Restaurant Menu Engineering UK 2026: Stars, Dogs and Profitability

Menu engineering is the systematic analysis of a restaurant's dishes using two variables: their contribution margin (what the restaurant earns per dish sold) and their popularity (how many are sold in a period).

The result is a classification of every dish into one of four categories:

• Stars: high margin and high popularity. The dishes that make money and customers enjoy. Protect and give prominence.
• Plough-horses: low margin and high popularity. Sell well but leave little. Candidates for recipe reformulation or gradual price increase.
• Puzzles: high margin and low popularity. Would be profitable if they sold more. Review positioning, description and price.
• Dogs: low margin and low popularity. Candidates to remove or replace, especially if they complicate kitchen operations.

The framework was developed by Michael Kasavana and Donald Smith at Cornell University and has been applied in hospitality since the 1980s. It remains the standard approach because it is actionable: every dish gets a direction, not just a number.

Contribution margin per dish is the difference between the selling price (ex-VAT) and the food cost of the ingredients:

Contribution margin = selling price (ex-VAT) − recipe ingredient cost

The ingredient cost comes from the recipe card: every ingredient in the dish (including preparation waste yield) costed at actual purchase price.

Example:

• Ribeye steak with chips: selling price £28.00 (ex-20% VAT: £23.33) − ingredient cost £8.40 = contribution margin £14.93
• Seasonal soup: selling price £7.50 (ex-VAT: £6.25) − ingredient cost £0.85 = contribution margin £5.40

To decide whether a margin is "high" or "low", calculate the weighted average contribution margin across the whole menu (sum of (margin × units sold) / total dishes sold) and compare each dish to that average. A dish above average is "high margin"; below average is "low margin".

Popularity is measured by the popularity index (PI):

PI = units sold of the dish / total units sold in the category

A dish is considered popular if its PI exceeds 70% of the expected uniform popularity (1 / number of dishes in the category × 0.70). With 8 main courses and 240 sold in a period, uniform popularity is 12.5% and the popularity threshold is 8.75%. A dish that does not reach that threshold is classified as low popularity regardless of its margin.

Sales data by dish comes from the EPOS. Without dish-level sales data, menu engineering relies on team estimates — useful as a starting point but much less precise than actual data.

The decisions that follow from each classification:

• Stars: do not change the recipe. Give good visibility on the menu (prime reading position, strong description). A small, gradual price increase is usually well tolerated if the dish has strong perceived value. Protect consistency so the margin does not erode through portioning drift.
• Plough-horses: look at the recipe for cost reduction opportunities that do not change the customer's perception. Consider a modest price increase. Reposition in the menu to give less prominence and steer customers toward higher-margin alternatives.
• Puzzles: check whether poor visibility is the problem — these are often buried in the middle of a long list. Improve the name or description. Review whether the price creates a barrier. Try as a daily special to test whether exposure increases volume.
• Dogs: assess whether they belong on the menu at all. If the operation is complex and the margin low, removing them simplifies the kitchen and may increase average margin across the remaining dishes. If there are good reasons to keep them (a signature dish, a dietary requirement anchor), at least do not give them prominence.

Menu engineering is not a once-a-year exercise. Popularity data changes seasonally, ingredient costs shift and customer preferences evolve.

A practical cadence for most UK restaurants:

• Full analysis with each seasonal menu change (two to four times a year): reclassify all dishes using sales data from the period and update recipe costs to current purchase prices.
• Monthly food cost check: compare actual food cost for the period to theoretical food cost (recipe cost × dishes sold). A meaningful gap indicates a problem with recipe accuracy, waste or a shift in the sales mix — all of which change the menu engineering picture.
• Price review when ingredient costs shift significantly: if a key ingredient rises more than 10%, revisit the dishes that use it most. A plough-horse can become a dog if ingredient costs rise without a corresponding price adjustment.

How many dishes do you need on a menu for menu engineering to work?

Menu engineering works with any size of menu but is most reliable when each category has at least 6–8 dishes, giving statistically meaningful popularity data. With fewer dishes, small changes in sales volume can flip a dish between categories. With very long menus (15+ dishes per category), the analysis typically reveals more candidates for removal than protection — long menus reduce the average popularity of each dish and complicate kitchen operations. Menu engineering often leads directly to a shorter, more profitable menu.

Does menu engineering work the same way for set menus and tasting menus?

The framework applies differently to set menus. For a prix-fixe or tasting menu, the relevant unit of analysis is the whole menu, not individual dishes. Contribution margin is the menu price minus total food cost for all courses. Popularity is the cover count versus alternatives (à la carte, other menu tiers). Within a tasting menu, ingredient cost by course can still be tracked, but the customer does not choose courses so popularity as a driver of behaviour does not apply in the same way.