A guide to menu engineering: 12 data-driven tips to maximise profits

Menu engineering is the process of optimising your menu by analysing the relationship between each dish's profitability and its popularity. The goal is a menu where your highest-margin dishes are the ones customers choose most often and where the items that are dragging down your margins are either fixed or removed. 

As the popularity of items in a restaurant is heavily influenced by seasonal variation, menu engineering, and analysis is often an ongoing activity to help make dynamic decisions.  

The standard tool for menu engineering is a four-quadrant matrix that places every dish based on its margin and its sales volume. 

Stars are high profitability, high popularity. These are your best-performing dishes, relatively inexpensive to make and consistently chosen by customers. They do not need to be changed, but they do need to be actively protected. If a key ingredient rises in price, the margin on a Star can erode quickly without anyone noticing. Monitor their cost closely and promote them prominently. 

Puzzles are high profitability but low in popularity. The margin is there, but customers are not choosing them often enough. The fix is usually in visibility, eg. better placement on the menu, stronger descriptions, or staff recommendations, rather than in the dish itself. 

Plowhorses are high in popularity but lower in profitability, typically because their ingredient costs are high relative to their menu price. Customers like them, which makes them worth keeping, but the economics need managing. Small adjustments to portion size, ingredient specification, or price can often improve the picture without affecting appeal. 

Dogs are low in profitability and low in popularity. These are costing you more than they are contributing. Unless there is a specific brand or menu-balance reason to keep them, the default decision should be removal.

Before you can engineer a menu, you need accurate sales data at dish level. Not category level, not weekly revenue, but which specific dishes sold, in what quantities, on which days and at which sites. This is the foundation of everything else. Without it, you are placing items in the matrix based on instinct, and the matrix is only as useful as the data behind it. 

Pull your sell-through data for a consistent period, at minimum four weeks, ideally a full quarter to smooth out weekly variation. If you have multiple sites, run the analysis separately for each one. A dish can be a Star at one location and a Dog at another and understanding why that difference exists is often as valuable as the classification itself. 

The most common mistake in menu engineering is calculating dish profitability from recipe costs that have not been updated since the menu launched. If a key ingredient has risen 15% in price since the recipe was built, the GP figure you are working from is wrong and any decision you make based on it may be wrong too. 

Dish GP should be calculated from current ingredient prices, not historical ones. That means your recipe management needs to be connected to your live purchasing data. For each dish: selling price minus current ingredient cost, divided by selling price, gives you GP percentage. Run this across your full menu and you will likely find that several dishes have drifted significantly from where you thought they were. 

Waste is not just a cost control problem - it is also a menu engineering signal. A dish that consistently generates high prep waste or spoilage may be telling you something about its design: it requires specialist ingredients that are not used elsewhere, it demands more prep than the kitchen can execute consistently at volume, or the portion specification is generating plate waste because it is larger than customers want. 

Track waste at dish level where possible, or at ingredient level and map it back to the recipes that use those ingredients. Items with high associated waste are candidates for redesign or removal regardless of where they sit on the popularity axis, because the true cost of running them is higher than the recipe cost alone suggests. 

The matrix is most useful when it reflects normal trading patterns rather than anomalies. A dish that sold poorly for three weeks because it was a hot summer and it is a heavy winter warmer should not be classified as a Dog based on that data alone. Equally, a dish that spiked in sell-through during a promotional period is not necessarily a Star in ordinary trading. 

Use your historical sales data to understand the demand patterns that are specific to your business - which dishes peak at which times of year, which days of the week drive volume for particular categories, how your sales mix shifts between lunch and dinner service. Build your matrix analysis on a period that represents typical trading, and factor seasonal variation into your decisions about what to keep and what to remove. 

A price file that is updated monthly, or whenever someone remembers to do it, is not sufficient for meaningful menu engineering. By the time a quarterly menu review arrives, a supplier may have implemented two or three price changes that have each individually seemed minor but together have moved a dish from a 68% GP to a 61% GP. Across a menu of 40 dishes, that kind of drift across multiple ingredients is very difficult to track manually. 

The fix is to connect your supplier pricing directly to your recipe costs so that every invoice price change flows through to the dish margin automatically. This transforms menu engineering from a periodic project into a continuous process - the margin picture is always current, and decisions can be made when they are needed rather than when a review is scheduled. 

Individual dish performance matters, but so does the composition of your overall sales mix. A menu where the majority of covers are driven by low-margin Plowhorses may show healthy revenue while delivering disappointing profit. The sales mix view tells you what percentage of your total volume each dish is contributing, weighted by its margin, which gives you a much clearer picture of where your profit is coming from. 

