Stars, plowhorses, puzzles and dogs
The menu matrix sorts every dish by how well it sells and how much it leaves you. What each group means, and what to do with it.
Menu engineering is a simple idea from the 1980s that still works. You compare every dish on two things at once: how popular it is, and how much contribution it leaves per portion. That gives four groups, and each one calls for a different action.
The two lines
Popularity. A dish counts as popular if its share of sales is at least 70% of an equal share. With 20 dishes, an equal share is 5%, so anything above 3.5% of portions sold is popular. The 70% rule keeps a few best sellers from making everything else look weak.
Contribution. A dish counts as high-contribution if it leaves more per portion than the menu's weighted average. That is total contribution divided by total portions sold.
The four groups
| Group | Popularity | Contribution | What to do |
|---|---|---|---|
| Stars | High | High | Protect the recipe and the portion |
| Plowhorses | High | Low | Re-engineer the cost, or raise the price a little |
| Puzzles | Low | High | Promote them: placement, name, staff recommending |
| Dogs | Low | Low | Replace or drop |
Stars
These are the dishes that pay the bills. Keep the spec tight. Watch their ingredient prices closely, because a supplier increase on a star costs more than on anything else. Don't discount them.
Plowhorses
Guests love them, but they leave less than average. Usually the fix is in the cost: a slightly smaller portion of the expensive part, a cheaper side, a better supplier price. Small price increases also tend to work, because demand is strong.
Puzzles
Good money per plate, but few people order them. Move them to a better spot on the menu, rename them, describe them better, or have the team suggest them. If nothing works after a month or two, treat them as dogs.
Dogs
Low sales and low contribution. Unless a dish has a real reason to stay, like a vegetarian option the menu needs, replace it with something new and test it.
Run the matrix per channel. A star in the dining room can be a plowhorse on Wolt once commission is paid.
Two common mistakes
- Using food cost percentage instead of contribution. A steak at 38% food cost can leave far more money than a pasta at 22%. The matrix should rank by euros left, not by percentage.
- Using old recipe costs. If plate costs are from last year, the groups are wrong. The matrix is only as good as the costing under it.
FoodNormIQ builds the matrix from current plate costs and real sales every month, for the whole menu and for each channel, so the groups move when supplier prices do.