Menu Engineering: Which Dishes Actually Make You Money
Most chefs can name their best-selling dish. Far fewer can name the dish that's quietly carrying the entire margin of the business. The gap between those two answers is where menu engineering lives.
Here's a small exercise. Look at last month's POS report. Pick the dish you think is your biggest contributor to the kitchen's bottom line. Now do the actual math: units sold × (price − food cost). The dish you guessed is rarely the dish on top. Usually it's something cheaper, less glamorous, and ordered more often than you remember — the side, the burger, the simple pasta.
Menu engineering is the discipline of looking at every dish through two lenses at once: how often it sells (popularity) and how much it contributes per sale (gross margin in dollars, not percent). Plot every dish on those two axes and you get a four-box matrix that tells you exactly what to do with each item. It's not new — the framework was formalised by Michael Kasavana and Donald Smith at Michigan State University in 1982, building on hospitality research at Cornell — but most independent kitchens still don't run it.
This is the working version: how to actually do it on a Sunday afternoon, what the four boxes mean in practice, and the awkward decisions the matrix forces you to make.
The two numbers that matter
For each dish on your menu, you need two figures from the same time period (a month works; a busy week works too):
- Units sold — straight from your POS. If your POS doesn't export this, tally manually for a representative week.
- Gross margin per unit (in dollars, not percent) — selling price minus accurate food cost. Note: dollars, not percent. A 50% margin on a $10 dish is $5. A 30% margin on a $40 dish is $12. The second contributes more to paying your rent, regardless of percentage.
If you've ever calculated food cost percentage and stopped there, this is why menu engineering matters. Two dishes can have identical food cost % and contribute wildly different amounts of cash. The matrix forces you to see both.
The four boxes
Plot popularity on one axis, contribution on the other. Above-average vs below-average on each. You get four quadrants:
Your hero dishes. People love them, they make real money per plate. The job here is "don't break it." Protect the recipe, ensure consistent execution, feature them prominently on the menu.
The crowd-pleasers that don't quite pay their weight. Burger sells like crazy at $14 with $5.50 in food cost — only $8.50 margin while your signature plate gives $18. The task is to engineer the margin upward without losing the volume.
Big margin per plate but nobody orders them. The whole-fish-for-two, the dry-aged ribeye, the dessert tasting. The work here is positioning: better description, better photography on the menu, server training, placement on the page.
Sells little and barely makes anything when it does. There's almost always one or two. The job is to be honest: drop them, or rework them so dramatically they become something else. Sentimental attachment to a dish you've made for ten years is the most expensive emotion in a kitchen.
What to do with each quadrant
Stars: leave alone, defend hard
The biggest risk to a Star is well-intentioned change. A chef "tweaks" the signature dish, popularity drops 15%, and a Star becomes a Plowhorse overnight. Document the recipe precisely (gram measurements, cooking times, plating diagrams). Make sure every cook on the line can produce it identically. If you're going to change anything, change the price upward — Stars can usually absorb a 5–10% price increase without losing volume.
Position Stars in the visual sweet spots of the menu: top-right of the page on a one-pager, top of the first section on a multi-section menu, with a clear name (no cute branding that obscures what the dish is). Use the "Chef's Pick" or "Signature" badge sparingly here — when everything is a star, nothing is.
Plowhorses: engineer the margin up
This is where the most money hides. Plowhorses have proven demand. The lever is cost or price.
- Reduce a portion by 10–15%. Most plate sizes are over-portioned by historical drift. Customers rarely notice 10g less protein. Save the cost.
- Swap an expensive component for a cheaper-but-equally-good one. Garnish, cheese, garnish oil. Run a side-by-side test for two weeks; if no complaints, lock it in.
- Raise the price by $1–2. Plowhorses are popular by definition — most customers won't react to small increases. Test on a quiet week; revert if you see volume drop.
- Bundle with a high-margin add-on. "Add fries for $4" turns a Plowhorse burger into a Plowhorse-plus-Star combo.
The aim is to move a Plowhorse rightward on the matrix (toward Star territory) without sacrificing the popularity that earned it the Plowhorse label in the first place. One change at a time. Measure.
