The 90-day menu-engineering quadrant: stars, dogs, plowhorses, puzzles.

Menu engineering plots every item on two axes — popularity (units sold) and contribution margin (€ profit per unit) — to sort the menu into four quadrants: stars (high margin, high volume), plowhorses (low margin, high volume), puzzles (high margin, low volume), and dogs (low margin, low volume). Each quadrant has a specific play. Run it every 90 days and the menu tunes itself.

Why a quadrant beats a list

Most café and restaurant owners look at their menu through two reports: a sales report (what sold) and a margin report (what made money). The problem is that the two reports rarely overlap in the right way. The best-selling item isn't always the highest-margin item. The highest-margin item often sells in tiny volumes. Reading either report alone leads to wrong calls — promoting a low-margin best-seller, or killing a high-margin slow-mover that just needs better placement.

The quadrant fixes this by forcing both dimensions onto the same picture. Every item lands somewhere on a 2x2 grid. The quadrant it lands in tells you what to do.

The two axes

X-axis: popularity. Units sold over the last 90 days. Use 90 days (not 30) so you smooth out weather, holidays and single-week anomalies.

Y-axis: contribution margin. Selling price minus variable cost (ingredients + portion of labour + payment fees). Not gross margin, not EBIT — contribution. Why? Because the menu decision is "should this item exist?" and contribution is the right cut: it's the money the item generates above its direct cost, before fixed overhead.

Split each axis at the median. The median of popularity = the middle-ranking item by units sold; the median of contribution = the middle-ranking item by € margin. The split lines carve the grid into four quadrants.

The four quadrants

The plays aren't suggestions — they're prescriptions. The only reason to keep a dog is if it's a strategic loss-leader (the gluten-free option that brings in a group, the kids' meal that brings in a family). Even then, you write down the strategic reason and review it next quarter.

Worked example: a Lisbon café

Café Olive in Lisbon ran the quadrant on their food menu (excludes drinks) for Q1 2026. Twelve items, 90 days of data. Median popularity was 320 units sold. Median contribution margin was €4,20.

The plays Café Olive ran:

• Stars (avocado toast, brioche French toast, egg muffin, croque madame): moved to top-left of the printed menu, staff trained to mention them on order. No price change.
• Plowhorses (pastel de nata, granola bowl, cinnamon roll): nata raised from €1,80 to €2,10 (sourcing locked at €0,50/unit). Granola bowl portion reduced 15g, price held. Cinnamon roll reformulated with cheaper butter blend, contribution lifted to €4,40.
• Puzzles (goat-cheese tartine, smoked-salmon plate): added small photos on the menu, server upsell prompt. Volume on tartine doubled in six weeks.
• Dogs (tuna sandwich, fruit salad): killed. Replaced with a chickpea-feta bowl (early data: 340 units in 30 days, €5,10 contribution — likely Star in next review).

Net result over 90 days: food-menu contribution rose 11,4% on flat covers. No price war, no concept change — just the four plays, executed on the right items.

The 90-day cadence

Re-run the quadrant every 90 days. Why 90? Two reasons:

• Shorter than 90 days, you don't have enough data on slow-movers to position them confidently. A puzzle that sold 18 units in 30 days might sell 60 in 90 — the picture stabilises.
• Longer than 90 days, you miss seasonal shifts. The avocado toast that's a star in summer is a plowhorse in February when the avocados arrive bruised.

The 90-day re-run takes about 45 minutes if your product margins are tracked in nouz — you export the per-item report, plot it (median × median), and run the four plays. Schedule it for the first Tuesday of each quarter.

I'd been thinking about the pastel de nata as my hero product because it was always the top seller. The quadrant showed me it was actually my biggest margin drag. I raised it 30 cents, lost zero volume, and added almost €800 to the month.

For the foundational pricing exercise to run on every star and plowhorse, see pricing the croissant. For the help-center walkthrough on the per-item margin report, see the per-product margin view.

FAQ