Star-plowhorse-dog-puzzle matrix: what changed in 2026 and what to do with your menu
The star-plowhorse-dog-puzzle matrix still works in 2026, but sorting by food cost percentage is pulling cash-generating dishes off menus: the real trend moves the vertical axis to contribution margin in currency per dish and weights popularity by channel sales mix, because a plate carrying 34% food cost and $9 of margin pays more payroll than one at 24% and $3.20. Traditional method: two static axes, annual review, decisions by ratio. Masterestaurant method: margin in dollars against a channel-level popularity index, quarterly review, and an exit rule written down before anyone looks at the numbers. Maximum acceptable food cost per dish: 32%.
A 68-item menu at a three-unit steakhouse showed nine stars in the POS report. Reclassify that same data by contribution margin in dollars and four survive; two of those supposed stars were plowhorses in costume, glowing only because a tiny food cost percentage sat on top of a tiny price.
That is the state of the craft in 2026. The matrix is not broken, it is badly fed. Michael Kasavana and Donald Smith published it in 1982 with two axes, popularity and contribution margin, and that second axis was ALWAYS money, never a ratio. The drift arrived later, when point-of-sale systems started exporting food cost percentage because it was the easy column.
Operations tables and the National Restaurant Association Show have been pushing since 2024 toward per-dish decisions instead of per-category ones. With sector prime cost sitting near 60% of sales and food-away-from-home inflation still running above headline CPI, every line on the menu competes against the venue's break-even point, not against its own ratio.
Side-by-side comparison
| Traditional method | Masterestaurant method | |
|---|---|---|
| Matrix vertical axis | ✕Food cost % per dish (cut at 30-35%) | ✓Contribution margin in $ per dish (cut at weighted mix average) |
| Horizontal axis (popularity) | ✕Total units sold, cut at 70% of average | ✓Popularity index by channel: dine-in, delivery and take-away measured apart |
| Review cadence | ✕Annual, or whenever the menu is redesigned (1 time per 12 months) | ✓Quarterly, plus a monthly read on 4 sentinel dishes (4 reviews/year) |
| Decision on a dog dish | ✕Cut from the menu with no further analysis | ✓Written exit rule: 2 quarters below threshold and <1.5% of mix before removal |
| Menu size it produces | ✕Menus of 60-90 lines that dilute the sales mix | ✓Physical menu of 24-32 lines plus QR for extended and seasonal items |
| Costs charged to the dish | ✕Often allocates prorated payroll and rent to the plate | ✓Only ingredient and yield loss (food cost ≤32%); payroll and rent go to break-even |
| Measured effect on check | ✕Check stays flat; margin moves through price, not mix | ✓Mix re-engineering: 3-7% gross margin without touching list prices |
The vertical axis returns to currency: the trend that reorders the whole menu
Move your matrix's vertical axis from percentage food cost to contribution margin in currency per dish, because the percentage hides the cash. A ceviche at 24% cost on a $13 price yields $9.88 per unit; a mushroom risotto at 31% on $34 yields $23.46, nearly two and a half times more, and payroll gets paid with those $23.46, not with the pretty percentage. Kasavana and Smith published the matrix in 1982 with two axes, popularity and contribution margin, and the second one was ALWAYS monetary; the drift arrived when point-of-sale systems started exporting the easy column. With full-service pre-tax profit at 2.8% of sales in 2024 according to the National Restaurant Association's Restaurant Operations Report, pulling a dish that contributes twenty-three dollars per cover is a mistake the sector median does not forgive. Build a separate matrix per channel, because blending dining room and delivery into the same scatter produces a dish that exists nowhere.
One matrix per channel, not an average matrix of three businesses
When the aggregator charges 22% commission, the $18 dish that left $11 of margin at the table drops to $7.04 in the app, and that gap of almost four dollars completely reorders the quadrant it lives in. Channel weight is no longer marginal: off-premises traffic in full service reached 30% in 2024 against 19% in 2019, and in limited service it climbed from 76% to 83%, per the National Restaurant Association's Off-Premises Report 2024. A single-shift location can live with two matrices on the same sheet; a three-unit operation needs to weight popularity by each channel's real sales mix before touching a single price. Reclassify the matrix every quarter. Input prices moved with food inflation above 2% year over year through 2025 and 2026 according to the USDA Economic Research Service, so a classification made in January describes a menu that stopped existing around May.
