolves on its own. And from 2,600 to 4,800 you add graphic redesign of the menu, price testing by zone, and support across two mix cycles. Against a 2.8% median pre-tax profit in full service (National Restaurant Association, Restaurant Operations Report 2024/25), that support is what decides it. Percentage food cost is for buying; contribution margin in money is for deciding the menu.

The axis almost everyone builds wrong

That distinction, which Diego F. Parra repeats in every Masterestaurant diagnostic, flips the quadrants of an entire menu. A dish at 33% cost returning 14 USD outranks one at 22% returning 6, always, unless the kitchen is hitting its capacity ceiling and the bottleneck is pass time rather than cash. On the percentage axis, the pasta of the day looks like a flawless star; on the money axis it turns out to be a plowhorse that fills the kitchen and never fills the till. Kasavana and Smith published the mechanics at Cornell in 1982 and nothing has changed in forty-four years: popularity against marginal profitability, four quadrants, four decisions. What does change, and what ruins the exercise, is the number you feed into the vertical axis. Five variables explain almost the entire span between 1,400 and 4,800 USD. Menu size weighs first: going from 30 to 61 items, like that 72-seat steakhouse, nearly doubles the costing hours and adds between 600 and 900 USD.

Five factors that move the consulting price

Recipe status comes next; if no written standard recipes exist, add 400 to 1,200 depending on kitchen complexity. A third factor is POS integration, because exporting ninety days of clean mix runs 150 to 300 when the system is old. Fourth: number of locations, with a typical 25 to 40% increase per additional unit, since waste and suppliers shift by site. And the fifth, the one most underestimated, is the graphic redesign of the menu, which rarely lands below 500 USD once new photography enters the picture. Your data window decides the classification as much as the costing does. Ninety days of sales mix by product is the defensible minimum, and a winter dish classified against July data ends up a dog by accident of the calendar, not on its own merits. Pulling it will cost you next season's menu, and that kind of mistake stays invisible for six months, by which point nobody remembers where the decision came from.

Ninety days of data, not thirty

There are contexts where the calendar outweighs the margin: non-alcoholic beverage sales grew 30% in 2024 (Restaurant Dive), a structural shift a thirty-day window would read as noise. Same with the GLP-1 effect, with 54% of users eating out less frequently (survey of 1,000 users via Fortune, 2025). If your mix moved because of that, a one-month matrix is lying to you. Almost nobody has real per-portion costing, and that is where the exercise gets lost before it begins. Supplier invoice prices go in without waste, without cleaning yield and without cooked weight, so a sirloin portion that invoices at 7.20 USD per kilo ends up costing 11.90 after 27% loss in butchering, a routine figure for beef cuts according to yields published by the USDA. That 65% deviation on unit cost does not spread evenly: it punishes protein cuts and forgives the sides, so the whole matrix tilts.

The costing error that dresses a plowhorse as a star

With the wrong input, a plowhorse dresses up as a star and you spend a year defending the dish that eats your margin. Before paying for any analysis, weigh three portions of each cut, raw and cooked, and write down the yield. Negotiate on scope, not on discount, which is where the consultant has real room. Ask for the exercise on your 20 highest-turnover dishes instead of all 61 items: that usually trims 30 to 40% off the fee and still captures most of the mix, because the long tail of a menu rarely clears 15% of sales. Second move: build the waste costing yourself with your head chef over two weeks before the first meeting, and deduct those hours from the budget. Third, pay by deliverable rather than by the hour, with the per-dish decision list as the closing condition. And demand the output in actionable form, not as a chart.

How to negotiate and optimize the spend?

If your final product is a four-quadrant diagram instead of a list carrying a new price, a new portion and a new menu position for every item, you paid for a drawing.

The matrix is not an analysis exercise: it is a PRICING exercise. The useful deliverable fits on one page: a column per dish with its assigned action and a date. Stars get protected, hold their visual position and stay untouched on price unless proven cost pressure appears. Plowhorses get adjusted by portion or by side, because raising their price outright is the fastest route to losing the frequency that made them popular. Puzzles get moved and rewritten, and the evidence there is hard: descriptive labels lift sales 27% against dishes with no description (Cornell Food and Brand Lab, Wansink). Dogs come off, unless they hold up a menu promise that costs more to break than to keep.

What has to be on the table the following Monday?

Start Monday with the puzzles, the only category where the change costs nothing and the effect shows up within two weeks of mix. The profitability axis.

The DIY version almost always builds it on food cost percentage, because that is the number the owner already has; the properly built version uses contribution margin in currency. Quadrants shift completely: on the percentage axis, the pasta of the day looks like a star, while on the currency axis it stands revealed as a plowhorse that fills the kitchen and leaves the till alone. The data window. Ninety days of item-level sales mix, never thirty. A winter dish classified on July data is a dog by accident of the calendar, and pulling it will cost you next season's menu. The decision list. A matrix without a per-dish action —new price, new portion, new position— is an academic exercise. Diego F. Parra puts it bluntly in Masterestaurant audits: if the deliverable does not fit on one sheet the head chef can tape to the wall, nobody will execute it.