How much of a bundle's sales are actually new?

The question that decides every bundle

A bundle only adds money when it captures spend that wouldn't otherwise have happened. That single sentence is the whole game, and it is easy to lose sight of behind a rising cover count. The trap is comparing the bundle to nothing — counting every deal sold as a win — when the only fair comparison is to the world in which you never ran it. Compared to what is everything.

Picture a customer who walks in for a Coke. At the till they're offered fries for an extra £1 as part of a combo, and they take it. Did the bundle work? It depends entirely on a question the receipt can't answer: would they have bought those fries anyway? If they were always going to add fries, the combo just handed them a discount on a sale you already had — pure give-back. If the fries only made it onto the tray because the deal made them feel like good value, that's genuinely new spend, and the bundle earned its keep. Same transaction, opposite verdicts. Multiply that across thousands of covers and the share that falls into each camp decides whether the whole programme makes or loses money.

That share is incrementality: the proportion of bundle sales that are genuinely additional rather than sales you'd have made regardless. Its mirror image is cannibalisation — the deals taken by people who would have bought the items at full price, where the discount comes straight out of contribution. Every bundle is some mix of the two, and the mix is rarely flattering. The job is to estimate it honestly, because nearly every other number — redemptions, revenue, even a headline volume lift — flatters the deal by treating cannibalised sales as if they were new.

What the research says (and the honest number)

So what is a realistic share? The most credible answer, drawn from the studies that actually measure it, is roughly half — and that is the figure to plan around until your own data says otherwise.

The foundational work comes from grocery and consumer-packaged-goods promotions, where scanner data lets researchers take a promotional sales bump apart. Van Heerde, Gupta and Wittink (2003), in the Journal of Marketing Research, decomposed exactly that bump and found that only about 33% came from brand-switching — customers defecting from a rival — which means roughly 67% was incremental to the brand rather than stolen share. A companion finding in the same research stream reinforces it: of the brand's own incremental sales, around two-thirds reflect genuine extra (primary) demand rather than borrowing from the future or other products. Encouraging — but that two-thirds is the optimistic end of the picture, measured for a single brand on a shelf rather than a self-funded basket.

Pull back to the category and the gloss comes off. Nijs and colleagues (2001), studying 560 consumer-goods categories in Marketing Science, found that price promotions overwhelmingly reshuffle sales within a category and only rarely expand it for any length of time. A bundle of your own products is much closer to that category view than to the brand-switching one — there is no competitor's product on the next peg to capture, so a chunk of every bundle is your own demand rearranging itself. That is precisely the dynamic that pulls the realistic number down from two-thirds toward half.

Practitioners who do this for a living have converged on the same place. The revenue-growth-management benchmark is blunt: aim for more than 50% of promoted volume to be incremental — and the framing matters, because it is set as a target to clear, an implicit admission that plenty of real promotions fall short of it. From the food world specifically, the CPG analysts at Circana work through a worked example in which around 40% of promoted sales were "subsidised" — sold at a discount to buyers who'd have purchased anyway — leaving about 60% incremental. Stack the brand-switching decomposition (~67% incremental), the category reality (often far less), the >50% target and the ~60% food example side by side and they bracket the same conclusion: plan around half incremental, with most credible estimates sitting in a roughly 40–65% band. Anything markedly above that should be treated as a pleasant surprise to be proven, not assumed.

It isn't about the size of the discount

A natural instinct is to read incrementality off the discount: surely a bigger price cut pulls in more genuinely new customers? The evidence says no — and the reasoning, once you see it, is uncomfortable. Deeper discounts don't make sales more incremental; if anything they make them less so, because the loudest responders to a deep cut are the people who were already going to buy and now simply buy more of it.

The cleanest controlled test comes from McColl, Macgilchrist and Rafiq (2020), who examined what actually drove cannibalisation in a retail promotional setting. The culprit wasn't the depth of the price cut — it was store format and context. Where promotions cannibalised, they did so regardless of how generous the discount was; deeper cuts mostly skewed the response toward existing buyers stockpiling — loading up at the lower price — rather than new customers appearing. For an operator that translates cleanly: a deeper meal deal is not a safer bet on incrementality. It mainly transfers more margin to the regulars who'd have come anyway, while the share of genuinely new spend barely moves. So don't infer incrementality from the discount — the headline cut tells you what the deal costs you, not how much new business it brings.

What actually drives it

If depth doesn't move incrementality, what does? Two forces, pulling in opposite directions.