From annual to quarterly review: January's menu no longer describes September
I got this wrong for years recommending the annual cycle, comfortable for the printer and ruinous for margin: if the core protein cost of a dish climbs eight points and the selling price stays put, a star becomes a plowhorse and nobody notices until year-end close. The operating rule we apply at Masterestaurant is short: twelve weeks, updated recipe costing for the ten highest-volume dishes, and a recalculation of margin in dollars. A family steakhouse can do it in one afternoon with the POS export and a spreadsheet. Popularity is not just demand, it is also exposure, and in 2026 menu design enters the matrix as a variable. A guest spends 109 seconds on average reading the menu according to NeatMenu's menu psychology analysis, enough time to work through twelve to fifteen lines with real attention. If your four highest-currency-margin dishes are buried in the third column of a 68-line menu, the POS will hand back a low popularity figure that measures invisibility rather than preference, and you will end up pulling exactly what you should have pushed.
The 109 seconds of reading rule over the popularity quadrant
Before moving a dish into the dog quadrant, raise it two positions and give it ninety days of testing. At a three-unit steakhouse we reviewed, that relocation alone pushed two supposed puzzles across the popularity median without touching price or recipe. Adopt two things now and watch a third without investing in it. What goes this quarter: contribution margin in dollars as the vertical axis, and a matrix split by channel with the aggregator's commission deducted dish by dish; both are assembled with data your point of sale already exports and cost no extra license. What deserves watching without buying yet are dynamic pricing models by time slot, which promise to recover margin in valley hours but punish fair-price perception in a market where morning daypart traffic barely returned to growth, up 3% in March 2025 and the first rise since the second quarter of 2023, according to Circana.
The horizon: what to adopt this quarter and what to watch from a distance
Diego F. Parra sums it up in one line: first fix the menu you have, then automate the one you want. Ignore for now the dashboards promising to reclassify your matrix automatically with artificial intelligence. The problem is not the algorithm, it is that the algorithm feeds on the same stale recipe costing and the same percentage food cost that produced the wrong classification in the first place; a model eating garbage returns garbage sorted with elegant charts. On that 68-dish steakhouse menu I mentioned, the POS software flagged nine stars and four survived reclassification by dollar margin: two were plowhorses in disguise that shone only because their cost percentage was tiny on a tiny price. No AI layer fixes a recipe cost nobody updated in fourteen months. Use AI for what it genuinely solves today, which is crossing sales mix by channel and hour, and keep the decision to pull a dish where it belongs, at the operations table with the costing sheet open.
Before killing a dish, measure what it drags down with it
Do not remove a dog without calculating the cash it drags along. A dish with low popularity and low margin may be holding up an input shared with four other lines, and by pulling it you lose purchase volume, raise that input's unit cost and punish the margin of dishes that were working fine. The alternative that usually pays better is reformulation: cut the gram weight of the expensive protein, change the side, raise price between 8% and 12% and give it a quarter. With sector prime cost hovering near 60% of sales and limited-service median profit at 4.0% in 2024 according to the National Restaurant Association, every menu decision competes against the location's break-even point, not against its own percentage. Open the costing sheet for your ten highest-volume dishes this week and recalculate margin in dollars before printing anything. Ratio versus money. Ceviche at 24% food cost on a $13 price leaves $9.88 of margin; a mushroom risotto at 31% on $34 leaves $23.46.
The differences that move cash
The percentage says ceviche wins, the till says otherwise, and payroll gets paid from the till. One channel versus three. When the aggregator charges 22%, an $18 dish earning $11 in the dining room drops to $7.04 on delivery. Blend both channels into one matrix and you are averaging two different businesses, then deciding about a dish that does not exist. Annual versus quarterly review. Ingredient prices moved with food-away-from-home inflation running above 2% year over year across 2025-2026 per USDA Economic Research Service, so a January matrix describes a menu that no longer exists by September. Cutting versus reframing. A puzzle is not a problem, it is a badly communicated opportunity: the margin is already there, the mix is missing. Move it into the high-read zone of the physical menu, give it a description with texture, and it banks more than deleting it ever would.
The differences that move cash — in practice
Long menu versus governed menu. Sixty lines dilute the sales mix until no star clears 4% of volume, and a kitchen running 68 recipe cards carries yield loss that nobody audits. List price versus price psychology. Raising everything 6% is the expensive shortcut; anchoring a high-priced plate next to the one you want to sell, dropping the currency symbol and breaking the aligned price column shifts the mix without shifting the perception of expense.
