Glossary
What is menu engineering?
Menu engineering is the practice of analyzing each dish on a restaurant menu by two measures, how much profit it contributes and how often guests order it, then deciding what to promote, reprice, rework or remove. The classic tool is a 2x2 matrix that sorts every dish into Stars, Plowhorses, Puzzles or Dogs.
The classic matrix
Four boxes,
four different moves.
Popular · Low margin
Plowhorses
Guests love them but each plate earns little. Rework the recipe cost, trim an expensive garnish, or nudge the price. Don't cut them, they bring people in.
Popular · High margin
Stars
Guests order them often and each one earns well. Keep them exactly as they are, give them the best spot on the menu, and train servers to point at them.
Unpopular · Low margin
Dogs
Rarely ordered and barely profitable. Unless a dish serves a purpose, a regular's favourite, a dietary anchor, it usually comes off the menu.
Unpopular · High margin
Puzzles
Profitable dishes almost nobody orders. Rename them, move them up the menu, have servers recommend them, or rethink the description. The margin is worth the effort.
Where the classic model stops
Sales data tells you what sold.
It can't tell you what guests wanted.
Classic menu engineering runs on POS numbers alone. That works for ranking what's already on the menu, and it stays silent about everything guests looked for and didn't find. Question data, what customers ask a menu before they order, fills that gap.
The dish nobody could find
Guests ask for dairy-free options that don't exist on your menu. Sales data will never show that demand, there was nothing to sell. Question data counts every ask.
The Puzzle that isn't puzzling
A high-margin dish sells poorly. Sales data calls it a Puzzle. Question data might show guests asked about it and walked away, which means the description or the price is the problem, and you know which fix to try.
The question before the order
When 32% of your customers ask about gluten before choosing a main, that shapes what Stars you should build next. The POS only sees the order that survived the doubt.
This is what tamr's menu analytics adds on top of your POS, powered by an AI waiter on the menu. For the raw material, read what questions diners actually ask waiters →
Common questions
Menu engineering, quick answers.
What are the four categories in menu engineering?
Stars are popular and profitable, keep and feature them. Plowhorses are popular but low margin, rework the cost or nudge the price. Puzzles are profitable but rarely ordered, promote or reposition them. Dogs are neither popular nor profitable, and usually come off the menu.
What data do you need for menu engineering?
Two numbers per dish: how many times it sold over a period, and its contribution margin (menu price minus food cost). Sales counts come from the POS; food cost comes from recipe costing. Question data from a conversational menu adds a third layer, what guests looked for and didn't order.
How often should a restaurant run menu engineering?
Most operators run it quarterly, or with every seasonal menu change. You need enough covers for the numbers to mean something, a few slow weeks can make a solid dish look like a Dog. Pair it with what guests are asking for so you catch demand shifts between cycles.
Engineer with both halves of the data.
Your POS knows what sold. tamr shows you what guests asked for. Put them side by side.