Promo Uplift Decomposition: Where a Promotional Bump Actually Comes From

Two published studies decompose the same uplift and come back with very different answers. One says about three quarters of it comes out of rival brands. The other says about a third. Both are right, and knowing why is what lets you hold your ground when somebody quotes the other number at you.

The first counts the elasticity. Working from household panel data, which follows what individual homes buy week after week, across 173 brands in 13 categories, it breaks a shopper's reaction to a promoted price into the three decisions they actually make: which brand to pick up, whether to buy the category at all on that trip, and how much to carry home. On average 75% of that reaction sits in the choice of brand, 11% in the decision to buy the category at all, and 14% in how much they take [Bell and colleagues]. Read as units, your promoted brand gains 100 and rival brands lose about 75 of them.

The second counts the units rivals actually lose. It works from store scanner data, which records every sale at the checkout rather than a sample of homes, and it subtracts something the first study leaves in. Your promotion also pulls extra shoppers into the category that week, and a share of those shoppers buy a rival, so part of the 75 comes straight back. The net loss to rivals lands nearer 33 units for every 100 you gain [van Heerde and colleagues]. The rest splits at roughly a third each between volume you borrowed from your own future weeks and demand that is genuinely new to the category.

Which number you reach for depends on who is in the room. Your share story runs on the gross figure, because that is what moved on the shelf. The retailer's category story runs on the net figure, because the category kept most of what you took. A buyer working from one paper and a brand manager working from the other can spend a whole meeting disagreeing about numbers they both have right.

The four buckets below work from the net figure, because that is the money a promotion has to pay itself back out of. They also split the borrowed third into its two sources, and that step is an illustration rather than a published set of numbers.

The four buckets, plain English

The single uplift number on an event report is four very different kinds of volume added together:

• Cross‑brand switching (~33%): shoppers who bought your brand instead of a competitor. Real volume, the manufacturer wins, the retailer's category is flat.
• Acceleration (~33%): shoppers who would have bought the brand anyway, but bought it during the promo window instead of next week. The dip after the event is where this volume came from.
• Own‑brand cannibalization (~8%): shoppers who switched from one of your other SKUs to the promoted SKU. Same brand, same shelf, thinner margin.
• Category expansion (~26%): genuinely new occasions or new shoppers who entered the category because of the promo. The only stream that grows the pie for everyone.

Add the two within‑brand pieces together, acceleration and cannibalization, and you get 41%: volume that either moved in time or moved between your own packs. Subtract that from 100 and you are left with the 33% switching plus the 26% expansion that together represent real, additional volume.

Whether the pack can be stored decides where the rest of the uplift goes

The share that comes out of rivals barely moves between categories. What changes completely is the quarter that does not, and one property of your product decides it. Where a shopper can store the pack, that same three‑way split runs 75 / 3 / 22: almost nobody new comes into the category, and the extra volume is existing buyers filling a pantry. Where they cannot store it, the split is 75 / 17 / 8: the promotion brings in people who were not going to buy at all, and they buy a normal amount.

That one property changes what your promotion is even for. On a storable pack you are buying pantry space, which is worth something against a competitor and nothing as new demand. On a pack nobody can hoard you are buying an extra shopper, and that is the only stream that grows the category for you and the retailer at the same time.

Where promotions genuinely raise consumption, and where they only pull the same volume forward

The same work compared how much shoppers bought on deal against how long they then waited before buying again, which separates two outcomes that look identical on an uplift report.

• They consumed more: bacon, potato chips, soft drinks and yogurt. Volume rose and the gap to the next purchase held steady, so the extra units were eaten and drunk rather than stored.
• They only bought earlier and bigger: toilet paper, coffee, laundry detergent and paper towels. Volume rose and the wait to the next purchase stretched to match it.

Detergent is the cleanest illustration. On promotion, buyers took about 60% more of it, and then waited about 60% longer before buying again. The pantry absorbed the entire difference.

One caution before you use that list as a verdict on your own category. The test only looks at people who bought, so it cannot see a promotion that brings a brand new household into the category. One of the categories in that work shows no extra consumption at all on this measure, and yet about 2% more households buy it when it is promoted, which is real growth the test has no way of seeing.