Puzzles: market harder, not engineer harder
The margin is already great. The problem is awareness. People aren't ordering because they don't notice, don't understand, or don't believe.
- Rewrite the description. "Slow-braised lamb shoulder, eight hours, Aleppo pepper, yoghurt" beats "Lamb shoulder with yoghurt sauce." Sensory language. Cooking method. Origin.
- Move it on the menu. Top of the section, visually offset, sometimes a "Chef recommends" mark. Menu eye-tracking studies are real — top-right corner of a page is most-read.
- Train servers to recommend it. "If you've never had a real shanklish, this is your chance" — server enthusiasm shifts orders dramatically. Brief at pre-service.
- Feature it photographically. If you use menu photos at all, this is where to use them.
If you've done all of that for three months and the Puzzle still doesn't sell, it might actually be a Dog wearing a Puzzle's clothes. Move it.
Dogs: cut them, with care
The honest job. Dogs come in three flavours:
- Sentimental Dogs — "we've always had it." Cut.
- Insurance Dogs — the vegan option, the kids' meal, the safe-bet pasta. These sell little but turn away no one. Keep one of each category, drop duplicates.
- Genuinely Broken Dogs — bad concept, wrong execution, mismatch with the rest of the menu. Cut, and learn from it for the next menu cycle.
Cutting dishes is the single most effective menu engineering move because it has compound benefits: fewer SKUs in inventory, fewer prep tasks, less waste, more focus on the dishes that work. The fear is always "what if a regular asks for it?" — in practice, regulars adapt within a week.
In ProChefDesk
The Menu Builder shows expected revenue and average margin live as you build. Each dish has a special-badge dropdown — Chef's Pick, Signature, New, Spicy — for visually elevating Puzzles and Stars on print. Themes, accent colours, and two-column layouts let you publish the same menu in formats appropriate for dining room, drinks list, or in-room card. And once you've decided to retire a Dog, dropping it cascades: it leaves the menu, the Kitchen Cards, the shopping list, and the prep checklists in one move.
The exercise: 90 minutes, once a quarter
This is a quarterly exercise. Not daily. Not "when the books are bad." Quarterly, on the calendar, so it actually happens.
- Export POS unit sales for the last 90 days. Dish name, units sold. CSV is fine.
- Pull current food cost per dish from your costing system (or recalculate using current invoice prices). This is where a maintained cost report pays off.
- Compute margin per dish: price − cost. And total contribution: margin × units sold.
- Calculate the median for both popularity (units sold) and contribution. Use median, not average — averages get skewed by one runaway hit.
- Classify each dish: above-median popularity / above-median contribution? Star. Below-popularity / above-contribution? Puzzle. Above-popularity / below-contribution? Plowhorse. Both below? Dog.
- Pick three changes for the next quarter. One per actionable category. Don't try to fix everything.
- Measure next quarter. Did the Plowhorse move? Did the Puzzle find an audience? Did dropping the Dog hurt total revenue?
The first time you do it the exercise takes maybe 90 minutes. By the third quarter it's 30 — most of the dishes haven't moved, you're just confirming the placement and choosing the next three changes.
The thing nobody tells you
Menu engineering, done honestly, will tell you uncomfortable things. Your favourite dish might be a Dog. The dish you're embarrassed to put on the menu might be the Star paying your wages. The clever Puzzle you developed in your last service might never find an audience no matter how prettily you describe it.
"The menu doesn't care what you love. It cares what works."
The chefs who get good at this aren't unsentimental. They just learn to separate the dish they want to cook from the dish that should be on the printed menu. The first lives in specials, R&D, the late-night staff meal. The second pays for the business.
Where to start tomorrow
If you've never done this before, do the minimum version. Just five dishes — your three best-known and two you suspect might be Dogs.
- Units sold last month for each (look at the POS).
- Current accurate food cost for each (cost the recipe with today's supplier prices).
- Total dollar contribution: (price − cost) × units. Rank them.
That single ranking — five dishes, three numbers each — will tell you more about your business than the last quarter of P&L summaries combined. Once you've seen the picture for five dishes, you'll want it for all of them. And then you'll have menu engineering as a regular habit, which is the actual win.
The matrix doesn't make decisions for you. It just removes the excuse for not making them.