How Do You Calculate Promotional Lift Correctly?

To calculate promotional lift, isolate the promoted store-weeks, take the provider's modeled baseline for those weeks, subtract it from actual sales to get incremental, and express lift as incremental divided by base. Do it per retailer, per event, at the promoted product group level, and say which lift convention you used.

Lift is the layer of promo analysis most brands already produce, and the one most often produced wrong. The errors are rarely arithmetic. They come from the wrong weeks, the wrong baseline, the wrong denominator, or the wrong level. This chapter is the procedure; the Syndicated Data guide's chapter on base, incremental, and promotion measures defines the measures the procedure uses.

The six steps

Step	What you do	Why it matters
1. Isolate the event	Select the store-weeks flagged under the promo condition you care about, and confirm the promoted price sits below the everyday price by a real margin	Nielsen's own benchmark framework begins here: identify promotion weeks by depth of discount against everyday price [8]. A "promotion" with no price gap is a data artifact
2. Confirm the baseline	Read the provider's base for those exact weeks. Base is modeled volume expected at regular price without merchandising, estimated from the item's history with promo effects removed [6]	The baseline is an estimate, not a count. Two providers model it differently, so never mix a Circana base with Nielsen actuals
3. Compute incremental	Incremental equals total sales minus baseline in the promoted weeks [7]	This is the only volume the promotion created
4. Compute lift and state the convention	Lift equals incremental divided by base, times 100 [4]	Other conventions exist (below). Name yours
5. Split subsidized from incremental	Subsidized volume equals promoted volume minus incremental volume	The units you discounted but would have sold anyway. Defined in the Syndicated Data guide; it decides the economics in Chapter 5
6. Break lift by tactic, with the right denominator	Lift under a display condition uses base in the display stores as its denominator, not total base [4]	A display lift divided by chain-wide base understates the display and misprices the tactic

Then read the result next to merchandising reach (% ACV with merchandising, covered in the Syndicated Data guide) so you know whether a weak number is an offer problem or an execution problem.

A worked example, illustrative

Take one item group at one retailer, two promoted weeks under feature and display, with round numbers that are invented for the purpose.

The provider's modeled base is 1,000 units a week. Actual sales were 1,600 units in the first promoted week and 1,400 in the second. Regular price is $4.00; promoted price is $3.00.

Now the arithmetic.

• Promoted volume over the two weeks: 1,600 + 1,400 = 3,000 units.
• Base over the two weeks: 2,000 units.
• Incremental: 3,000 minus 2,000 = 1,000 units.
• Lift, incremental over base: 1,000 / 2,000 = 50%.
• Subsidized volume: 3,000 promoted minus 1,000 incremental = 2,000 units sold at a $1.00 discount that would have sold at $4.00 anyway.
• Incremental retail dollars: 1,000 units at $3.00 = $3,000.

A 50% lift reads well on a slide. The two numbers that will decide Chapter 5 are the 2,000 subsidized units and the $1.00 discount on every one of the 3,000 units that scanned. Hold on to them.

Which lift is this?

The word carries several meanings and the providers do not all use the same one. This guide uses incremental divided by base, the Tip Sheet's glossary definition [4], because it answers the question the buyer asks: how much more than expected did the event sell?

NielsenIQ's dictionary defines promotional lift as the proportion of promotional sales that are incremental, with the example that selling $100 on promotion against $80 expected is a 20% lift [1]. On our example that convention gives 1,000 / 3,000 = 33%, the same event described as the share of promoted sales that the promotion created. Neither is wrong. Reporting one and comparing it to a benchmark built on the other is. The Syndicated Data guide's chapter lists a third usage, total sales above base in a period whatever caused them, which is broader still.

NielsenIQ's dictionary also makes a point analysts forget: incremental lift can be negative when the item does not sell enough extra volume to compensate for the price reduction [2]. In dollar terms, a shallow lift on a deep discount can leave promoted dollars below base dollars. That is a real result, and a good workbook reports it rather than clipping it at zero.

Denominators and levels

Two choices quietly change the answer.

The denominator. Databases carry both a general promotional lift, which uses total base as the denominator, and tactic-specific lifts, where a display lift uses the base in the display stores as its denominator [4]. If a display ran in a quarter of the chain, dividing its incremental by chain-wide base shrinks the display's apparent effect by roughly a factor of four. When you build a lift ladder by tactic, each rung needs its own base.

The level. Analyze at the promoted product group, the set of items the retailer prices and promotes together, rather than the individual UPC [5]. Shoppers saw the group on the shelf; the retailer planned the event around it; a single UPC inside a multi-pack group can show a lift that is really the group's mix shifting.

And always at the retailer. Promotions are executed retailer by retailer, so a channel-level lift blends events that had nothing to do with each other. The Syndicated Data guide calls this "valid but meaningless," and it is the first thing a buyer will notice if you bring a channel number to a single-account meeting.

The two reads NielsenIQ recommends

NielsenIQ's own guidance for brands on its platform names two reports worth mirroring in any tool. The first is a promo-versus-non-promo decomposition: break total sales into incremental and non-incremental (base), then follow the incremental down to the tactic that produced it, and run it on strategically chosen markets inside the promotion window [3]. The second is a support-and-lift comparison across retailers for the same event, so that when one account trails the others on the same offer you know whom to speak to; NielsenIQ frames this as holding retail partners accountable, and notes the same report can be run against competitors instead [3].

Both reads are retailer-level by construction. Neither averages anything across accounts.

What "good" lift means

There is no universal benchmark and this guide will not invent one. Nielsen's own conclusion after analyzing 76 million event weeks was that there is no single formula for efficient trade promotion that fits every market, category, channel, brand, and item [8]. Your benchmark is your own lift ladder: this item group, at this retailer, under this tactic, at this depth, over the last several events. A 50% lift is strong if the same event last quarter did 30% and weak if it did 90%. The ladder is what you compare to, and the ladder only exists if the events were measured the same way each time.

Where Sous fits

Promo lift analysis is one of the saved template workflows on the Sous homepage: a 10-step analysis that measures promotional effectiveness with lift metrics by tactic. Structurally the workbook it produces holds an event table (store-weeks, condition, promoted price, depth), a lift-by-tactic table with each rung on its own base, a merchandising reach read, and a written narrative, and it re-runs when the next period's file lands. No performance figures are attached to that description, and none should be; the numbers are yours.

One check inside it is worth naming because any tool should have it. Sous's verifier confirms that base plus incremental equals total on every promo query. When a baseline has been mismatched to the wrong weeks or the wrong provider, that identity fails, and the query is stopped before a wrong lift reaches a slide. It is the mechanical version of the habit in step 2.

Common questions

How do you calculate promotional lift? Isolate the promoted weeks, subtract the provider's baseline from actual sales to get incremental, and divide incremental by base. State the convention and compute per retailer and per event.

What is a good lift percentage? There is no universal answer. Compare the event to your own prior events at the same retailer under the same tactic. Nielsen's 2015 benchmark concluded there is no single formula that fits every category and account.

Can lift be negative? Yes. If the extra volume does not compensate for the price reduction, incremental dollar lift is negative. Report it.

Why does my display lift look small? Check the denominator. If you divided display incremental by chain-wide base rather than base in the display stores, you understated it.

Should I compute lift by UPC? Compute it at the promoted product group, the items the retailer priced and featured together. UPC-level lift inside a multi-pack group mostly measures mix.

Is the baseline a real number? No. It is the provider's model of what would have sold without promotion, estimated from the item's history. Two providers produce two different bases for the same weeks.