Run your sales mix analysis at least monthly. Look for shifts in the mix over time: if a high-margin Puzzle has been slowly gaining traction and is approaching Plowhorse popularity levels, that is worth accelerating through promotion. If a Star's share of covers is declining, it is worth investigating whether a pricing or presentation change is behind it. 

Most recipes are not as simple as a single ingredient from a single supplier at a single price. Proteins come in different cuts and sizes, dairy products have multiple specifications, seasonal produce changes supplier and specification throughout the year, etc. If your recipe management only captures one version of each ingredient, your costings will be wrong every time the kitchen uses a different variant, which, in practice, is often. 

Grouped Products let you capture every variant of an ingredient -different pack sizes, different suppliers, seasonal alternatives, and assign them to a single recipe component. The recipe cost then reflects what is in the kitchen rather than a theoretical ideal specification, and when a substitution is made, the cost updates automatically. 

The matrix is a classification tool, not a decision in itself. Once you have placed your dishes in their quadrants using current margin and sell-through data, the value comes from acting on what you find. Each quadrant has a default action: 

There is no single food cost percentage that is right for every dish on every menu. A cocktail and a slow-braised short rib have completely different cost structures and different customer price expectations. Applying a blanket GP target across your full menu will push you towards over-pricing simple dishes and under-pricing complex ones. 

Build margin targets by category: starters, mains, desserts, drinks, and use those category targets as the benchmark for your matrix analysis rather than a single overall figure. This also makes it easier to understand the contribution of each category to total menu GP, and to make deliberate decisions about how you want the mix to look. 

Cost increases from suppliers are rarely announced far in advance, and they are rarely uniform. A 5% rise in one protein, a 3% rise in a key dairy ingredient and a 10% rise in a seasonal vegetable can each seem manageable individually, but if they all affect the same three dishes, those dishes' margins can move significantly in a short period. 

A practical way to manage this is to build a sensitivity assumption into your margin model: review what happens to each dish's GP if key ingredient costs rise by a given percentage. This is not forecasting - it is a stress test. It tells you which dishes are most exposed to cost movement, which suppliers represent the most concentration risk, and where a price change or recipe modification now would protect you against a move you can see coming.

Menu engineering is not just about numbers; it is also about making sure the products behind each dish fit your concept and your guests’ expectations. A neighbourhood café might prioritise affordability and familiarity, while a premium casual brand leans into higher‑quality, local or sustainable products that customers are willing to pay more for.  

Build regular sense‑checks into your process – market research, reviews, and quick customer surveys – so you understand whether your guests care most about price, provenance, sustainability, or indulgence, and reflect that in the products you source. 

Guests are increasingly looking for operators whose sustainability claims are reflected in what appears on the plate, not just in marketing copy: surveys show that around a third of Brits are more likely to choose venues that are transparent about their environmental impact, and over half want restaurants to make sustainable choices easier at the point of ordering. 

Choosing products with stronger sustainability credentials – better traceability, lower carbon impact, or more ethical supply chains – can strengthen trust with your regulars and support a premium positioning when it is backed up by a clear story on the menu.  

The challenge is doing this consistently across a full menu without adding a layer of admin that the kitchen and procurement teams cannot realistically maintain.

The tips above describe a data-driven process. The limitation of running that process manually is time: gathering current cost data, updating recipe costings, pulling sales figures and building the matrix takes hours, and by the time it is done the underlying data has already moved on. 

Procure Wizard Evo Recipes removes the manual data-gathering step. Ingredient costs are driven by live stock levels and current supplier prices, so recipe GP reflects what is on hand, not a historic average. Grouped Products capture every variant of an ingredient, which means substitutions do not create silent cost variances, and FIFO stock management keeps costs aligned to real usage.  

On top of that, the reporting suite covers sales mix, GP by dish and by site, and waste analysis, with the ability to build custom reports from any data point across the platform. The practical result is that the decisions described in the tips above – promote, reprice, delist – can be made when the data says they are needed, not just when a scheduled menu review comes around. 

Ready to put this into practice? 

Menu engineering works best as a habit rather than a project. The 12 tips in this guide give you a practical framework for building that habit, starting with accurate, current data and ending with clear decisions about every dish on your menu. 

If you would like to see how Procure Wizard Evo Recipes connects your menu engineering to live procurement and stock data, take a look at our recipe and menu engineering module or speak to the team.