Menu engineering: how to use the stars, plowhorses, puzzles and dogs matrix

01What menu engineering actually measures

Menu engineering is a simple way to answer a hard question: which dishes on your menu are actually pulling their weight? It plots every item on two measures. The first is contribution margin, the cash each sale leaves after paying for the ingredients on the plate. The second is popularity, the share of guests who order that dish compared with everything else in its category. Put those two measures together and every dish lands in one of four quadrants, each with its own clear set of actions.

The method has been taught in hospitality schools for decades because it replaces gut feeling with arithmetic. Most chef-owners already have strong opinions about their menu: the lamb shank everyone raves about, the salad that never moves, the curry that keeps the lights on. Some of those opinions are right and some are expensive habits. A spreadsheet and a few weeks of clean sales data are usually enough to tell the difference, and the exercise often surprises even experienced operators.

It is worth being clear about what menu engineering does not do. It does not tell you whether a dish is delicious, whether it fits your concept, or whether it brings new guests through the door. It does not account for labor, which can make a high-margin dish costly to produce on a busy Saturday. Treat it as one strong input alongside your palate, your kitchen’s capacity and the dishes your regulars come back for.

02Collect item sales from your POS or sales records

You need two numbers for every dish: how many portions sold over the period, and what each portion costs you to make. Sales counts come from your POS or sales records. Most systems can export an item sales or product mix report as a spreadsheet. If you still work from handwritten checks, a tally from a few weeks of order pads will do, though it takes discipline to count accurately and consistently.

Choose the period with care. Four to twelve weeks is a practical window: long enough to smooth out a slow Tuesday or a large party, short enough that the menu and prices did not change midway. Avoid stretches dominated by a festival rush, a long closure or a promotion that distorted ordering, such as a Diwali set menu or a Christmas week. If the menu changed partway through, analyze only the stretch where it stayed stable.

Analyze one category at a time. Comparing a starter with a main course tells you nothing useful, because guests are not choosing between them. Group dishes the way guests choose: starters against starters, mains against mains, desserts against desserts. In a bar, cocktails sit apart from beer and wine. Exclude comps and staff meals from the counts, and treat dishes with several sizes consistently, either combined or split by size.

1. Export item-level sales for the chosen period from your POS or sales records, including units sold and selling price.
2. Remove voids, comps, staff meals and test items so the counts reflect what paying guests ordered.
3. Group the remaining items into the categories guests actually choose between.
4. Add a current plate cost for every item, using today’s supplier prices rather than last year’s recipe card.
5. Subtract any sales tax, VAT or GST included in the menu price so you work from net revenue.
6. Save the sheet with the period dates in the file name so the next review can be compared against it.

03Calculate contribution margin for every dish

Contribution margin is the selling price, net of tax, minus the plate cost. If a dish sells for 18.00 before tax and its ingredients cost 5.40, its contribution margin is 12.60. That 12.60 is what each sale contributes toward labor, rent, utilities and profit. Multiply it by the number of portions sold and you have the dish’s total contribution for the period, often the most revealing figure in the whole exercise.

Next, calculate the weighted average contribution margin for the category. Add up the total contribution of every dish, then divide by the total number of portions sold in that category. A simple average of the margins would give a rarely ordered dish the same weight as your bestseller and skew the benchmark. The weighted figure is the line every dish is measured against: above it is high margin, below it is low margin.

04Measure popularity against a fair share

Popularity is each dish’s share of the portions sold in its category, often called its menu mix. If a category sold 1,000 portions and one dish accounted for 250 of them, its menu mix is 25 percent. The question is where to draw the line between popular and unpopular. With five dishes, each would sell 20 percent if guests chose evenly, so a dish selling 18 percent is close to its fair share, not a failure.

The usual convention sets the threshold at 70 percent of an equal share. In a five-dish category the equal share is 20 percent, so the popularity line sits at 14 percent. In a ten-dish category the equal share is 10 percent and the line falls to 7 percent. The 70 percent factor acknowledges that some dishes will always outsell others, and it stops you labeling a respectable item unpopular simply because one bestseller dominates.

Watch for distortions. A dish that sold out early every night because of a supply problem will look less popular than it really is. A dish that servers push because it is quick to plate will look more popular. Note those cases next to the numbers. Seasonal dishes deserve the same caution: an alphonso mango dessert analyzed in May and again in October will tell two very different stories.

05Sort dishes into stars, plowhorses, puzzles and dogs

With margin and popularity in hand, every dish falls into one of four quadrants. Stars are high margin and high popularity. Plowhorses are popular but earn below-average margin. Puzzles earn a strong margin but sell below the popularity line. Dogs are low on both. The names are memorable on purpose, so the whole team can discuss the menu in a shared language without anyone feeling their favorite dish is being attacked.

