How incrementality works: baseline, lift and incremental sales
Every incrementality figure starts from a counterfactual: what the same shoppers would have bought if the ads had never run. IAB names a credible counterfactual as the first requirement of any causal measure. IAB/MRC note that incrementality methods rest on an estimate of baseline sales. A randomized controlled trial is designed to give the least biased estimate. It splits shoppers into those eligible to see the ads and those held back.
Three numbers follow from that comparison:
- Incremental sales: sales of the advertised products among exposed shoppers, minus the baseline set by the control group.
- Incremental lift: the percentage increase over that baseline.
- Incremental ROAS (iROAS): incremental sales divided by ad spend.
In a June 2023 engineering post, Instacart described comparing total sales, organic plus attributed, between the two groups. Counting total sales matters because the held-back shoppers have no attributed sales at all.
Incrementality vs attributed ROAS
Attributed return on ad spend (ROAS) counts every sale that followed an ad inside a network's attribution window. IAB separates the two measures by their job: attribution and ROAS show what happened, and incrementality measures whether marketing caused the result. The difference between them is the sales credited to ads that shoppers would have made anyway.
In January 2025, Digiday reported that advertisers were expected to put incrementality metrics ahead of ROAS that year. ROAS had long been the default way to judge retail media campaigns. The question behind the shift, in the words of one analytics CEO: "Would that sale have happened regardless of the ad?" One agency executive said marketers doubt the incrementality retailers report because some networks "always take 100% attribution for sales". In 2023, PepsiCo's then retail media lead told AdExchanger that what advertisers really need is incremental return on ad spend.
Incrementality at Instacart and Kroger
Instacart argues that retailers hold what incrementality measurement needs: control over who sees an ad, and complete sales data to compare. In its 2023 post, it described a randomized experiment for each advertiser. Shoppers were split into an ads group and a holdout group whose experience was "as if the advertiser stopped spending entirely." Competitors' ads still appeared.
Kroger Precision Marketing makes the baseline argument directly. "Retailers can report incremental sales with greater accuracy than any other media company," a KPM vice president told Digiday in January 2025. "We know exactly what the baseline organic sales should be in any category we sell." That month KPM announced incremental sales reporting for self-service programmatic campaigns. KPM's FAQ says its incrementality methods compare exposed and unexposed groups to isolate true impact.
Ways to measure incrementality
IAB sorts incrementality methods into four families with different causal strength:
| Method family | Examples | Causal strength |
|---|---|---|
| Experiment-based | Randomized control tests, holdouts and ghost ads, matched markets | Strong |
| Model-based counterfactual | Synthetic control, machine learning propensity models | Strong to moderate |
| Econometric | Marketing mix modeling (MMM), time-series regression | Moderate to weak |
| Hybrid proxies | New-to-brand percentage, baseline vs exposed analysis, platform-reported incrementality | Weak |
The trade-off runs from one end of the table to the other. Experiments bring high rigor but are costly, time-intensive and prone to contamination without proper controls. Hybrid proxies are easy, fast and always on, and they lack causal rigor. IAB/MRC recommend randomized controlled trials where feasible. Designing and running one is covered under incrementality testing, and building the unexposed group under control groups. Lift studies and media mix modeling have their own entries.
Why it matters for brands and sellers
For big calls such as budget allocation, IAB advises methods with stronger causal footing. It also notes that incrementality shifts as competition, shopper behavior and tactics change, so a result needs re-measuring over time.
Scope matters too. IAB says experiments usually stay inside one platform and are rarely practical for comparing networks. It leaves that job to marketing mix modeling. Products differ as well, and IAB/MRC expect results reported by SKU where relevant, since most spend promotes specific SKUs. A lift read on one network, or on one hero SKU, is weak evidence for the rest of a portfolio. Brands get more from reading incrementality network by network and product by product. The results can then feed how they plan a retail media budget and judge retail media effectiveness.
Related terms
- Incrementality testing: designing and running a test.
- Control groups: the held-back shoppers who set the baseline.
- Closed-loop attribution: how networks match ads to sales.
- Return on ad spend (ROAS): the attributed metric incrementality corrects.
- Lift studies: studies that report a campaign's lift.
- Media mix modeling: the econometric route to incrementality.
Frequently asked questions
What does incrementality mean in marketing?
In marketing and advertising, IAB defines incrementality as the extra business outcomes a campaign or tactic directly drives. The comparison point is what would have happened with no marketing at all. In retail media that outcome is usually incremental sales: purchases above the baseline of what shoppers would have bought anyway.
What is the difference between incrementality and attribution?
Attribution assigns credit for sales that followed an ad, at the individual or household level. Incrementality measures the sales the ad caused by comparing exposed shoppers with a baseline, usually a control group. A campaign can show strong attributed sales and little incremental impact when shoppers would have bought anyway.
How do you calculate incrementality?
Compare sales per shopper of the advertised products between an exposed group and a comparable control group over the same period. Multiply the difference by the number of exposed shoppers to get incremental sales. Divide that by the baseline for lift, or by ad spend for incremental ROAS. Incrementality testing covers the design.
Is new-to-brand the same as incrementality?
New-to-brand counts how many sales came from a brand's first-time customers. It does not show whether the ad caused those purchases. IAB classes new-to-brand percentage as a hybrid proxy with weak causal strength, useful for quick directional reads that need confirming with causal methods.
What is incremental ROAS (iROAS)?
Incremental ROAS is incremental sales divided by ad spend: the sales an ad caused for each dollar, measured against a control group. It sits beside attributed ROAS, which also credits purchases shoppers would have made anyway. In 2023, PepsiCo's then retail media lead called it the return advertisers really need.
