← The Growth Lab · MEASUREMENT
Incrementality Testing on a Growth-Stage Budget
Oct 17, 2026 · Read time: 3 min · DV8 Pacific
Incrementality testing has a reputation for requiring large budgets, data science teams and expensive tools. Large-scale programs do. But growth-stage brands can get reliable, decision-ready answers with simpler designs, as long as the test is planned before the campaign starts.
The principle
Every incrementality test compares two groups: one that could see the advertising and one that could not. The groups must be as similar as possible in every other way. The difference in outcomes between them is the effect of the advertising.
Four designs that work for smaller brands
Geo split. Run ads in some regions and hold out comparable regions, then compare sales trends. This is often the most practical design for brands selling through retailers, because retail sales can be tracked by region. Choose regions with similar distribution, seasonality and sales history.
Audience holdout. Randomly withhold ads from a portion of the eligible audience and compare purchase rates. Some ad platforms and retail media networks support this directly. It is precise but needs enough volume to detect a difference.
On/off periods. Alternate periods with and without advertising and compare sales. This is the simplest design and works with very small budgets, but it is the least reliable, because seasonality and promotions can distort results.
Retailer sales lift. Match ad exposure to a retailer's own purchase data, usually through a privacy-safe data environment such as a clean room. This is powerful for consumer packaged goods brands that sell mainly through one large retailer.
How to run a test well
Decide the question first. For example: does this campaign add sales at our main retailer among new customers?
Choose one primary measure. Sales, new-to-brand orders or branded search volume. Do not change it after the results arrive.
Set the comparison before launch. Pick test and control regions, or the holdout share, before the first ad runs.
Record a baseline. Capture several weeks of data before the campaign to understand normal variation.
Run long enough. Allow enough time to see a stable difference, longer for products that are bought infrequently.
Keep everything else steady. Avoid launching promotions or price changes in only one group during the test.
Read results honestly
Small tests produce estimates, not certainties. A modest lift with a wide margin of error is still useful: it tells you whether a channel deserves a larger, better-powered test. Report results as ranges, compare them with attributed results, and use the gap to recalibrate how you read platform dashboards.
Look beyond the first order
Many growth campaigns look expensive on first-order returns and profitable over time. Track new-to-brand customers and their value over several months, including repeat purchases. A channel that brings in loyal new customers at a higher upfront cost can be a better investment than one that efficiently recycles existing buyers.
Sources: IAB and MRC, "Retail Media Measurement Guidelines," January 2024; IAB, "Guidelines for Incremental Measurement in Commerce Media," 2025; DV8 Pacific analysis.