The worked example below uses a five-dish category that sold 1,000 portions over eight weeks. The popularity line is 14 percent, or 140 portions, and the weighted average contribution margin works out to 8.83 in the restaurant’s currency. The figures are illustrative, but the pattern is typical: two stars carry the category, a plowhorse moves volume, a puzzle earns well but rarely gets ordered, and one dish is quietly taking up space.

Plotting the result on a simple scatter chart, with popularity along the bottom and margin up the side, makes it easier to explain to partners and staff. Draw the two threshold lines, label each dot with the dish name, and the menu’s strengths and weak spots become visible at a glance. Keep each review’s chart, because placing two quarters side by side shows which dishes are drifting toward a weaker quadrant long before the change becomes obvious on a single night’s takings.

The names are memorable on purpose, so the whole team can discuss the menu in a shared language without anyone feeling their favorite dish is being attacked.

06What to do with each quadrant

The quadrant is a starting point for a decision, not the decision itself. Before acting, ask why a dish sits where it does. A puzzle might be a great dish described badly, a plowhorse might be underpriced compared with its neighbors, and a dog might be the only vegan main on the menu. The aim is to move dishes toward the star quadrant, or to free their space for something that will earn it.

Change one or two variables at a time and keep notes. If you reprice, rewrite and move a dish in the same week, you will not know which change worked. Small, deliberate moves also protect the guest experience, because regulars notice when a favorite suddenly costs more, looks smaller and has moved to the back page all at once. A simple log with the dish, the change, the date and the reason is enough to keep the next review honest.

07Turn the analysis into menu changes

Once you know what you want each dish to do, use the levers on the menu itself. Position, description, price, portion and recipe are the main tools, and they work best alongside a short briefing for the front-of-house team. Write a plan that lists each dish, its quadrant, the change you will make and the date you made it, so the next review can test whether the change actually worked.

Give every change enough time. A rewritten description or a new position usually needs several weeks of service before the sales data means anything, especially in a restaurant with many regulars who order by habit. Resist reversing a decision after one slow weekend, and look at the trend across the full review period instead. Weather, local events and school holidays all nudge ordering, so compare like with like wherever you can.

• Place stars and puzzles where the eye lands first in each category, such as the top of a list or a lightly boxed feature.
• Rewrite puzzle descriptions with concrete ingredients, technique and origin rather than generic praise.
• Rebalance plowhorse plates by adjusting garnish, sides or the portion of the costliest component, then re-cost them.
• Drop dotted leader lines and right-aligned price columns that invite guests to shop by price alone.
• Offer a second, smaller size where a lighter format could convert hesitant guests.
• Brief servers on which dishes to recommend and let them taste those dishes before service.
• Retire dogs with no strategic role and use the space to trial a new dish.

08Set a review cadence and keep the menu honest

Menu engineering works best as a routine rather than a one-off project. A full analysis every quarter suits most restaurants, with a lighter monthly look at the top sellers and any dishes you changed. Run an extra review after a big supplier price increase, a menu-wide price change, a seasonal rotation or a new concept launch. Each time, compare the new matrix with the previous one and note which dishes moved quadrants and why.

Between reviews, keep the guest-facing menu accurate. A menu that lists dishes you cannot serve undermines the whole exercise, because the guest who wanted the lamb shank ends up disappointed rather than steered toward another star. If you run a digital menu, a tool such as Menulo lets owners and managers mark a dish or a single portion sold out, rewrite a description or adjust a price in minutes, while the sales figures for your analysis still come from your POS or sales records.

Share the results with the kitchen and floor teams. Cooks who understand why the lamb shank matters are more careful with its portion, and servers who know which dishes are puzzles can recommend them with genuine enthusiasm. A short monthly menu meeting before family meal is often enough to keep everyone pulling in the same direction. Use it to taste any reworked dish together, so every server can describe it from experience rather than from a list.

09Common menu engineering mistakes to avoid

Most problems with menu engineering come from bad inputs or impatient conclusions rather than from the method itself. The arithmetic is simple; the discipline lies in costing accurately, choosing a clean period and resisting the urge to change everything at once. It also helps to remember that the matrix describes the past. It shows how guests responded to last quarter’s menu, prices and service, and it cannot predict how a new dish will perform until that dish has sold for a few weeks.

Keep the spreadsheet simple enough that someone else on the team could update it next quarter. A model that only the owner understands tends to be abandoned the first time the owner gets busy, and an abandoned analysis is worth nothing. One tab for raw sales, one for plate costs and one for the matrix is usually plenty, with the formulas visible so a sous chef or manager can follow the logic and spot a costing error.

• Using outdated plate costs, which turns stars into plowhorses on paper and hides real margin problems.
• Comparing dishes across categories instead of within the set a guest is actually choosing from.
• Ranking dishes by food cost percentage instead of contribution margin in cash.
• Ignoring labor and station load, such as a puzzle that ties up the grill cook for ten minutes at peak.
• Cutting every dog without checking dietary coverage, shared prep or group-dining needs.
• Changing price, description and position together, so you cannot tell which change worked.