Menu Engineering With Real Data: From Stars and Dogs to Profit

A practical four-quadrant menu analysis that uses POS data, not gut feel, to decide what to push, what to fix, what to demote and what to cut.

By NASS Editorial · · 9 min read · Menu, Analytics

Menu engineering is one of the oldest disciplines in restaurant management, and one of the most consistently neglected. Most operators run the same menu for years, occasionally swapping seasonal items, and never systematically ask which dishes are quietly destroying margin and which are quietly carrying the business.

This piece walks through a four-quadrant menu analysis you can run in 90 minutes with two pieces of data: 90 days of POS item-level sales, and the recipe cost of each menu item.

If you cannot easily get recipe costs, that is itself a useful finding and the place to start.

The four quadrants

Plot every menu item on two axes: - Horizontal: number of times sold in the period (popularity). - Vertical: contribution margin per sale (selling price minus food cost).

Draw a vertical line at the average popularity and a horizontal line at the average margin. You now have four quadrants:

  • High popularity, high margin: stars. Protect them.
  • High popularity, low margin: workhorses (sometimes "plowhorses"). Improve their margin or quietly reposition.
  • Low popularity, high margin: puzzles. Promote them.
  • Low popularity, low margin: dogs. Cut them.

This is not a new framework. The discipline of actually running it on real data is where the value is.

Building the data set

Export the following from your POS for the last 90 days:

  • Item ID, item name, category.
  • Number sold.
  • Total revenue.
  • Optionally, average price (in case you ran discounts).

Cross-reference with your recipe book to get the food cost per item. If your recipes are out of date or non-existent, build a simple cost-per-recipe spreadsheet for the top 40 items. Anything below the top 40 is statistical noise for a typical independent restaurant.

Calculate per-item: - Contribution margin = selling price − food cost. - Margin % = contribution margin / selling price.

What to do with each quadrant

Stars: these are the items that fund the restaurant. Make them easy to find on the menu, photograph them well, and protect them from cost increases by maintaining tight portion control. Do not over-engineer them. Do not raise their price unless your competitors do first.

Workhorses: popular but low-margin. They are usually high-quality value items that customers come for, but the margin is weak. Options, in order of preference: - Engineer the recipe to reduce food cost (smaller portion of premium garnish, lower-cost protein cut). Test in a soft launch. - Increase the price by 5-8%. Workhorses are price-elastic but rarely as sensitive as operators fear. - Pair the workhorse with a high-margin add-on that customers reliably take.

Puzzles: high-margin but unpopular. Usually a marketing problem, not a product problem. Move them up the menu, add a photo, train servers to recommend them. A 25% lift in volume on a high-margin item is a much bigger profit gain than the same lift on a workhorse.

Dogs: low-margin and unpopular. Cut them, unless they serve a strategic purpose (a vegan option, a kids' menu, a top-selling item's "ingredient cousin"). Each item cut reduces inventory complexity, prep time and waste.

The price-test discipline

When you change a price on a star or workhorse, watch the metrics that matter for the next 4 weeks:

  • Volume of that item.
  • Volume of substitutes in the same category.
  • Average ticket size.
  • Customer comments / reviews.

If volume drops more than 8% and substitutes do not pick up the slack, roll back the price. If volume drops less than 5% and revenue per cover rises, the price change worked.

Do not run multiple price changes simultaneously. You will not be able to tell which one caused which effect.

Beyond the four quadrants

The classic four-quadrant analysis is a great starting point but misses a few important dynamics:

  • Menu psychology. Items at the top-right of a printed menu, or in a "specials" box, sell more regardless of merit. The placement, not just the dish, drives volume.
  • Time-of-day effects. A dish that is a dog at dinner may be a star at lunch. Run the analysis separately by daypart if you have the data.
  • Ingredient leverage. Sometimes a low-volume item carries a high-volume one. Cutting "miso soup" might quietly tank ramen sales.
  • Customer mix. The same item performs differently for tourists vs. regulars, dine-in vs. takeaway. A dish that is a dog overall might be a star in your takeaway channel.

A more sophisticated analysis breaks these out separately. A first pass with the four-quadrant model is still 80% of the value.

How often to run this

Quarterly is the right cadence for most operations. Monthly is overkill — menu items take time to settle into a sales pattern. Annual is too infrequent — by the time you notice a dog, you've already lost a year of margin.

Block 90 minutes in the calendar on the first Monday of each quarter, run the analysis, decide on 2-4 changes, and implement them within the following two weeks.

What this analysis will reveal

Almost every restaurant we have done this with for the first time discovers some version of the same three findings:

  • A dish nobody thought about is the highest-margin item on the menu, selling well below its potential.
  • A "signature" dish that staff are proud of has terrible margin and is rarely re-ordered. Customers try it once.
  • 4-6 items can be cut entirely with no measurable customer impact and meaningful operational simplification.

Each of those findings is a 2-5% boost to profit. Together, they are usually enough to fund a real wage rise for the staff who held the menu together while it was working against them.

The bottom line

Menu engineering is the highest-leverage 90 minutes a restaurant operator can spend per quarter. The data is already in your POS. The recipe costs are knowable. The four-quadrant model is simple. The reason most restaurants do not do it is not technical — it is the willingness to act on what the data says, including cutting items the chef loves and raising prices on items customers expect to stay cheap.

The restaurants that do it consistently quietly outperform their peers, year after year, on the only metric that matters: profit per cover.