Where Does Promotional Volume Come From? Switching, Stockpiling, Dips, and Cannibalization

A promotion's incremental volume is a mix of shoppers switching from a competitor, shoppers buying earlier or buying more than they otherwise would, and shoppers switching from another of your own items. Only part of it is new demand for the category, and part of it is borrowed from the weeks after the event.

This is the fourth layer of promo analysis and the one the buyer cares about most, because the retailer's economics depend on where the volume came from. It is also the layer the syndicated file answers least directly, so this chapter leans on the research that established how promotions behave and then says what you can and cannot see in your own data.

How promotions work has been studied empirically for decades; the field's canonical summary of what was known by the mid-1990s is Blattberg, Briesch, and Fox's review "How Promotions Work" [8]. The findings below are specific studies in specific categories. They describe mechanisms, and your own data supersedes them as a benchmark.

Switching: less than the elasticity numbers suggest

Household-panel research long attributed roughly 74% of the promotion elasticity to brand switching, with the remainder to timing acceleration and quantity increases. Van Heerde, Gupta, and Wittink showed that this does not mean competitors lose 74 of every 100 units a promoted brand gains. Converting the elasticity decomposition into unit sales, they found that about 33% of the unit increase is attributable to losses by other brands in the category [1]. The rest is shoppers buying sooner or buying more.

The practical reading is that a promotion which "steals share" in the panel data may be borrowing from its own future in the store data. That matters when you claim category growth to a buyer, and it matters when you plan the next event.

Acceleration and the post-promotion dip

A post-promotion dip is the fall in sales after an event ends because shoppers bought earlier or stocked up during it. Household studies predict it. Store data rarely shows it plainly, which Van Heerde, Leeflang, and Wittink called one of the mysteries of store-level scanner modeling: the dip should be there, and visual inspection almost never finds it [2]. Using distributed lead-and-lag models on two categories, tuna and toilet tissue, they estimated the combined pre- and post-promotion dip at 4% to 25% of the promotion's current-period sales effect, consistent with household-level findings [2].

Macé and Neslin then asked what makes dips bigger. Across 39,441 dip elasticities in 83 stores and 10 categories, both pre- and post-promotion dips were stronger for high-priced, frequently promoted, mature, high-market-share items, and post-promotion dips were more prominent where the promotion pattern was less predictable [3]. If your flagship item is the one you promote most, expect the largest dip there.

Neither estimate is a constant to apply to your business. Both are reasons to look at the four to eight weeks after an event before crediting the event with everything that happened inside it.

Cannibalization: incremental to the item is not incremental to the brand

When a promoted item's incremental volume comes out of a sibling item on the same shelf, the item gained and the brand did not. Circana's dictionary defines "incremental to brand" as incremental that excludes cannibalization of your own items, calculated by subtracting sales lost on sister items from the focal item's lift, and gives the example of a launch that lifts the flagship by 200 units while the old size loses 50, for a net 150 incremental to brand [4]. The same arithmetic applies to a promotion: read the sibling items in the same store-weeks before you report the number.

Circana's promoted volume entry frames the timing question the same way: compare promoted against base mix to evaluate whether discounts drive incremental demand or shift timing [5].

Category expansion: what the retailer is paying for

A retailer's category manager does not earn anything when a shopper switches from one brand to another inside the category at a lower price. The pass-through study found exactly that in retailer behavior: retailers passed through less on products with higher "demand clout," the ability to steal sales from other products, and the authors note that retailers generally stand to gain from category expansion, not brand switching [6].

Nielsen's European source-of-volume modeling, published alongside its 2015 benchmark, quantified the split for the events it studied. It classed 32% of promotional volume as subsidization (low benefit to both parties), 24% as mutual growth, 27% as manufacturer growth at the retailer's expense, and 17% as retailer growth at the manufacturer's [7].

Those are European figures from a decade ago and they will not match your category. The frame is what travels. Retailers fund category growth; the other three quadrants are a harder sell.

What to net out before you claim a return

Chapter 5 computed return on incremental units. This chapter says which of those units to trust.

Net out	Why	Where you find it
Subsidized volume	Would have sold anyway; you paid the discount on it	Promoted minus incremental, in the syndicated file
The post-event dip	Borrowed from the following weeks	Base and actual in the weeks after the event, same retailer
Cannibalized volume	Came from your own sibling items	Sibling items' incremental (usually negative) in the same store-weeks
Switched volume, if you are claiming category growth	The retailer did not gain from it	Competitor and category movement in the same store-weeks

After netting, the incremental that remains is the volume you can defend in the meeting. It is smaller than the lift table's number. It is also the only number the buyer will believe once she has run the same query.

What POS can and cannot see

Point-of-sale data shows what sold, where, and when; it cannot show which household bought it or what they would have bought otherwise. That is panel data's job, and the Syndicated Data guide's opening chapter covers the difference. In practice: cannibalization, competitor movement, and the post-event dip are all readable from POS at the retailer level, because they are store-week facts. Switching at the household level and true stockpiling are panel questions, and without panel access the research figures above are your best proxy for their size.

The question to ask any tool

"Show me who lost the volume I gained." It is the source-layer question in one sentence, and it can only be answered if the competitive set is defined and stored. Ask it of your analyst and of any software you evaluate. If the answer requires re-entering the competitor list every time, the tool does not hold your brand context.

Sous can answer it because the competitive set lives in brand context: the products, categories, retailers, and competitors the brand has told it about persist across sessions, so a promo read can put the named competitors' and the category's movement in the same store-weeks alongside your event without anyone re-typing the list. The brand context chapter of the Agents in CPG guide and the four components chapter of the Brand Context Layer guide cover how that memory is built and kept.

Common questions

What is a post-promotion dip? The fall in sales after an event because shoppers bought earlier or stocked up. One store-data study in two categories estimated it at 4% to 25% of the event's current-period effect. Dips are stronger for high-priced, frequently promoted, mature, high-share items.

How much of a promotion's lift comes from competitors? Household-panel elasticity decompositions attribute about 74% to switching, but in unit terms the cross-brand loss is about one third. The rest is shoppers buying sooner or buying more.

What is cannibalization in promo analysis? Incremental volume on the promoted item that came out of your own sibling items. Circana's "incremental to brand" measure nets it out.

Why does the retailer care where the volume came from? Because the retailer gains when the category grows. A shopper switching brands inside the category at a lower price gives the account nothing, and the pass-through study found retailers pass through less on products that mostly steal share.

Can I see switching in POS data? Not at the household level. You can see competitor and category movement in the same store-weeks, which is the retailer-level proxy. Household switching is a panel question.

What should I net out before claiming return? Subsidized volume, the post-event dip, and cannibalized volume. If you are claiming category growth, also the switched volume.