The spread is far wider than any single number admits

Across those 13 categories the switching share runs from 49% in butter to 94% in margarine. Two products a shopper will happily swap for each other, and their promotions do close to opposite things. Every benchmark on this page is the middle of a very wide distribution, and the only number that decides anything is the one from your own category.

Applying the split is easy. The hard part is setting the four shares for your own category and your own event, and being straight with yourself while you do it, instead of reaching for the ~33% / ~33% / ~8% / ~26% averages because they are there.

The decomposition equation

The share you actually get to bank as new volume is:

This four‑bucket split has no store switching or halo in it, so net incremental volume here is what is left once acceleration (the dip after) and own‑brand cannibalization come out of the uplift. Or equivalently:

Within‑Brand Share is acceleration plus own‑brand cannibalization added together: the volume that moved in time and the volume that moved between your own packs. On the benchmark averages that comes to about 41%.

Worked numbers at the standard split

Take an event that shows a 2.0x uplift on a baseline of 1,000 units per week, sustained over a 2‑week window. Apparent uplift is 2,000 units. Split it four ways on the benchmark shares, which are an illustration built around the one finding that has been verified, that about a third of the uplift is switching:

• Brand switching: 2,000 x 33% = 660 units (real, manufacturer wins)
• Acceleration: 2,000 x 33% = 660 units (timing‑shifted, dips next month)
• Cannibalization: 2,000 x 8% = 160 units (own‑brand swap, no net manufacturer gain)
• Category expansion: 2,000 x 26% = 520 units (real, retailer category grows)

Net incremental = 660 + 520 = 1,180 units, a net share of uplift of 1,180 / 2,000 = 59 percent of the apparent uplift (net incremental volume divided by the uplift)

A 2.0x event report becomes a much smaller 1.59x event after credible decomposition. Work the return on investment off the 2,000‑unit uplift and you overstate it by roughly 70% against the same calculation run on the 1,180 units that are really yours.

Why the split is rarely the textbook split

Categories where the pack keeps for months (pasta, detergent, multipack soft drinks) run a higher acceleration share, because a shopper can buy six and put five away. Categories where it does not keep (bakery, dairy, fresh produce) run less acceleration and more category expansion, because the promotion created an extra occasion the shopper had to take that week or lose. A deep price cut runs more within‑brand volume than a display worth the same money, because the price cut mostly reaches people who already buy you. New product launches run higher category expansion in the early weeks while distribution is still building.

The ROI consequence

Plug the decomposed uplift into the standard ROI formula:

Promo ROI = (Net Incremental Volume x GP per unit minus Promo Cost) / Promo Cost

Take a pack carrying a 35% gross margin, cut it 20%, and read a 2.0x uplift off the report. The return looks positive on that number, and it frequently turns negative once you take out the within‑brand 41%. That swing is exactly why so many FMCG brands run promotional calendars that look healthy on event reports and look broken on the annual P&L.

An illustrative scenario in carbonated soft drinks. A premium brand with a 6‑can multipack at $5.99 shelf price runs two events at the same retailer in the same quarter. Same brand, same SKU, same retailer, same baseline. Both events come back showing the same 2.4x uplift on the report written after the event. The CMO is happy with both. Split the volume four ways and they stop looking like the same event.

Event A: 25% off TPR on the brand's hero SKU, 2 weeks, no display

Apparent uplift: +1,400 units over the window. Split it with shares set for this kind of product, a mainstream pack that keeps for months. They put a little more on within‑brand volume than the textbook numbers, because a multipack of cans stores easily in a pantry or a garage:

• Brand switching: 1,400 x 28% = 392 units
• Acceleration: 1,400 x 38% = 532 units
• Cannibalization: 1,400 x 12% = 168 units (the brand has 4 other soft‑drink SKUs at the same retailer)
• Category expansion: 1,400 x 22% = 308 units

Net incremental = 392 + 308 = 700 units, a net share of uplift of 700 / 1,400 = 50 percent of the apparent uplift (net incremental volume divided by the uplift)

Within‑brand share: 50%. The price cut is mostly paying regular buyers to shop a week early, and taking a further slice off the brand's own value‑tier multipack sitting on the same shelf.

