What Is Incrementality Testing? A Practical Guide for Marketers
Incrementality testing measures whether your ads actually caused sales — or whether those customers would have bought anyway. Here's how it works, when to use it, and how it fits with MMM.
Incrementality testing measures whether your advertising actually caused a sale — or whether that customer would have bought anyway without seeing your ad. It's the difference between knowing your ads are correlated with sales and proving they caused them.
This matters more than most marketers realize. Every ad platform reports conversions, but those numbers include people who were already going to buy. Your Google Ads dashboard might show 500 conversions last month, but how many of those 500 people would have purchased even if they never saw your ad? That gap — between reported conversions and truly incremental ones — is where most wasted ad spend hides.
How Does Incrementality Testing Work?
Incrementality testing works like a scientific experiment. You split your audience into two groups: one that sees your ads (the test group) and one that doesn't (the control group). Then you compare the outcomes.
If the test group converts at 4.2% and the control group converts at 3.1%, the difference — 1.1 percentage points — is your incremental lift. That lift is what your advertising actually produced. Everything else would have happened without you spending a dollar.
There are two main approaches:
Geo-Based Testing
Geo-based incrementality testing (also called geo-testing or matched-market testing) splits geographic regions instead of individual users. You run ads in some markets and go dark in others, then compare sales between them.
This is the most common approach in 2026 because it doesn't require user-level tracking — which makes it immune to iOS privacy changes, cookie deprecation, and consent requirements that have broken user-level measurement. You're comparing aggregate sales in Dallas (ads running) versus Houston (ads paused), not tracking individual clicks.
How to run a geo test:
- Pick matched markets. Find pairs of regions with similar demographics, sales history, and seasonal patterns. Dallas and Houston. Portland and Seattle. Chicago and Detroit.
- Establish a baseline. Run both markets identically for 2–4 weeks to confirm they behave similarly.
- Run the test. Turn ads on in one market and off (or reduce spend) in the other for 4–8 weeks.
- Measure the difference. Compare sales between the test and control markets. The gap is your incremental impact.
User-Level Holdout Testing
User-level holdout tests randomly assign individual users to test and control groups within a single platform. Meta, Google, and most DSPs offer built-in holdout testing tools.
The advantage is precision — you're comparing identical audiences. The disadvantage is that it only works within a single platform and depends on that platform's ability to track users, which is increasingly limited.
Why Is Incrementality Testing Important?
Attribution models — whether last-click, multi-touch, or platform-reported — all share the same fundamental flaw: they measure correlation, not causation. They tell you what a customer touched before converting, but not whether that touchpoint actually mattered.
This creates two expensive problems:
Over-crediting channels that intercept existing demand. Branded search is the classic example. If someone types "your brand name" into Google and clicks your ad, did that ad cause the sale? Almost certainly not — they were already looking for you. But your attribution model gives full credit to paid search.
Under-crediting channels that create demand. TV, podcast ads, YouTube pre-roll, and out-of-home advertising often plant the seed that eventually leads to a search or direct visit. Attribution models rarely credit these channels because they can't track the full path. Media mix modeling captures this effect at the aggregate level, and incrementality testing validates it with experimental evidence.
Brands that measure incrementality consistently find that 20–40% of their "conversions" would have happened without any advertising. For some channels — especially retargeting and branded search — that number can exceed 60%.
How Does Incrementality Testing Fit with MMM?
Incrementality testing and media mix modeling answer different questions and work best together:
| MMM | Incrementality Testing | |
|---|---|---|
| What it answers | Which channels drive the most incremental revenue across my entire budget? | Did this specific campaign or channel actually cause sales? |
| How it works | Statistical modeling of historical spend and sales data | Controlled experiments with test/control groups |
| Scope | All channels at once | One channel or campaign at a time |
| Speed | Initial insights in 10–15 days, continuous updates | Each test takes 4–8 weeks |
| Best for | Budget allocation, scenario planning, cross-channel strategy | Validating specific channels, calibrating your MMM |
The most effective measurement approach uses both. MMM tells you where to allocate your budget across all channels. Incrementality tests validate those recommendations by experimentally proving whether a specific channel is actually driving the lift your model predicts.
Think of it this way: MMM is your GPS. Incrementality testing is checking the road signs to make sure the GPS is calibrated correctly.
When Should You Run an Incrementality Test?
Not every channel or campaign needs an incrementality test. They require pausing or reducing spend in some markets, which means accepting short-term revenue risk. Run them when the stakes justify it:
- High-spend channels you've never validated. If you're spending $50K+/month on a channel and your only proof it works is the platform's own reporting, that's a test worth running.
- Channels with suspiciously high reported ROAS. If a platform claims 8x ROAS but your overall business results don't reflect it, incrementality testing will show you the real number.
- Before major budget increases. Planning to double your Meta spend next quarter? Test incrementality first to make sure you're scaling real performance, not inflated reporting.
- To calibrate your MMM. Run incrementality tests periodically to validate and refine your media mix model's channel-level estimates.
What Are the Limitations of Incrementality Testing?
Incrementality testing is the gold standard for proving causation, but it has practical constraints:
- You can only test one or two things at a time. Each test takes weeks and requires pausing spend. You can't test every channel every quarter.
- Geo tests need sufficient market volume. If your business only operates in three metro areas, you don't have enough markets to create valid test and control groups.
- Results are point-in-time. A test run in January might not reflect July performance. Seasonality, competitive activity, and market conditions change.
- Short-term measurement only. A 6-week test captures immediate sales impact but misses longer-term brand effects that build over months.
These limitations are exactly why incrementality testing works best as a complement to MMM, not a replacement. MMM captures long-term and cross-channel effects continuously. Incrementality tests provide periodic ground-truth validation.
How Do You Get Started?
If you've never run an incrementality test, start simple:
1. Pick your highest-spend, lowest-confidence channel. Which channel gets the most budget with the least proof it works? That's your first test.
2. Design a geo test. Identify 4–6 matched market pairs. Use historical sales data to confirm they behave similarly. Plan for a 2-week baseline period followed by a 4–6 week test period.
3. Go dark in the control markets. Pause or significantly reduce spend in the control markets. Keep everything else — pricing, promotions, organic activity — identical between test and control.
4. Measure the lift. Compare sales between test and control markets. Calculate the cost per incremental sale. Compare that to what your attribution model was reporting.
5. Use the results. If the channel is truly incremental, scale it with confidence. If it's not, reallocate that budget to channels that are actually driving growth. Feed the results into your MMM to improve its accuracy going forward.
The brands seeing the best results in 2026 aren't choosing between attribution, MMM, and incrementality testing. They're using all three — attribution for tactical optimization, MMM for strategic budget allocation, and incrementality testing to keep the whole system honest.