All articles

July 22, 2026 · 5 min

How to Run an Incrementality Test for Influencer Marketing

You ran the campaign, the promo codes were redeemed, and the dashboard says the program paid for itself. Then someone in a budget meeting asks the question every marketer dreads: how many of those sales would have happened anyway? That is the incrementality question, and no attribution report can answer it, because attribution only tells you which touchpoint got credit, not whether the sale depended on the campaign at all.

Here is the direct answer. You prove incrementality by comparing people who could see your creator campaign against a similar group that could not. For most consumer brands the most practical version is a geo holdout test: run the campaign in some regions, hold it back in comparable regions, and compare sales between the two. If the exposed regions outperform the holdout regions beyond normal variation, the difference is your incremental lift. You do not need a data science team to run a useful version of this.

What does incremental mean in influencer marketing?

An incremental sale is a sale that would not have happened without the campaign. That distinction matters because creator content often reaches people who already know your brand. Some of them were going to buy this month regardless. If your measurement counts those sales as campaign wins, you will keep funding activity that merely collects credit instead of creating demand.

Attribution tools, promo codes, and UTM links all measure credit, not causation. A creator whose audience overlaps heavily with your existing customers can post great looking numbers while adding very little. Incrementality testing is how you tell the difference.

Why can't promo codes and UTMs answer this?

Tracking links and codes are still worth using, but they have two blind spots that no amount of dashboard work fixes:

  • They only capture buyers who click or redeem. Many people watch a video, remember the brand, and buy later through search. Codes miss them entirely.
  • They cannot show the counterfactual. A redeemed code proves a purchase touched the campaign. It does not prove the purchase needed the campaign. Discount hunters who were already in your funnel will happily use a creator's code for ten percent off.

So treat codes and links as evidence of reach into the purchase path, and treat a holdout comparison as the proof of whether that path created new demand.

What is a geo holdout test?

A geo holdout test uses geography as the control. You pick regions that behave similarly, activate creator campaigns in one set, keep the campaign out of the other set, and compare outcomes over the same window. Geography works well for influencer marketing because you can brief creators toward regional audiences, run paid amplification of creator content by region, and read results from sales data you already have.

The logic is simple: the holdout regions show you the world where the campaign never ran. Whatever the exposed regions do beyond that baseline is the lift.

How do you run one step by step?

  1. Pick a single outcome metric before you start: orders, signups, or revenue for a defined product line. One metric keeps the readout honest.
  2. Choose matched regions. Look for pairs of cities or areas with similar sales history over the past few months. Comparable trend lines matter more than comparable size.
  3. Assign exposure and holdout groups. Keep every other marketing input as stable as you can in both groups during the test.
  4. Run the campaign for a clean window. Four to eight weeks is a common range, long enough for creator content to circulate and convert.
  5. Compare each group against its own baseline. Calculate the change from baseline in the exposed regions and subtract the change in the holdout regions. What remains is your estimate of lift.
  6. Sanity check with supporting signals: branded search volume, direct traffic, and survey responses from the exposed regions should move in the same direction.

How do you read the results?

Expect noise. Weather, retail promotions, and seasonality all move regional sales, which is why the holdout exists. A useful reading is directional, not decimal precise. If exposed regions grew clearly faster than holdouts and your supporting signals moved with them, the program is creating demand. If the groups moved together, the campaign mostly harvested existing demand, and you should revisit creator selection or the offer before you revisit the budget.

Resist the urge to end a test early because early numbers look good or bad. Creator content compounds through shares, rewatches, and search, so short windows systematically undercount lift.

What if your budget is too small for a geo test?

You can still get a rough answer with lighter methods:

  • Stagger creator launches by a few weeks and watch for steps in your baseline sales after each wave.
  • Compare branded search and direct traffic during campaign windows against a pre campaign baseline of similar length.
  • Add a post purchase survey asking where the buyer first heard about you, and watch how the creator share shifts during campaign windows.

Each of these is weaker than a true holdout, but together they triangulate the same question: did new demand appear when the campaign ran?

The Bottom Line

Attribution tells you where credit landed. Incrementality tells you whether the sale needed the campaign. A geo holdout test is the most accessible way to answer that question: matched regions, one clean metric, a stable test window, and a directional reading. Even an imperfect test will teach you more about your creator program than another month of staring at last click reports.

Working with creators who are verified on real performance and conversion history makes tests like this far more likely to show lift worth paying for, which is exactly how ViralFusion builds its programs.