Menu cannibalisation: the real cost when a new dish eats an old one

The till lies by omission

Every launch review starts with the same chart: the new item’s weekly sales. It is the one chart that cannot answer the question being asked of it. Marketing scientists who decompose new-product demand find it splits three ways — sales pulled from your own products, sales pulled from competitors, and genuinely new consumption — and that a new product’s net demand “may be considerably less than its total demand” once the first slice is stripped out.[1] Their flagship case was a car, not a curry: when Lexus launched the RX300, its demand came partly out of Lexus’s own SUVs and its own saloons — a double bite the headline sales figure never showed.[1]

A menu behaves the same way, only more so, because the room is smaller. The guest choosing your new smash burger was, more often than not, going to buy something from you tonight. Even the giants file this as a formal risk: at the height of its expansion, Starbucks listed “opening less productive stores and cannibalizing existing stores” among the named threats to its growth in its annual SEC filings.[3] If a company with that much data treats self-competition as a board-level risk, a single site launching a fourth burger shouldn’t assume its new sales are new money.

The whole question fits in one subtraction

Here is the useful surprise: you don’t need to predict the cannibalisation rate to know whether cannibalisation will hurt you. You need one subtraction. Every guest who switches swaps the contribution of the old dish for the contribution of the new one, so:

gain or loss per switched sale = new contribution − old contribution

If that gap is positive, every switch helps and the switching rate only sets the size of the win. If it’s negative, every switch costs you and no rate can rescue it — more take-up just digs faster. The rate scales the answer; the gap decides its sign. That is the whole discipline, and it’s why the modelling has to happen before launch, when the gap is still a choice you control through price and spec.

Two launches, one menu

Put real numbers on it — VAT at 20%, one burger doing 520 sales a period. It sells at £14.00 and costs £4.10 to plate, so each sale nets £11.67 and leaves £7.57 of contribution. Now launch a £16.50 premium version and suppose three in ten buyers trade across (156 sales). Watch how the same launch goes two completely different ways depending on one line of the spec sheet:

Same selling price, same take-up, same fanfare. A £1.90 difference in plate cost is the entire difference between a launch that adds £91 a period and one that quietly burns £205.

Notice what the price told you: nothing. Both versions are “premiumisation” at £2.50 more on the menu. The dearer dish with the generous spec is a trade-down in contribution wearing a trade-up’s price tag. This is the single most common way cannibalisation hurts operators — not a cheap item stealing from a dear one, but a new build whose extra cost outruns its extra price.

One more trap hides in the good scenario. At £5.60 cost the new burger carries a lower margin percentage than the old one (59% of net, against 65%) — yet it makes you £91 more cash. Judge switches in pounds of contribution, not points of margin: percentages pay no bills, and chasing them would have talked you out of a profitable launch.

Sometimes eating your own lunch is the strategy

None of this makes cannibalisation a dirty word. Deliberately taking sales off your own menu is sound strategy in at least three situations, and the same subtraction governs all of them:

• Trading up. If the new item carries more contribution per sale, you want the switch — the faster the better. The model simply tells you what each point of switching is worth, so you can decide what it’s worth spending to encourage it.
• Defensive launches. If guests are going to defect to a competitor’s new category anyway, sales you cannibalise from yourself are sales you didn’t lose to the rival — the relevant comparison isn’t “before”, it’s “the before that was about to disappear”.[1]
• Planned replacement. If the old line is due for deletion, high cannibalisation is the point: you’re migrating its buyers before it goes. Here the switch rate matters more than the gap — you need the buyers to actually move.

What’s never sound strategy is the accidental version: a lookalike item added because the category “needed refreshing”, at a spec nobody costed against the line it shadows. Thirty years ago Quelch and Kenny watched packaged-goods companies do exactly this at scale and concluded that unchecked line extensions rarely expand demand, do disguise rising costs, and muddle the strategic role of every product in the range — which is why their first fix was better cost accounting, not better marketing.[2] Swap “line extension” for “fourth chicken dish” and the finding maps onto a menu without editing.

What to do with this before your next launch

Three habits turn the theory into a routine that takes about twenty minutes per launch:

• Cost the gap first. Before anything else, compute new-minus-old contribution against every item the newcomer plausibly shadows. A negative gap is a design problem — fix the price or the spec before the menu goes to print, because after launch your options shrink to delisting.
• Bracket the switch rate, don’t guess it. You can’t know whether 20% or 50% of buyers will move. You don’t need to: run both. Because the effect is linear in the rate, the two runs bound every outcome in between — if both ends look acceptable, launch; if one end is ugly, you know exactly what to watch.
• Read the mix after launch, not the item. The launch review chart should be total category contribution — old items plus new — against the pre-launch baseline. That’s the only chart on which cannibalisation is visible at all, and it’s the one the tool below prints for you, before and after.

And keep one caveat in view: the pure-switch arithmetic here is deliberately conservative. A genuinely distinctive new item also brings incremental demand — guests and occasions you weren’t getting — and none of that is counted above. How much of a new line’s take-up is truly incremental is its own measurement problem, with its own tricks and traps, and it has its own article. But incrementality is the upside case. The subtraction in this one is the downside case — and the downside is the part you can compute, in advance, with the figures already in your costing sheets.