Menu Engineering With Data, Not Guesswork: How NYC Restaurants Use POS Data to Decide What Stays

Every menu has a dish that everyone assumes is a hit because it sells constantly, and a dish everyone assumes is a dud because it barely moves. Both assumptions are frequently wrong, and the only way to actually know is to look at the two numbers that matter for every single item: how often it sells, and how much profit it actually generates per sale.

The framework, explained simply

Menu engineering is a well-established restaurant analysis method that sorts every item on your menu into one of four categories, based on two factors: popularity and profitability. The four categories, commonly called Stars, Plowhorses, Puzzles, and Dogs, give you a clear, repeatable way to decide what to protect, what to fix, and what to remove.

Stars are both popular and profitable, your best-performing items in every sense. Plowhorses are popular but not very profitable, the dishes everyone orders that quietly aren’t making you much money. Puzzles are profitable but not popular, hidden opportunities that most guests simply aren’t ordering. Dogs are neither popular nor profitable, and generally the strongest candidates for removal.

Why this beats gut instinct

A chef or owner’s intuition about which dishes are “working” is usually based on how often something gets ordered, which only tells half the story. A dish that sells constantly but has thin margins, a Plowhorse, can quietly be dragging down your overall profitability even while it looks like a runaway success on paper. Meanwhile, a dish with excellent margins that rarely gets ordered, a Puzzle, might just need better placement or a clearer description rather than removal.

This isn’t a theoretical exercise. Industry analysis on menu engineering has found that restaurants applying this framework and acting on it, repositioning and repricing existing items rather than just adding new ones, have seen profit increases in the range of 10 to 15 percent, without needing to invent a single new dish.

How to actually do this with your own POS data

Pull sales volume and cost data for every item over a meaningful period. Most modern POS systems can generate this directly. You need, at minimum, units sold and food cost per item over the past month or quarter, long enough to smooth out any single unusual week.

Calculate contribution margin for each item. This is simply the price charged minus the food cost, which tells you the actual profit generated per sale, not just the revenue.

Compare each item against your menu’s average popularity and average profitability. An item above both averages is a Star. Above popularity but below profitability is a Plowhorse. Below popularity but above profitability is a Puzzle. Below both is a Dog. This sorts your entire menu into the four categories without any guesswork.

Act differently on each category, not the same way across the board. Stars deserve prime placement and should rarely, if ever, be discounted. Plowhorses often need a quiet cost adjustment, a slightly smaller portion, a cheaper garnish swap, without changing the price or the guest’s perception. Puzzles usually need better menu placement or a more compelling description rather than a price change. Dogs, in most cases, should be removed unless they serve a specific strategic purpose like accommodating a dietary need.

Revisit the matrix regularly, not once and done. Ingredient costs shift, guest preferences change, and a dish that was a Star last quarter can drift into Plowhorse territory as food costs rise. Reviewing this quarterly, or seasonally alongside a menu update, keeps the analysis current rather than stale.

This connects directly to the cost pressures you’re already managing

Menu engineering isn’t a separate project from the margin protection work already covered on the blog. It’s one of the most direct tools for actually acting on that pressure, since it tells you exactly which items are absorbing rising costs quietly versus which ones have room to adjust. It also connects to reducing food waste, since Dogs identified through this process are often also the items generating disproportionate waste relative to how rarely they sell.

Where this fits into your financial habits

If you’re already committed to reviewing your P&L every week, menu engineering is the natural extension of that discipline down to the individual dish level. The same POS data powering your financial review is exactly what this analysis needs, which means this isn’t an additional data-gathering project so much as a different way of looking at information you likely already have.