The menu is the P&L. It is not a design document, not a marketing artifact, not a chef's love letter. It is a set of price and margin choices, and every guest interaction with it is a mix decision that shows up in the following week's contribution number. Operators who understand this run their menu quarterly and never spend a single quarter without margin lift. Operators who do not run it at all leave 1 to 3 points of contribution on the table every year, and they do it while congratulating themselves on holding food cost.
This is the mechanics of menu engineering, done the way I run it in field operations, with the numbers that actually move.
The matrix in plain English
Score every item in a category (say, all entrees) on two dimensions:
- Popularity: Percent of items sold within the category. If your category has 12 entrees and one item sells 15 percent of category volume, that item is above the category average of 8.3 percent.
- Profitability: Contribution margin dollars per item sold. Not percent. Dollars.
Then plot each item on a 2x2 grid. Above-average popularity and above-average margin dollars is the top-right quadrant. Below-average on both is the bottom-left. The four quadrants each carry a name and an action.
Fig. 1 · Two axes. Four quadrants. Four different actions. Run it every 90 days.
What to do in each quadrant
Stars: feature and protect
These are the items making you money. Feature them on the printed menu, on the daily specials board, in server callouts. Protect the recipe. Do not let the kitchen substitute a cheaper ingredient to shave 40 cents. That shaves the reason it is a star. Do not raise price aggressively. Guests know the item and will notice.
Stars need one more thing: succession. Every star has a shelf life. Something will eventually replace it. The star quadrant should always have a "next star" candidate being moved up from the puzzles quadrant so you are not caught when the current one fades.
Plow horses: raise price or reengineer
Popular items with weak margin. The guest loves them. The P&L does not. Two moves:
- Small price increase. 3 to 5 percent, timed with a menu refresh. On a $22 item that is 66 cents to $1.10. Rarely visible to the guest. Adds 3 to 5 points of margin on a popular item.
- Recipe reengineering. Same guest experience, lower cost of goods. Portion audit. Ingredient swap on a garnish that no one photographs. Plate optimization.
The math example that always lands: a plow horse burger selling 800 units a month at $14 with 42 percent food cost. Contribution per unit is $8.12. Bring food cost down 2 points via portion discipline and small changes to the bun spec. New contribution per unit is $8.40. Times 800 units, times 12 months, that is $2,688 in annual contribution recovered on one item. Multiply across five plow horses and you have $13k of margin per unit, per year, from one afternoon of work.
Puzzles: reposition
These are the frustrating quadrant and the highest-leverage one. High margin, low sales. Something is stopping the guest from ordering them. Common reasons and fixes:
- Menu position: Move it to the top-right of the printed menu, where the eye tracks first. This alone can lift mix by 15 to 25 percent on an item.
- Description weakness: Rewrite it. "Roast chicken" versus "Herb-brined half chicken with pan jus and Yukon gold potatoes" is a different sales pitch.
- Server unfamiliarity: If servers do not know how the item eats, they will not recommend it. Sample it at pre-shift.
- Bad pricing signal: Sometimes the puzzle is priced too low, and guests assume it must be a lesser option. Counterintuitive but I have seen this move mix.
The goal is to move puzzles into the star quadrant. A puzzle promoted successfully turns into your next star. This is how a menu compounds margin without any menu redesign.
Dogs: kill unless strategic
Low sales, low margin. Default is delete. Every menu item has an operational cost even if it does not sell: it takes up prep space, it requires a spec, it complicates receiving, it distracts the server describing it. The exceptions are strategic:
- A vegan or gluten-free option that guests do not order much but that keeps the group table from walking.
- A kids item that closes a family visit.
- A high-price signaling item that anchors the perceived value of the rest of the menu (a $54 steak that never sells makes the $32 pasta feel reasonable).
Every dog stays only if someone can name the strategic reason. Default: kill.
The single most important scoring rule
Score profitability on margin dollars per item, not margin percent. This is the one thing I see operators get wrong every single time.
Item A: 78% margin, sells 40/month Margin dollars: $6.00 × 40 = $240 Item B: 62% margin, sells 400/month Margin dollars: $5.20 × 400 = $2,080 Item B is 8.6x more valuable to the P&L, despite having 16 points lower margin percent.
If you rank on percent, you will feature Item A. If you rank on dollars, you will feature Item B. The P&L cares about dollars. Rank on dollars.
Rank on contribution dollars, not gross margin percent. This one rule is worth more than most menu redesigns.
The server callout is 60 percent of the work
The single biggest lever for shifting mix is not the printed menu. It is the server callout. What the server names when the guest asks "what do you recommend" moves mix more than any layout change.
