Your best-reviewed dine-in dish can be your worst delivery item. The crispy chicken sandwich that gets raves at the table shows up as a steamed sponge, and the customer blames you, not physics; the one-star review mentions neither the traffic nor the box. Classic menu engineering sorts items on two axes, popularity and margin. Delivery adds a third one nobody teaches you: travel-worthiness. Ignore it and the old matrix will point you, confidently, at the wrong menu.
Travel-worthiness: the 25-minute box test
Before you touch a spreadsheet, run the physical test. Cook your ten best sellers, pack them exactly the way they'd go out for delivery, and leave them sealed for 25 minutes, a realistic door-to-door time for most city orders. Then open the boxes, look, and eat. Braises, curries, rice bowls, stews, and pizza mostly survive; centuries of leftovers already engineered them for it. Fried textures die first: fries, tempura, anything whose whole appeal is crunch. Composed dishes fail in a way you won't notice until you look. The sauce migrates, the greens wilt, and the plating that justified $24 is now a pile.
A failed dish has three exits. You can re-engineer it (sauce on the side, vented packaging, a sturdier cut of fry), re-price it into honesty (call it "market fries," charge less, promise less), or drop it from the delivery menu altogether. Treat the delivery menu as its own list, not a photocopy of the dining room. The strongest delivery brands run something shorter, where every item has passed the box test.
The four quadrants, redrawn for delivery
Stars are the easy call: popular, high-margin, and they survive the trip. Open the menu with them, on the first screen, in the biggest photos, in the first bundle slot. A plowhorse sells well but earns thin, and delivery makes it riskier than the dining room does, because packaging and courier costs eat whatever margin was left. Fix one with a delivery-specific portion, a $1–2 reprice, or a bundle that pairs it with a high-margin side. With puzzles the trouble is visibility, not the food: they're profitable and nobody orders them, which online almost always means a merchandising miss. This is where a digital menu beats paper, through a better photo, a higher slot, an "add for $4" prompt at checkout. Dogs are unpopular, unprofitable, and usually bad travelers on top of it. Cut them from delivery even when the chef defends them at dinner, because every screen slot a dog holds is a slot a star could be using.
Put real numbers on it. This is contribution per item sold on a worked example, after food cost, packaging, and an allocated slice of delivery cost:
Bring your menu to a 30-minute call — we'll build the matrix from your real order data.
Photos move items between quadrants
At a table, the server sells the puzzle dish. Online, the photo does that job. Operators consistently report that a real, well-lit photo lifts an item's order rate by 20–30%, and more than that for dishes people don't recognize, which makes photography the cheapest way you have to move an item between quadrants. Three rules keep it honest. Shoot the actual dish, because customers photograph what arrives and any gap becomes a refund. Photograph at least some items in their delivery packaging, so expectations stay ones you can meet. And don't leave a puzzle without a photo while a dog has one. On your ordering website and branded app you control every pixel of this, which is one reason the same menu converts better on channels you own than inside a marketplace template.
Names and descriptions carry more weight than they look. At home nobody can ask the server what "Nduja Rigatoni" is, so an unfamiliar name needs one plain line: what it is, what comes with it, how spicy. Operators say that adding a concrete description to a high-margin item that's underselling often does half of what a photo does, and it costs nothing. Write for a hungry person reading on a phone, not a food critic.
Placement is the second server
Digital menus have prime real estate the same way paper ones do: the first screen, the top of each category, the checkout upsell slot. That's where stars and puzzles belong. Bundles earn special attention in delivery because they hit two numbers at once, raising the ticket and giving a plowhorse a high-margin escort. A "burger + wings + drink" set at $21.90 can carry a $3.10 plowhorse to an $8 combined contribution. We ranked the ticket-size levers in our AOV breakdown.
Re-run the matrix quarterly, with delivery-only data
The matrix is a snapshot, and delivery data moves faster than dine-in data does: weather, seasons, and a competitor opening down the street all redraw it. Rebuild it every quarter from item-level sales and margins in reports and analytics, and use delivery orders only, because dine-in popularity will mislead you about what people order at home. Two signals the classic framework skips are worth watching: items that show up together in refunds, which is a travel-worthiness alarm, and items customers keep adding at checkout, which are your natural bundle partners.
Do this next
Tonight, run the 25-minute box test on your top ten sellers. This weekend, pull one quarter of delivery-only item sales and sort them by contribution per item. If a dish fails the test while sitting on your first screen, that's the most expensive pixel on your menu, and you can fix it by Monday.
Item-level analytics, bundles, and a menu you fully control — app and site live in about two weeks.