Head to head, criterion by criterion
How the classic matrix operates1982-2020
- Four fixed quadrants: star (high popularity, high margin), plowhorse (high popularity, low margin), puzzle (low popularity, high margin) and dog (low popularity, low margin).
- Popularity cut set at 70% of average sales per item, a 1982 convention almost nobody recalculates.
- Profitability axis read as food cost percentage, which rewards the cheap plate and punishes the expensive one even when the expensive one banks twice the cash.
- One matrix for the whole business, dine-in and delivery blended, despite aggregator commissions taking 15% to 30% of the price.
- Review chained to the menu redesign, which in practice means once a year at best.
- Binary calls: the dog goes, the star stays, and the puzzle waits for somebody to push it.
How we rebuilt it at MasterestaurantMasterestaurant
- Vertical axis in money: contribution margin per dish equals selling price minus ingredient cost and yield loss. The cut is the weighted average of the real mix, recalculated every quarter.
- Horizontal axis by channel, since the same dish can be a star in the dining room and a dog on delivery once the aggregator commission enters the arithmetic.
- A WRITTEN exit rule, signed before anyone reads the data, so no chef rescues a favourite plate with after-service arguments.
- Four sentinel dishes tracked monthly; the rest of the menu, quarterly.
- A physical menu of 24 to 32 lines that governs the guest experience, with a QR menu extending it for seasonals, allergens and delivery.
- Food cost ceiling of 32% per dish as a MAXIMUM tolerance, never a target; payroll, rent and utilities get settled at the venue's break-even point.
Side-by-side comparison
| Traditional method | Masterestaurant method | |
|---|---|---|
| Matrix vertical axis | ✕Food cost % per dish (cut at 30-35%) | ✓Contribution margin in $ per dish (cut at weighted mix average) |
| Horizontal axis (popularity) | ✕Total units sold, cut at 70% of average | ✓Popularity index by channel: dine-in, delivery and take-away measured apart |
| Review cadence | ✕Annual, or whenever the menu is redesigned (1 time per 12 months) | ✓Quarterly, plus a monthly read on 4 sentinel dishes (4 reviews/year) |
| Decision on a dog dish | ✕Cut from the menu with no further analysis | ✓Written exit rule: 2 quarters below threshold and <1.5% of mix before removal |
| Menu size it produces | ✕Menus of 60-90 lines that dilute the sales mix | ✓Physical menu of 24-32 lines plus QR for extended and seasonal items |
| Costs charged to the dish | ✕Often allocates prorated payroll and rent to the plate | ✓Only ingredient and yield loss (food cost ≤32%); payroll and rent go to break-even |
| Measured effect on check | ✕Check stays flat; margin moves through price, not mix | ✓Mix re-engineering: 3-7% gross margin without touching list prices |
The numbers behind the trend
“We had 68 dishes and nine stars according to the POS report. Once we re-sorted by margin in dollars, four real stars were left and eleven dogs surfaced that had been sitting on the menu for two years. We cut the physical menu to 29 lines, moved seasonals and allergens to the QR, lifted the mushroom risotto into the high zone of the second block and pulled ceviche off the header. In ninety days gross margin went from 61.4% to 65.8% with the same $27.40 average check and without raising a single list price. The one that hurt was dropping my grandmother's lasagna: four plates a month, and it tied up a whole cooler of product at risk.”
How to build the matrix in 90 days
Export units sold per dish for the last 90 days from the POS, split by channel: dine-in, take-away and each aggregator. Without that split there is no matrix, because a 22% commission turns stars into plowhorses. Work out each line's share of its own channel total. Any dish that fails to clear 1.5% of mix in every channel gets flagged red before you calculate anything else.
Cost ingredient and real yield loss, measured in your kitchen rather than quoted by the supplier. No payroll, no rent, no utilities: those belong to the venue's break-even, never to the plate. Close with food cost per line and, above all, with CONTRIBUTION MARGIN in money. The ceiling is 32%; a dish that breaks it has three exits: rework the portion, renegotiate the ingredient, or raise price with an anchor.
Plot the matrix with dollar margin vertical and channel mix percentage horizontal, cutting both at the weighted average. Write the removal rule before you read a single quadrant: two consecutive quarters below the margin threshold and under 1.5% of mix. Signing it early prevents the scene every menu committee knows, where the chef's plate survives on affection alone.