Event B: 10% off plus gondola end display, 2 weeks

Apparent uplift: also +1,400 units over the window. The offer behind it is very different: a shallower cut, backed by a display that catches shoppers who were not going to walk down the soft‑drinks aisle at all.

• Brand switching: 1,400 x 36% = 504 units (display puts the brand in front of competitor‑brand shoppers)
• Acceleration: 1,400 x 22% = 308 units (smaller, because the depth is too shallow to motivate stockpiling)
• Cannibalization: 1,400 x 8% = 112 units
• Category expansion: 1,400 x 34% = 476 units (display drove genuinely new occasions)

Net incremental = 504 + 476 = 980 units, a net share of uplift of 980 / 1,400 = 70 percent of the apparent uplift

Within‑brand share: 30%. The shallow cut with a display behind it delivered 40% more genuinely new volume than the deep price cut, on the same apparent uplift.

Reading the comparison

The post‑event report would tell the CMO that both events were "+140% uplift" wins. The source‑of‑volume decomposition tells the trade‑spend committee that one event delivered 700 real units and the other delivered 980. At the brand's $2.20 gross profit per unit, that 280‑unit gap is $616 of incremental gross profit per store, multiplied across 200 matched stores = $123,200 of profit difference between two events that scanned identically on the uplift report.

Use the published split to check your own numbers rather than to produce them, and do not carry it out to three decimal places. Whatever model your trade promotion optimization (TPO) team is running, this is the reading you hold it up against.

Three places the rule pays off

First, pre‑approval challenge. When a planner submits a projected ROI based on a 2.5x uplift forecast, ask: "How much of that uplift are you assuming comes from within our own brand, and why is it different from the roughly 41% benchmark?" If they cannot answer, the projection is sitting on a number they have not interrogated. The split turns a vague forecast into a defensible one.

Second, post‑event reality check. The post‑event report shows a 2.2x uplift. Take roughly 41% off the uplift as a floor for the within‑brand share. Does the resulting net‑incremental number still look good? If the event only looks good before that deduction, the original story was probably overstated.

Third, category‑mix prioritization. Some categories sit reliably above the 41% benchmark on within‑brand share: shoppers stock up, they rarely swap brands, and the range on shelf is narrow. Promotional returns in those categories are reliably lower. The benchmark lets you spot them early, while the annual plan is still being built, and move trade money toward categories where more shoppers are willing to switch.

Two questions that tell you which case you are in

Neither needs a new model, and most teams have never been asked either one.

Does our category store? If the pack sits happily in a pantry for a month, expect your promotions to buy pantry space rather than new shoppers, and leave category growth out of the numbers you present unless you can show it happening. If the pack has to be used within the week, it runs the other way, and every dollar you put behind a display works harder than a dollar you put into a deeper cut.

When we promote, do buyers wait longer before they come back? Work out the average time between one purchase and the next, first for the weeks you promoted and then for the weeks you did not. If the gap stretches by about as much as the volume rose, the promotion moved sales in time and created none. If volume rose and the gap held steady, people genuinely consumed more, and that is the event worth funding again.

Three places the rule misleads if used naively

First, new product introductions. Early in a launch, switching and category expansion both run far above the 33% norm, because a brand with almost no buyers yet has almost nothing of its own to take volume from. Judge a launch event against these benchmarks and its return will look worse than it is.

Second, competitive defense events. When you are promoting because a competitor moved first, the switching share can jump to 60% or more, since you are pulling back shoppers who had just left. Held against that, the 33% benchmark understates how much volume you genuinely won back.

Third, bonus‑pack and pack‑size mechanics. The split was estimated on price‑promotion data. Bonus packs and pack‑size moves change how much a household uses in a week, on top of changing whether they buy at all, so the volume splits differently. Use the benchmark on those as a rough direction check, and do not read it as the split itself.

How the rule travels with you

When you move from one company to another, or one category to another, the actual decomposition shares will differ. What carries across is the habit of writing the share down. Senior RGM leaders who have internalized this split rarely ask "what was the uplift?" without immediately asking "and what share of it was within‑brand?". Asking that second question is what separates the people who look after trade money from the people who only spend it.