Most menus are what they are because nobody has looked at the numbers in a year. Items are on the menu because they have always been on the menu. Prices are what they are because that was the price at the last redesign. Meanwhile the food cost has drifted 3 points, the mix has shifted, and the highest-contribution items on the menu are the ones nobody is bothering to sell.
Menu engineering is the quarterly operator process that fixes this. Not menu design, which is the visual work. Menu engineering is the numbers underneath: which items to cut, which to feature, which to reprice. It takes about six hours a quarter and it is one of the highest-return margin moves available to a full-service operator. Across the five-location Michelin-recognized group I ran, this process alone contributed 2 to 4 points of margin recovery per unit per quarter, compounding across four quarters.
What menu engineering actually is
The whole discipline lives on one simple 2x2 matrix. Every item on your menu goes into one of four boxes based on two axes:
- Popularity: units sold in the last 90 days.
- Profitability: dollars of contribution margin per unit sold.
Split each axis at the median for your menu. That produces four quadrants: high-high, high-low, low-high, low-low. The framework has been around since the 1980s (the terms come from Kasavana and Smith), and it is still the cleanest way to see a menu as an operator instead of as a chef.
Fig. 1 · Four boxes, four actions. Simple to plot, hard to execute.
The four boxes and what each one means
Stars: high popularity, high margin
These are your best items. Guests love them and they pay you well. Typically 8 to 15 percent of the menu produces 40 to 55 percent of contribution. Stars need protection, not tweaks. Do not change the recipe. Do not raise the price aggressively. Do give them the best menu real estate, the server call-out, and the photograph if you use one.
Plow horses: high popularity, low margin
These are the workhorses. Guests order them constantly but each unit contributes modest margin. They are your quiet leak. The action here is a price increase, usually 4 to 8 percent, which almost never dents mix and directly recovers contribution. This is where most of the immediate P&L upside sits.
Puzzles: low popularity, high margin
These items would help you if you could sell more of them. They usually do not sell because they are described poorly, priced awkwardly, or hidden in the menu. Puzzles are a marketing problem, not a menu problem. Try repositioning, renaming, or featuring them for two quarters. If they still do not move, they graduate to dogs.
Dogs: low popularity, low margin
These items are earning nothing and taking up shelf space. They also add hidden cost because their ingredients often exist only for them and produce walk-in waste. Cut them, with two exceptions I will get to below.
What to cut
The dogs list is where most operators freeze. Cutting a dish feels like a bigger decision than it is. The item was created for a reason, someone still orders it, and someone on the team advocates for it. Cutting it looks like admitting a mistake.
But the math is unambiguous. A typical menu can lose 15 to 25 percent of its items without a mix impact, and often with a small improvement because the remaining items get more attention from both guests and cooks. Here is the practical cut rule:
- Cut every dog that uses ingredients unique to that item. Those ingredients are hidden waste. The dish is not just failing on the menu, it is also creating walk-in spoilage on your ingredient side.
- Cut every dog that requires a specialized station or a specialized skill. The complexity cost is real. Every dish on the menu that requires the sauté cook to do one specific thing they do for nothing else is a productivity leak.
- Keep a dog only if it anchors a category with no substitute. If it is the only vegetarian entree, or the only gluten-free option, or the only item under $18, it earns its slot by covering a demand you would otherwise lose.
Every item on the menu earns its shelf space, or it takes shelf space from an item that would. There is no neutral position on a menu.
What to feature
Featuring a star is not just menu design. It is server behavior, kitchen priority, and photograph placement all working together.
The mechanics that actually move mix on a star:
- Physical menu placement. Top right of the category, or top of the first page in a single-page menu. Guests scan in a Z pattern and land there first. Move a star into that position and its mix share can lift 8 to 15 percent in a quarter.
- Server call-outs in pre-shift. Every pre-shift, name the two items you want the servers pushing tonight. Servers who know what to lead with will lift the mix on those items by measurable amounts within a week.
- Kitchen protection. Every cook on every station knows how to fire a star, and stars never 86. Nothing kills a star faster than a Friday night 86 in month three of its lifecycle.
- Cross-selling pairings. Servers offer a specific pairing (a specific side, a specific dessert, a specific beverage) with the star. Attachment rate on paired items is often 30 to 45 percent when the pairing is coached.
What to raise
Pricing is the highest-return move on this whole list, and it is also the one operators are most afraid of. The fear is out of proportion to the risk if you do it right.
