Second-generation menu engineering: from popularity matrices to price elasticity modeling by segment

Chapter 1 — Why the classic matrix no longer protects margin in 2026

The popularity-margin matrix stops protecting margin because it is a static diagnosis born in 1982, back when inputs rose 2-3 % a year. Kasavana and Smith, at Michigan State University, sorted every dish as Star, Plowhorse, Puzzle or Dog along two fixed axes: how much it sells and how much it leaves. The model was brilliant for its time and remains a good starting point. The problem in 2026 is double-digit volatility in proteins and dairy: the FAO food price index has posted year-over-year jumps above 20 %, and the USDA reports persistent food-away-from-home inflation. A dish tagged Star in January can slide to Puzzle by March without anyone touching the menu. I have seen it in dozens of kitchens: a cut's food cost jumped from 28 % to 41 % in one quarter while the matrix still flagged it green. A map that updates once or twice a year leaves money on the table every week.

Chapter 1 — Why the classic matrix no longer protects margin in 2026 — in practice

Masterestaurant's second generation does not discard the sales mix or contribution margin; it keeps them as the starting point and adds three layers the static snapshot ignores by design: elasticity by segment, cost variance, and scenario stress. The core difference is that the classic matrix describes while second generation prescribes: it tells you how many cents to raise on each dish, not just which quadrant it sits in. A diagnosis that only paints four boxes leaves the pricing call to the owner's gut, and gut does not defend margin under double-digit inflation. The prescriptive model closes that gap with an operable rule: each dish gets an adjustment computed from its elasticity, its cost variance and the inflation scenario you run. Compare the two by decision variables —the matrix uses 2 (popularity and margin); second generation uses 5 (popularity, margin, elasticity, variance, scenario)— and by frequency —1-2 times a year versus monthly on 60 % of the mix.

Chapter 2 — From diagnosis to prescriptive model: how much to raise each price

In practice, a Plowhorse (high sales, low margin) that tolerates a 6-8 % bump recovers 3-4 points of operating margin with no measurable traffic loss. Diego F. Parra puts it this way in Masterestaurant board meetings: the matrix tells you where it hurts, the model tells you how hard to squeeze. Running the model every 30 days instead of every 12 months turns margin into a variable you defend, not one you endure. That cadence change is half the value. Elasticity turns the sales mix into a lever because it measures how much each dish's demand falls when you raise its price, and that number varies brutally across dishes. It is the first of the three new layers and the one that unlocks the most margin without touching a single recipe. An anchor dish, heavily ordered and loaded with perceived value, usually shows low elasticity: raise it 10 % and you lose 2-3 % of units, leaving more net margin.

Chapter 3 — Price elasticity: not every dish tolerates the same adjustment

A commodity dish, comparable to the one down the street, shows high elasticity: the same 10 % costs you 12-15 % of units. The classic matrix treats both the same, and that is its most expensive mistake. At Masterestaurant we measure elasticity by customer segment over 8-12 weeks of real sales, not by hunch, cross-referencing the POS price history with units sold. A three-band map orders the menu: inelastic (<1, raise hard), unit (≈1, fine-tune), elastic (>1, freeze or cut). The practical result: you push hard where the customer does not react and freeze the price where they do, capturing 2-5 points of margin. That segmentation is what the static snapshot never saw, because the matrix only looks at the room average, not the sensitivity dish by dish. Portion costing with a standard recipe is the base of the model, because without controlled variance any price rests on false figures.

Chapter 4 — Variance between theoretical and real cost: the base without which everything collapses

This is the second layer, and though it sounds less glamorous than elasticity, it is where 80 % of lost margin lives. Variance is the gap between your theoretical cost —what the recipe says the dish costs— and your real cost —what the register records at close. In kitchens with no control I see gaps of 4-7 %: waste, portions by eye, petty theft, unlogged substitutions. If your theoretical cost says 30 % and the real is 36 %, any price adjustment computed on the theoretical figure is born broken: you are optimizing over a number that does not exist. Remember the hard Masterestaurant costing rule: 32 % food cost per dish is the ceiling, not the target, and payroll and rent are not charged to the dish —they go to the break-even point. The first step of second generation is not raising prices: it is closing variance below 2 % with a weighed standard recipe and weekly inventory counts.

Chapter 4 — Variance between theoretical and real cost: the base without which everything collapses — in practice

Only then does the elasticity model have true data to chew on. Diego F. Parra insists 80 % of lost margin is not in the menu, it is in the gap nobody measures, which is why the phase order matters: variance first, price second. Pricing psychology multiplies the effect of the adjustment because it changes what the customer perceives without changing a gram of the dish. Three concrete levers: anchoring, price endings and position on the menu. Anchoring places a premium dish up top so the rest reads as reasonable; a $48 anchor makes the $29 dish read as value. Endings pay off differently by category: in fine dining, a round price with no symbol ($34, not $33.99) lifts the perceived ticket. Position matters: the eye lands first on the top-right corner and on the first dish of each block, where your highest margin belongs, not your cheapest plate.

