How Do You Read Base, Incremental, and Promotion Measures?

Total volume can be cut three different ways, and confusing them causes most promotional misreads. Promo against non-promo is observed. Base against incremental is modeled. And within promoted volume, incremental against subsidized is another split that helps determine whether a promotion was worth running.

Three cuts, not one

Let's start with the observed cut of promo against non-promo. Promo volume is what sold while the item was under any promotional condition; said another way, what the shopper bought while a deal was active. Non-promo volume is what sold at regular shelf price with no promotional support. Both are recorded, not inferred [1].

The modeled cut is base against incremental. Base volume is what the item was expected to sell in the period. Incremental volume is what sold above that. This split is estimated by the provider from the item's own selling history, and each provider uses its own algorithm, so two providers can report different base for the same item in the same week.

The third cut is the one that matters commercially. Promoted volume contains both incremental volume, which the promotion created, and subsidized volume, which would have sold anyway but received the discount. Subsidized volume is promoted volume minus incremental volume, and can appear in data extracts under column labels like "Base Units, Promo".

What subsidized volume costs you

Work an example. 20% of your volume sold on promotion. Incremental came in at 10% of total volume. That means half of everything you promoted was volume you already had, sold at a discount you funded.

This is the entire economics of trade spend, and it is the number a recap deck almost never shows. A promotion that moved 10,000 units did not buy you 10,000 units. If base was 7,000, it bought 3,000, and the real cost per incremental unit is a multiple of what the deck implies.

Every promotion carries some subsidized volume; the objective is to keep it small. The two levers are timing and frequency, the subject of "When to promote" below.

Price architecture: four prices, not three

Syndicated pricing comes in layers, and each price is computed across a different slice of volume.

• Non-promoted price: the average price paid on volume that sold with no promotional support. This excludes promoted base volume and is calculated as non-promoted dollars divided by non-promoted units.
• Promoted price: the average price paid on volume that sold under a promotional condition.
• Base price: the provider's modeled estimate of the item's price absent promotion, computed across base volume. Because base volume includes the subsidized units that sold at a discount, base price sits below non-promoted price in any period with real promotional activity. The two are not synonyms, though they are constantly used as though they were.
• Average retail price (ARP): total dollars divided by total units, blending everything. This is what "price per unit" usually means in a topline report.

Two misreads to know. The classic one: a brand believes its price position eroded because ARP fell, when base price never moved; it simply promoted deeper or more often. The reverse hides real erosion under a quiet promotional pullback.

The less obvious one: promoted ARP can sit above total ARP. That looks impossible until you remember ARP is a weighted average of a mix. If your promotions skew toward larger packs or higher-priced items, the promoted mix carries a higher average price than the portfolio does overall, even though every promoted item sold at a discount.

Lift: what one promoted week is worth

Lift expresses response to promotion, and the word carries at least three meanings in active use. It can mean incremental volume itself. It can mean total sales above base in a period, whatever caused them. And it can mean promoted lift, restricted to weeks with merchandising support. Confirm which one a report uses before comparing anything to anything.

As a rate, a lift of 150% means a promoted week sold 150% above its base. Different merchandising conditions produce different lift, and your own data will put numbers on that ladder per account. Those numbers, rather than general expectations about which tactic works best, are what should decide where the next trade dollar goes.

Merchandising reach, and retailer compliance

Whether the promotion ran at all is its own measure. % ACV with merchandising divides the ACV of stores where your product had merchandising support (feature, display, or price reduction) by the market's total ACV [2]. It answers "how much of the market's volume saw this event?", weighted by store size like every ACV measure.

Read it next to lift. Strong lift on 8% merch ACV is an execution gap: the event works, almost nobody ran it. Weak lift on 60% merch ACV is an offer problem: everybody ran it and shoppers shrugged. Both cases produce the same disappointing incremental total, but they call for different fixes, and the retailer conversation that follows [3] goes very differently in each case.

When to promote: aligning spend with demand

Subsidized volume is largely a timing problem. If you promote into a period when demand is already strong, more of what sells would have sold anyway. If you promote into a weak period, you are discounting to move volume the category is not asking for.

The tool is a seasonality index: base units in each week divided by average weekly base units over a trailing window, indexed to 100 [4]. Building it on base rather than total volume is deliberate: it keeps past promotional activity from contaminating your read on underlying demand.

Lay percentage of units on promotion over that index and the mismatch, where it exists, is immediate. It is common to find a brand's heaviest promotional week of the year sitting in a period where the seasonality index is below 100, while the genuine demand peak carries far lighter support. That is subsidized volume being manufactured on purpose, usually because the promotional calendar was set by the retailer's schedule rather than by the category's demand curve.

One further point worth carrying into provider choice: seasonality curves differ by channel for the same category. A natural-channel demand peak and a conventional one do not necessarily fall in the same weeks, so a total-US roll-up can average away the pattern you actually needed to see.

Handle with care: the decomposition is modeled

Because base and incremental come from a model, they carry restrictions the raw sales columns do not, and promotional analysis has two of them worth naming.

First, modeled decompositions are estimated per cut, so the split does not survive naive re-aggregation. Base and incremental computed at item level in one market do not meaningfully sum into a brand-level national split, because the baseline was estimated per item, per store cluster, per week, and adding the pieces across cuts compounds the modeling error. Re-pull the decomposition at the level you need rather than assembling it from parts. That rule and its siblings get a full treatment in the aggregation trap.

Second, promotional response is retailer-specific. Promotions are executed retailer by retailer, so a channel-level promotional read is arithmetically valid and analytically empty. Promotional assessment can show you what worked in one geography and should be tried in another; it cannot sensibly evaluate promotion at a channel total.

All of this makes promotional analysis the most tedious part of reading syndicated data, and the most expensive part of a CPG business to get wrong. It rewards weekly-level work, because that is the grain at which individual events can be isolated and understood. The period-over-period grind, meanwhile, is mechanical: the same promo recap rebuilt every 4 weeks with new data. In Sous, that recap is a workbook saved as a template; when the next period lands, the lift table, the merch reach read, and the narrative recalculate on the new file, and the analysis you built in March is still working for you in September.

Common questions

What is the difference between base sales and incremental sales? Base is the volume expected to sell in a period. Incremental is what sold above that. Base is modeled by the provider from the item's selling history; only the total is directly scanned.

What is subsidized volume? Volume that would have sold anyway but was purchased under a promotional condition, so you funded a discount on sales you already had. It equals promoted volume minus incremental volume.

What counts as merchandising? Feature, display, price reduction and special pack, plus their combinations, collectively the causal conditions. % ACV with merchandising tells you the size-weighted share of the market where any of them ran.

Why is my average price falling if I never changed price? More or deeper promotion. ARP blends promoted and non-promoted selling, so a heavier promotional calendar drags it down while base price sits still. Check base price first.

Why is my promoted price higher than my average price? Mix. If your promotions skew toward larger packs or higher-priced items, the promoted mix can carry a higher average price than the portfolio overall, even at a discount.

Can I sum incremental sales across markets? You can, and the result will not mean much. The decomposition is modeled per cut, so re-pull at the level you need rather than adding the pieces.

When should I promote? Into demand rather than away from it. Build a seasonality index from base units, lay your percentage of units on promotion over it, and check whether your heaviest support is landing in your strongest weeks.