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

Chapter 1 — What is the classic menu-engineering matrix missing?

The Kasavana and Smith matrix classifies the past but never predicts how demand reacts when you move the price.

Born in 1982, it plots popularity against contribution margin across four quadrants —star, plowhorse, puzzle, dog— and still remains the only tool most operators use. The flaw is architectural: it tells you a dish sells poorly, not the price at which it would start selling for a specific segment. And menu layout matters more than most think: 71% of guests decide their order based on menu design and placement, according to OneHubPOS (Menu Engineering 2024). In 2026, with food-cost inflation still volatile, classifying yesterday does not protect tomorrow's margin. The photograph is well taken; the film is missing. Second generation keeps the matrix as its base layer and overlays a price-elasticity model that segments demand by guest, occasion and channel. The first generation optimizes the photo of the menu; the second models the film of how demand shifts when you change the price.

Chapter 2 — What does second-generation menu engineering involve?

Instead of one contribution margin, you cross that margin with distinct sensitivity curves: the midday business diner barely notices a one-dollar increase, while the Saturday occasion diner clearly does.

The behavioral evidence backs the approach: 56% of guests chose dishes with descriptive labels in Cornell's study (Wansink, Food & Brand Lab). That same sensitivity to framing is what 2.0 models with data, not intuition. You do not replace the matrix: you turn it into the first layer of a living model. Most restaurants leave 2 to 4 margin points hidden because they charge one price to demand curves that look nothing alike. This is Diego F. Parra's central thesis at Masterestaurant, and I have seen it identically in single-location operations and in twenty-unit chains. Willingness to pay varies by attribute with enormous gaps: 72% of guests would pay more for sustainability and 18% would accept a 6% to 10% premium, according to Toast (Restaurant Sustainability Survey 2025); 38% pay more for protein-rich dishes, according to Nation's Restaurant News (2025).

Chapter 3 — How much margin does a single price hide?

A flat price ignores both populations at once. The second generation does not raise everything: it identifies which dishes tolerate the increase for which segment and leaves the sensitive ones intact.

Those 2 to 4 points are the difference between surviving and growing. Engineering 2.0 requires crossing the POS with the CRM and the time slot, because the classic matrix works only with sales mix and that is not enough to segment the curve. The POS tells you what sold; the CRM and the hour tell you who bought it and on what occasion, which is where elasticity lives. Without that cross you cannot separate the midday diner from the weekend one, and 42% of younger guests share a main dish more frequently, according to Acosta Group (2025): a pattern that alters the ticket per person and that an aggregated POS hides. Adding the time slot reveals another layer: 37% of consumers seek quick bites instead of large meals, according to Circana (2024).

Chapter 4 — What data does the model need that the POS alone lacks?

Modeling without those axes is averaging three different restaurants inside the same location and setting one single price.

A dish the classic matrix flags as a 'dog' —low popularity, low margin— stops being one when you re-price or reposition it for the right segment, not when you delete it. The classic version tells you which dish is a dog; 2.0 tells you at what price that same dish stops being a dog for a specific diner. Framing moves the needle: 56% of guests chose options with descriptive labels at Cornell (Wansink), and more than half of consumers lean toward a dish labeled 'spicy' in 2025 versus 39% in 2015, according to Datassential (Spicy Food Trends 2025). A 'dog' with the right label and a price calibrated to its occasion can become the margin that rescues the menu. Deleting it destroys an option that was merely miscommunicated and mispriced, not dead.

Chapter 5 — Why does aggressive dynamic pricing destroy restaurant margin?

Airline-style dynamic pricing is the trap of a poorly executed second generation: it punishes the perception of fairness and drives the diner away.

The data is blunt: 52% of consumers see dynamic pricing in restaurants as 'price gouging' and 36% would order less frequently if it is applied, according to Capterra (2024 survey). A well-built second generation is not surge pricing: it is structural segmentation by menu, channel and occasion —lunch prices distinct from dinner, a delivery menu with its own architecture— that the diner perceives as coherent. Diego F. Parra sums it up this way at Masterestaurant: you model elasticity to set the right price per segment, not to change it every hour. The same sensitivity that makes people pay more for value —44% are motivated by locally sourced ingredients, according to Toast 2025— resents anything that smells like manipulation. Elasticity yes; visible opportunism, never. Implementation is layered: first the classic matrix as diagnosis, then the data cross, then elasticity modeling and finally differentiated prices by segment.

Chapter 6 — How is it implemented in layers without collapsing operations?

Starting with surge pricing is the error that leads to the 52% who perceive it as abuse (Capterra 2024); starting with a clean diagnosis avoids that blow.

The data layer requires joining POS, CRM and time slot; the elasticity layer, reading willingness to pay by attribute —61% seek 'natural' items, according to Nation's Restaurant News (2024), and 1 in 3 would pay more for plant-forward dishes, according to Datassential (2024)—. Each layer is validated before the next, measuring real margin per segment against the model. This way the chef-owner does not risk the operation: they build a system that, layer by layer, recovers the hidden 2 to 4 points without touching the diner's sense of fairness or the coherence of the menu. The first generation optimizes the PHOTO of the menu; the second models the FILM of how demand reacts when you move price. The classic uses contribution margin as its axis; 2.0 crosses it with price elasticity, which differs for the weekday-lunch business guest and the Saturday-night occasion guest.

Chapter 7 — What actually changes between generations

The classic works off POS data; 2.0 needs to cross POS with CRM and daypart to segment the demand curve. The classic tells you which dish is a 'dog'; 2.0 tells you at what price that same dish stops being a dog for a specific segment.