Menu Engineering: Find Your Winners and Fix Your Weak Sellers

What menu engineering measures, and what it does not

The classic method, as taught in culinary management training, works on one category at a time. For each item you calculate a food-only contribution margin: the selling price minus the standard portion food cost. Multiply that by units sold to get the item's total contribution. Add the totals for the category and divide by total units sold to get the weighted average contribution per item.

Popularity is each item's share of units sold in the category. The usual convention treats an item as popular if its share is at least 70 per cent of an equal share. With four active items an equal share is 25 per cent, so the cut-off is 0.7 divided by 4, or 17.5 per cent. Items at or above the weighted average contribution count as high contribution.

Two cautions belong right at the start. First, food-only contribution is not net profit. It leaves out labour, packaging, delivery commission, utilities and overheads, and those can differ sharply between items. Second, the 70 per cent rule is a convention for sorting items, not a business target and not evidence of demand. It tells you where to look, not what to do.

Prepare the data before you calculate anything

The analysis is only as reliable as its inputs, and most mistakes happen here rather than in the arithmetic. Decide the boundaries first, then collect the numbers.

• Compare like with like: one category, such as lunch mains, over one period and one sales channel. Dine-in and delivery often have different prices, packaging and customers, so analyse them separately.
• Use realised revenue per unit after discounts and excluding taxes collected for government, before channel fees that are outside this food-only comparison. Apply the same rule to every item; net bank settlements may already have fees deducted and need reconciliation.
• Use standard portion cost from a current recipe and current purchase prices. If an item has no tested recipe cost, mark it incomplete rather than guessing.
• Decide how to treat combos, variants, add-ons and complimentary items before you start, and write the rule down. A bundle sold at a discount should not silently inflate one component's sales or depress its price.
• Count active items that genuinely sold nothing, because a confirmed zero is real information. Do not count empty rows, discontinued items or placeholders.
• Note availability. An item that ran out on several days, or was only offered part of the period, will look less popular than its demand suggests.

Worked example: one lunch category

The figures below are illustrative only, invented for this example. They describe four lunch mains sold through dine-in over the same four-week period, using realised prices after discounts and excluding taxes.

Totals: 250 units sold, revenue of INR 55,400, food cost of INR 21,800 and total food-only contribution of INR 33,600. The weighted average contribution is INR 33,600 divided by 250, which is INR 134.40 per item. With four active items, the popularity cut-off is 17.5 per cent.

Read each result as a question, not a verdict

The veg bowl sells well and contributes above average. The temptation is to leave it alone, but its job is to stay reliable. The useful questions are whether it ever runs out at peak, whether portions are consistent across shifts and whether the queue or kitchen station it depends on is under pressure. Protect its availability and portion before thinking about price.

The paneer wrap is popular but contributes INR 130 against an average of INR 134.40. The gap is small, so start by checking the inputs: is the recipe cost current, is the paneer yield what the recipe assumes, and is the realised price being reduced by frequent discounts? If a change is needed, look at yield, recipe build or price openly. Do not quietly cut the portion, because regular guests notice and the item's popularity is the reason it matters.

Lemon rice sells below the cut-off and contributes below average. Before removing it, check whether it was available every day, whether its description is clear and whether it serves a purpose such as a lower-priced or simpler option some guests rely on. It may also share prep with other dishes, which changes the cost of dropping it. If it still underperforms after those checks, it becomes a candidate for redesign or replacement.

The mushroom bowl contributes the most per unit but sold only 10. That could mean it is hard to find on the menu, poorly described, priced above what this lunch crowd expects or simply unfamiliar. It is a candidate for a visibility test, provided the kitchen can handle extra demand and the ingredients do not create waste if sales stay low.

Factors the matrix cannot see

A menu is also an operating system. An item with a strong contribution may need a long prep, a scarce skill or a busy station, which can slow the whole service. Another may use trim from a different dish and reduce waste. Some items exist because a group of guests needs a dietary choice, and removing them can lose the whole table, not just one order.

Consider these factors alongside the classification. Record them in a short notes column so that the reasoning behind each decision is visible later, especially when the person who made it has moved on.

Run one small experiment at a time

Changing several items at once makes attribution difficult. Choose one defined change, run it over a comparable cycle, and compare it with a clear baseline. Use a similar duration and the same channel, and record seasonality, promotions, holidays and availability differences that may affect the result. A before-and-after comparison suggests what happened; it does not by itself establish causation.

For the mushroom bowl, test a more prominent menu position while keeping the recipe, price and service briefing stable. A service-description trial can follow separately if needed. Write down the experiment before it starts.

• Hypothesis: more visible menu placement will raise mushroom bowl sales without reducing total category contribution.
• Owner: the named manager responsible for recording results and deciding the outcome.
• Baseline: units, share and total contribution from the previous comparable cycle.
• Change: the single change being made, with its start and end dates.
• Success measure: total category contribution in rupees and the full mix, not the item's share alone. Higher mushroom sales that simply replace veg bowls may add little.
• Stop condition: what ends the trial early, such as rising waste, prep delays or complaints.

Common mistakes

Treating food-only contribution as profit is the most frequent error, followed closely by mixing channels or periods in one analysis. Others include using menu price instead of realised price, comparing items from different categories, removing an item because of its label alone, and ignoring stockouts that suppressed sales.

Another mistake is classifying items when the data is incomplete. If an item has no sales record for the period or no tested portion cost, leave it unclassified and fix the input first. A blank is more honest than a label built on a guess.