Raise on plow horses, not on stars
The counterintuitive part: raise prices on high-popularity, low-margin items (plow horses), not on high-popularity, high-margin items (stars). Plow horses are already ordered by inertia. Guests are not calibrated to their price and a modest raise almost never appears in the mix. Stars are ordered by love. Guests remember what they paid and a raise there tests that love unnecessarily.
Size the raise carefully
The sweet spot is 4 to 8 percent per plow horse per pricing cycle. Below 4 percent and the raise is not worth the printing cost. Above 8 percent and you start to test guest tolerance and produce mix shift. In a two-year window, most plow horses can absorb two rounds of a 4 to 6 percent raise without visible mix impact.
Change the price and the description together
A price increase alone reads as a price increase. A price increase paired with a small description update or a slight recipe refresh reads as a menu evolution. This is not deception. It is context. A guest who sees the same dish at a higher price feels different from a guest who sees a subtly refreshed version at a higher price.
Never raise a puzzle price
Puzzles need trial. The whole point is to get more guests to order them. Raising the price on a puzzle makes it a stranger puzzle. Fix the description, fix the placement, fix the pairing, and if it still does not move it becomes a dog, not a more expensive puzzle.
The 90-day process
Once a quarter, block six hours. Pull the last 90 days of item-level POS data (Toast, Square, whatever your system is). Sort by units sold and by contribution margin per unit. Plot the four-box matrix.
Then run the four decisions in order:
- List every dog. Apply the three-part cut rule. Confirm the cuts with the chef.
- List every star. Confirm menu placement, pre-shift call-out, and pairing coaching are in place.
- List every plow horse. Identify the two to four that will get a 4 to 8 percent raise this cycle.
- List every puzzle. Choose one lever per puzzle (renaming, repositioning, featuring) and commit to a two-quarter test.
Publish the changes at the start of the next month. Track mix and contribution weekly for the following 90 days. That data becomes the input for the next quarterly pass.
The mix shift you should expect after each pass
Operators often ask what a successful menu engineering pass looks like on the numbers side. A good rule of thumb, based on the four passes per year I ran across five units at Zareen's:
- Contribution margin lift of 2 to 4 points per quarter in the first two quarters, then 1 to 2 points per quarter after that as the low-hanging items get addressed.
- Menu item count reduction of 15 to 25 percent in the first pass, then 3 to 5 percent per pass after that. The menu gets tighter, not larger.
- Star mix increase of 6 to 12 percent as featuring, placement, and pre-shift call-outs shift guest orders toward the highest-margin items.
- Puzzle graduation rate of 40 to 60 percent in two quarters. Meaning: nearly half of puzzles either move up to stars with a reposition, or get cut. The other half stay in the puzzle box for another cycle.
The compounding matters. A menu that improves 2 to 4 points of contribution margin per quarter for four quarters produces a much larger annual improvement than any single price increase, and it does so without the guest ever noticing a jarring change. The whole strategy is quiet, iterative, and boring. It is also one of the highest-return operator disciplines available.
Menu engineering across a multi-unit group
One additional wrinkle for operators running more than one location: the matrix is per-unit, not per-group. The same item can be a star in one location and a plow horse in another because the guest mix is different. Do not paint the group with one brush. Run four separate matrices for four separate units, and let each unit's data drive the decision for that unit. The exception is a signature item that has to appear on every menu for brand reasons, which becomes a fixed constraint rather than an engineering decision.
What I got wrong the first time
I skipped the chef conversation on my first menu engineering pass at Zareen's. I emailed the chef a list of items I wanted to cut and asked him to sign off. Two of the items on the list were dishes he had built the reputation of the restaurant on. He signed off, but the trust between us took months to fully recover, and I lost time we could have been using on the actual work.
The lesson: menu engineering is a partnership between the operator and the chef, not a spreadsheet delivered as a decision. Every quarterly pass now starts with a two-hour conversation with the chef before I touch the data. The chef sees the numbers, the operator sees the craft, and both agree on the trade-offs. The cuts stick and the relationship holds.
The point
Menu engineering is the quiet quarterly discipline that produces some of the largest margin gains available to a full-service operator. Six hours a quarter, one matrix, four actions. Cut what is not earning its shelf space, feature what already sells, raise prices where the guest will not notice, and test the items that could be selling. Skip the process and your menu drifts into a museum of decisions nobody remembers making. Run it consistently and your contribution margin climbs 2 to 4 points a year without a single change to the kitchen equipment or the vendor list.