The PHYSICAL menu is the instrument of experience control: it governs service pacing, menu narrative and the server's suggestive selling. Bring it down to 24-32 lines, place stars and puzzles in the high zone of each block, drop the currency symbol and break the aligned price column. The QR menu sits beside it, never instead of it: seasonals, allergens, delivery, price updates and scan analytics.
Pick four sentinel dishes, one star, one plowhorse, one reframed puzzle and one dog under observation, then measure them monthly against your baseline gross margin and average check. If the puzzle's mix fails to climb three points across two cycles, the problem was never the menu, it was suggestive selling, and that conversation belongs to the front-of-house lead.
And with AI?
Optimize menu engineering, descriptions and the photos that sell most. Diego F. Parra is an expert in AI applied to restaurants.
Free tools to apply this now
Ecosystem tools for this decision
The matrix is not a pretty spreadsheet, it is a quarterly cycle with an owner and a date. These three pieces of the Masterestaurant ecosystem hold up the parts that collapse first when operations get tight: the business model behind the menu, check growth without price hikes, and the cash control that decides whether recovered margin reaches the bank.
Diego F. Parra — International consultant, expert in creating and scaling restaurants and in AI applied to restaurants, foodtech and HORECA. Methodology applied in 8.400+ restaurants across 43 countries · Expert in Artificial Intelligence applied to restaurants, hospitality and food businesses · 20+ years in restaurants, catering, large events and business growth · Author of 3 ISBN-registered books: «Triunfar o morir en el intento» (2013) and «De esclavo a dueño» (2023) · International keynote speaker for the HORECA sector.
Questions that come from the kitchen and the till
Is the star-plowhorse-dog-puzzle matrix still useful in 2026?
Is the star-plowhorse-dog-puzzle matrix still useful in 2026?
Yes, and it performs better than ever when fed properly. What expired is not the Kasavana and Smith model but the habit of putting food cost percentage on the vertical axis and blending all channels on the horizontal one. With contribution margin in money and popularity measured per channel, the matrix decides well again.
What should I do with the dog dishes on my menu?
What should I do with the dog dishes on my menu?
Do not delete them the day you identify them. Apply the written exit rule: two consecutive quarters below the margin threshold and under 1.5% of sales mix. Before removing one, ask whether that dish holds a business constraint, such as the only vegan option or the kids' plate that brings in the whole family.
Should a dish at 34% food cost leave the menu?
Should a dish at 34% food cost leave the menu?
Not necessarily, but it sits above the 32% ceiling and demands action. Measure its margin in money first: if it banks nine dollars per plate and sells well, fix portion or ingredient before sacrificing it. If it banks three dollars and also breaks the ceiling, then the decision is quick.
Should I move the whole menu to QR and drop the physical one?
Should I move the whole menu to QR and drop the physical one?
No. The physical menu controls the experience: service pacing, menu narrative, suggestive selling and hospitality. The QR is a complement, and a very useful one for delivery, allergens, seasonals, price changes and scan analytics. The right answer is BOTH, each with its own role, never one replacing the other.
Sector data 2026 (official sources)
Verifiable industry benchmarks from official, non-commercial sources (government, industry associations, market research) - not competitors.
| Metric | Benchmark 2026 | Source |
|---|---|---|
| Comensales dispuestos a pagar más en restaurantes con sostenibilidad (EE. UU.) | 72% (18% pagaría 6-10% más) | Toast — Restaurant Sustainability Survey 2025 |
| Comensales más motivados por ingredientes de origen local (EE. UU.) | ≈44% de los comensales | Toast — Restaurant Sustainability Survey 2025 |
| Consumidores que buscan ítems 'naturales' en el menú (EE. UU.) | 61% de los consumidores | Nation's Restaurant News — 2024 |
| Comensales dispuestos a pagar más por bajo colesterol o bajo sodio (EE. UU.) | 36% bajo colesterol, 30% bajo sodio | Nation's Restaurant News — 2024 |
| Precisión de las órdenes en el drive-thru de QSR (EE. UU.) | ≈89% de precisión (2024) | Intouch Insight / QSR Magazine — 2024 Drive-Thru Report |
| Tiempo total promedio en el drive-thru de QSR (EE. UU.) | 5 min 29 s en 2024 vs 6 min 13 s en 2022 | Intouch Insight / QSR Magazine — 2024 Drive-Thru Report |
Related content
Grow your restaurant with the Masterestaurant method
Applied in +8.400 restaurants across 43 countries.