Concrete example: on one Bay Area unit I ran we changed the server callout from "the salmon is popular" to "the mushroom bowl, and the branzino for the table." The mushroom bowl was a $16 star, the branzino a $32 puzzle. Over 60 days branzino mix went from 4 percent of entree covers to 11 percent. Mushroom bowl held. Contribution per cover on entrees moved up $1.85. Annualized that is roughly $47k on a single-unit basis. Zero menu redesign. Zero marketing spend. One line change in pre-shift talk track.
Update the callout every quarter to align with the current matrix. Servers will pick it up if you make it specific.
How I run the actual quarterly cycle
- Week 1 of the quarter: Pull 90 days of item-level sales from POS. One row per item: name, category, quantity, price, food cost, contribution per unit.
- Week 1: Score into the matrix. Publish a one-page grid per category (entrees, apps, desserts, beverages).
- Week 2: Meet with the chef. Reengineering targets on plow horses. Recipe protection on stars. Kill list on dogs. Reposition strategy on puzzles.
- Week 3: Menu updates, print run, POS updates. Server callout update.
- Weeks 4 to 12: Track mix weekly against the plan. If a puzzle is not moving, revisit the callout or the description.
- End of quarter: Re-run the matrix. Publish before-and-after contribution per cover.
The whole cycle is roughly 12 hours of operator work per quarter. On a $2M unit it moves contribution by 1 to 2 points, which is $20k to $40k. That is a $2k to $3k return per operator hour. There is almost no other exercise in restaurant operations with that math.
The interaction between price, mix, and cost
Where the matrix gets interesting is the interaction between the three variables. A move in one changes the others in ways that surprise operators who only look at one at a time.
Concrete example. You raise a plow horse from $14 to $15, a 7 percent price increase. Volume drops 4 percent because some guests trade down. On the surface, revenue on the item is up 2.7 percent. But the trade-down guests moved to a lower-margin item, which pulled category mix toward a lower-contribution position. Category contribution per cover ended up down $0.30, not up.
The lesson: any price change gets scored on category contribution per cover, not item revenue. If you cannot measure the mix shift, the price move is a guess.
Same with ingredient cost. Beef price goes up 12 percent. Your burger food cost moves from 32 to 36 percent. If you leave the price unchanged, you have quietly converted a plow horse into a bigger plow horse. If you raise the price 5 percent, you might convert it into a puzzle (guests trade away). The right move depends on how price-sensitive the burger buyer is. Test with a 30-day price change on one unit before rolling group-wide.
Beverage mix is the hidden lever
Most menu engineering conversations focus on entrees. The margin math actually lives in beverage. A beer at 78 percent margin, wine by the glass at 72 to 82, cocktails at 75 to 85. If beverage attach rate is 55 percent versus 72 percent, that is a full point of contribution margin at the P&L level, from one variable, and it moves through server behavior more than any other lever.
Track beverage attach rate weekly, by server, and coach the bottom third every month. This one exercise, run consistently, has moved beverage attach rate by 10 to 15 points in every operation I have coached. On a $2M unit that is $25k to $40k in annual contribution recovered. It does not show up on the menu engineering matrix if you only score the food side. Do a beverage-only matrix quarterly and run it alongside.
What I got wrong the first time
Two mistakes I made in the first year I ran matrix cycles.
First, I ranked on percent margin. I featured items with beautiful percentage margins that sold in tiny volume, and my "top star" was a $19 salad selling 22 times a month. I moved menu real estate to it. Nothing happened at the P&L level because $19 × 25 percent volume lift × 78 percent margin was still just a few hundred dollars a month. I could have featured a plow horse burger and moved five times the margin dollars.
Second, I redesigned the menu once a year instead of running the matrix quarterly. Mix drifted every quarter. I was catching it 9 months late. The delta between quarterly cadence and annual cadence, on the group I ran, was roughly 0.8 points of contribution margin per year. On a $30M revenue base that is $240k. From meeting cadence alone.
The point
The menu is the P&L. Every guest interaction with it is a mix vote. Mix drifts constantly and margin drifts with it, quietly, in a direction operators cannot see until quarter close. The matrix is a 12-hour quarterly exercise that catches the drift and pushes it back the other way.
Rank on dollars. Coach the callout. Kill the dogs. Reposition the puzzles. Raise the plow horses. Protect the stars. Do it every 90 days. If you do only one thing on this list, it should be re-run the matrix on the same Monday every quarter. Cadence beats intensity. The menu compounds when you look at it on a schedule.