Chapter 5 — Pricing psychology and ticket by segment: multiplying the effect without touching the recipe

These tactics move 3-6 % of the average ticket at zero input cost. The average ticket, moreover, stops being a blind number when you break it down by segment, because a $32 average can hide a $58 segment and a $19 segment that cancel each other out. Breaking it down by business diner at midday, weekend family, and delivery reveals where margin truly grows and where there is only volume with no profit. In a real Masterestaurant case, the delivery segment represented 40 % of orders and only 12 % of operating margin: lots of noise, little cash. By reweighting the mix and adjusting prices by channel, that same location lifted operating margin 4 points in one quarter without raising total volume. The classic matrix ignores the menu as a selling interface and treats the customer as a single average; second generation treats it as the best-paid salesperson in the room and reads the customer by segment.

Chapter 6 — Scenario stress: defensible margin under uncertainty, not a chart for the wall

Inflation scenario stress is the third layer that makes margin defensible, because it simulates what happens to your mix when the input jumps before it actually jumps. Instead of reacting when the supplier raises protein 15 %, you run that shock today against your current menu and see which dishes flip from profitable to red. The stated goal of the second-generation model is a defensible operating margin under uncertainty, not a snapshot that ages on the wall. In practice, I define three scenarios —base, tension (+10 %) and crisis (+20 %)— and for each one the model prescribes which prices to move and which dishes to redesign or retire. This three-scenario map crosses with the three elasticity bands from Chapter 3: an inelastic dish in the crisis scenario is your first pricing lever; an elastic dish that turns red at +20 % is a candidate for portion redesign or retirement, not a hike.

Chapter 6 — Scenario stress: defensible margin under uncertainty, not a chart for the wall — in practice

A location that runs this exercise every quarter absorbs an input shock with 2-3 weeks of lead time, not with two months of margin bleeding out. Diego F. Parra calls it budgeting for the scare: volatility is not predicted, it is rehearsed. With the six chapters chained —diagnosis, prescription, elasticity, variance, psychology and stress— the menu shifts from decorative chart to a living margin system. The group recovered 3.1 points of mix-weighted contribution margin in 90 days by applying the full model to its 10 main dishes. Starting point: classic matrix pinned in the office, average food cost of those 10 dishes at 31 %, unmeasured variance, and a $32 average ticket treated as a single number.

Chapter 10 — Quantified mini-case: 4-location full-service group in 90 days

The sequence was the method's: weeks 1-2, standard recipe and actual costing, which revealed a 5.4 % variance (real 36.4 % against theoretical 31 % on four dishes); weeks 3-6, elasticity measurement by segment over 10 weeks of sales, which classified 6 dishes as inelastic and 2 as elastic; weeks 7-10, variance reconciled below 2 % and stress at 5 %/12 %/20 %; weeks 11-12, execution. Six inelastic prices were raised between 6 % and 9 % and two elastic ones cut by 4 %. Result at 90 days: food cost of the 10 dishes from 31 % to 28 %, mix-weighted contribution margin +3.1 points, average ticket +4.80 USD, with no measurable traffic drop and no customer complaints. The key, per the operations director, was not raising prices: it was knowing which to raise and which to protect. That discernment is exactly what the 1982 matrix cannot give. The second-generation model fails when there is no quality data, and it must be said plainly: it is not magic, it is data discipline.

Chapter 11 — Limitations, assumptions and failure conditions of the model

Assumption #1: a price history per dish exists or a controlled 2-week test can be run. In a freshly opened venue, with under 8-12 weeks of sales, elasticity is estimated with a larger margin of error and prices should move with more caution. Assumption #2: variance can be reconciled, which demands weekly inventory and a weighed standard recipe; without that discipline, the model optimizes over false figures and does more damage than the matrix. Assumption #3: volume per dish is enough for elasticity to be statistically legible —dishes selling 1-2 units a day give no reliable signal and are grouped by category. Known failure conditions: external demand shocks (a competitor opening, a seasonal event) contaminate the elasticity reading and force recalibration; and a perceived-quality change (shrinking the portion to lower food cost) alters future elasticity and can turn a previously inelastic dish elastic. The food cost ceiling remains 32 % per dish: no elasticity model authorizes breaching that ceiling by raising portion at the expense of margin.

Chapter 12 — Limitations, assumptions and failure conditions of the model — in practice

That is why Diego F. Parra insists on treating the model as a living system: it is monitored at 3, 6, and 12 months and recalibrated, not decreed once and forgotten. The classic matrix is a diagnosis; second generation is a prescriptive model that says how many cents to raise each price, dish by dish. Demand elasticity turns the sales mix into a lever: an inelastic dish tolerates +10 % losing only 2-3 % of units; an elastic one loses 12-15 % on the same move. Per-portion costing with a standard recipe is the base; without variance controlled below 2 %, any pricing model rests on false numbers (the typical uncontrolled gap is 4-7 %). Pricing psychology (anchoring, price endings, menu position) multiplies the effect of the adjustment 3-6 % of the ticket without touching the recipe or the food cost. The average ticket stops being a blind average —a $32 can hide $58 and $19— and breaks down by segment to reveal where real margin grows.

Chapter 13 — The differences that define margin

Scenario stress rehearses the shock before it happens: 3 scenarios (base, +10 %, +20 %) leave the trigger price defined with 2-3 weeks of anticipation.