The first pushes it up: complementarity and the ability to reach a genuinely new occasion. When a bundle pairs things that naturally go together — and gives someone a reason to buy on a visit they'd otherwise have skipped or spent less on — the extra sales are far more likely to be real. Derdenger and Kumar (2013), in a structural study in Marketing Science, found that mixed bundling of complementary products drove roughly one million extra software units and around a hundred thousand extra hardware units — large, genuinely incremental volume, not reshuffled demand. The crucial detail in their result is the word mixed: the items stayed available to buy separately as well as together. Keeping standalone options open is what lets the bundle add new buyers without simply herding existing ones into a cheaper basket. The hospitality read-across is direct — a side and a drink that genuinely complete a main, offered alongside the à la carte menu rather than replacing it, is the kind of deal most likely to bring new spend.

The second force pulls it down: deal-seekers and trade-down. Some customers don't respond to a bundle by spending more — they respond by spending less than they otherwise would, swapping a higher full-price order for the cheaper set. Restaurant data makes the tension vivid. Quick-service combos reliably lift average order value, and value menus have become a central battleground for the chains; Datassential's 2025 work on value-meal pricing documents how heavily operators now lean on combos and value bundles to drive spend per visit. But the very customers those deals attract are the most price-driven: value buyers are markedly more likely to shop deals across chains, chasing whoever has the keenest offer this week. That is the live signature of trading down — covers that look like wins on the count but represent existing demand rotating to the cheapest available deal rather than net-new spend. A bundle's true incrementality is the net of these two pulls, which is exactly why the honest number lands near half rather than at the optimistic two-thirds.

Break-even incrementality — the number to watch

Here is where the concept turns into a decision rule. You don't actually need to know a bundle's exact incrementality to judge it — you need to know how high it has to be. That threshold is the break-even incrementality: the minimum share of bundle buyers that must be genuinely new for the deal to pay for itself.

It falls straight out of the contribution maths. Define it plainly:

An example makes it concrete. If a bundle leaves £8 of contribution where the items sold apart leave £10.50, then break-even incrementality is 1 − (8 ÷ 10.50) ≈ 24%. Roughly a quarter of the buyers must be people who genuinely wouldn't have bought otherwise; if more than that are incremental, the deal adds money, and if fewer are, it erodes it. Now hold that 24% against the ~50% planning benchmark and the deal looks comfortably achievable. Set the discount deeper so the bundle leaves only £5 of contribution and break-even climbs to over 50% — right at the edge of what's realistic, and a deal you should think hard about. The bundle analyser does this contribution maths for you — the £8, the £10.50 and the break-even that falls out of them — for your own items, prices and bundle price. What it deliberately won't do is guess your incrementality: that figure is a judgement, not a measurement, so the tool hands you the hard numbers to weigh against rather than a false-precise estimate. You bring the ~50% benchmark, and you confirm it with a test.

Most promotions lose money — so test it

A reasonable response at this point is discomfort, and that discomfort is well founded. Across the wider promotional world, the base rate is bleak, not gain. McKinsey, drawing on Nielsen data, reports that roughly 59% of trade promotions fail to break even — and in the US the figure is closer to 72%. The academic record agrees: a well-cited analysis by Ailawadi and colleagues found that more than half of retail promotions are unprofitable once all the costs and the give-back are counted. The majority of deals, run on instinct, lose money. A bundle is not exempt from that gravity.

Which leads to the one piece of advice that every credible source in hospitality and retail converges on: test it. You cannot reliably reason your way to a bundle's true incrementality from the armchair — the only honest read comes from running the deal in a subset of sites or markets and comparing them against comparable sites that didn't run it. Restaurant Dive, reporting on operators' value-promotion strategy, frames it exactly this way: the question to answer with a controlled test is whether a deal actually drives new visits and new spend, or merely trades existing spend down to a lower price point. A small, clean test — a handful of sites, a defined window, a like-for-like control — answers in weeks what no spreadsheet can settle on its own, and it converts incrementality from a guess into a measured number you can roll out on with confidence.

The bottom line

Incrementality is the hinge every bundle turns on, and the honest place to start is humble. Begin from the assumption that about half of a discounted bundle's sales are genuinely new. Adjust that up for deals that pair real complements and open a new occasion — a side and a drink that complete a main, offered alongside the full menu rather than instead of it. Adjust it down for discounts on the staples your regulars already buy week in, week out, where most of the take-up is simply your own loyal demand handed a cheaper price. Then stop adjusting and confirm by testing: run it in some sites against a fair control and read the new visits and new spend, not the redemptions. Set that measured incrementality against the deal's break-even — the contribution figures the analyser works out for you — and you have the one comparison that tells you whether a bundle earns its place, or quietly gives your margin away.