Key takeaways

• Incrementality testing compares a treatment group against a holdout group to reveal marketing's true causal impact — what would have happened without the campaign.
• Incrementality is really just Haus' plain-language word for causality, and GeoLift geo-experiments are Haus' core method for running these tests.
• Attribution assigns credit to touchpoints based on correlation; incrementality testing measures whether a channel or tactic actually caused a lift. In Haus' 2025 industry survey, only 20% of marketers named first- or last-touch attribution as their most-trusted measurement solution, and only 39% named MTAs.
• A/B tests tell you which version won; incrementality tests tell you whether the whole channel or tactic drove real results in the first place.
• Use incrementality results, like incremental ROAS and cost per incremental acquisition, to guide budget decisions, and let them anchor deeper tools like Causal MMM and Causal Attribution.

You increase your Meta budget for a month, and sales tick up. Was it the ad spend, or would those sales have happened anyway? Most marketers can't answer that with confidence, and the guesswork gets expensive once you're managing spend across a dozen channels.

Incrementality testing is how you get a real answer. Instead of guessing which touchpoints deserve credit, you run a controlled experiment and measure what actually changed.

This guide walks through what incrementality testing means, how it works, how it's different from attribution and A/B testing, and what you can test to make smarter budget calls.

What is incrementality testing?

Incrementality testing (sometimes called incrementality experiments) shows you what would have happened without your marketing. You compare business outcomes between a group exposed to a campaign and a group that isn't, and the difference between them isolates the true impact of that campaign, as Incrementality 101 puts it.

That's a fairly technical way of describing something simple: Incrementality is basically the non-scientist-friendly word for causality. An incremental conversion is one that happened specifically because of ad exposure, not one that would have happened regardless. Haus' Incrementality School explains it with a healthcare comparison: In a drug trial, a treatment group gets the medication and a statistically similar control group gets a placebo. The difference in outcomes between the two groups reveals the drug's actual effect. Marketing incrementality tests work the same way, just with a campaign standing in for the drug.

How incrementality testing works

Treatment and holdout groups

Every incrementality test needs two comparable groups: a treatment group that sees the marketing and a holdout group that doesn't. Because the two groups are otherwise similar, any gap in outcomes can be attributed to the marketing itself, rather than to a hunch about which channel gets the credit.

GeoLift: Haus' core method

GeoLift geo-experiments are Haus' core incrementality methodology. In a standard 2-cell test, regions are split into a treatment group exposed to the marketing and a holdout group that is not; comparing the two isolates true incremental impact — what would have happened without the campaign. A spend-increase test is a different design: Treatment regions raise budget while control regions stay at current spend, which answers whether extra investment still drives lift. A direct-to-consumer brand testing incremental Meta spend, for example, might raise budget in treatment regions and hold control regions at normal spend, then measure the difference in sales. Haus' Incrementality Experiments help businesses measure the causal impact of marketing across online and offline channels using these rigorous, geography-based tests, producing clear, defensible answers about which channels deserve more investment and which don't.

Other ways to measure incrementality

Geo-experiments aren't the only methodology. User-level experiments, typically run inside ad platforms, randomize individual users into exposed and unexposed groups. Observational studies compare before-and-after periods without a true control group, which makes them faster but less rigorous. Geo-experiments tend to offer the best balance of statistical rigor and practicality, which is why they're Haus' default.

Incrementality testing vs. attribution

Attribution models credit a conversion to the touchpoints a customer interacted with along the way — last-touch, first-touch, or multi-touch attribution (MTA), which splits credit across several interactions. The problem is that a click happening before a purchase doesn't prove the click caused it, and how incrementality differs from attribution comes down to exactly that: Attribution frames impact as crediting a customer event to a tactic, while incrementality frames impact as the change in customer events caused by a change in that tactic. That's a meaningful difference, and it shows up in Haus' 2025 industry survey: Only 20% of marketers named first- or last-touch attribution as their most-trusted measurement solution, and only 39% named MTAs.

Multi-touch attribution tries to fix single-touch attribution's blind spots by dividing credit across the ads, emails, and search queries that preceded a purchase instead of handing 100% of the credit to one click. But multi-touch attribution still measures correlation, not causation. A customer might have converted anyway without ever seeing a given touchpoint, and MTA has no reliable way to separate that possibility from a true incremental effect.

None of this makes attribution useless. Attribution is fast and gives you an ongoing, directional read on tactical performance. Incrementality testing takes more setup but earns its keep when you need to justify a big investment, evaluate a brand-new channel, or settle a real debate about where budget should go.

Incrementality testing vs. A/B testing

It's easy to lump incrementality tests in with A/B tests, but they answer different questions. An A/B test tells you which of two variants performs better — which subject line, which landing page, which creative. An incrementality test asks a more fundamental question: Did this channel or tactic cause a real lift at all, compared to not running it? You can run a beautifully designed A/B test on a channel that isn't driving any incremental revenue in the first place; incrementality testing is what catches that.

What you can test — and why it matters for budget decisions

Incrementality testing applies well beyond a single campaign. Businesses use it to measure channel effectiveness, halo effects between channels, efficiency, and diminishing returns at higher spend levels, upper-funnel channels that are notoriously hard to attribute, long-term impact, full media mix testing, secondary metrics like predicted lifetime value, and the true lift from promotions and seasonal pushes.

Haus' Principal Economist Phil Erickson frames the goal simply: Incrementality measures how a change in strategy causes a change in business outcomes, like how revenue would shift with a 10% higher ad budget. That's the kind of question finance teams ask, and it's why incrementality results, summarized in metrics like incremental ROAS and cost per incremental acquisition, tend to carry more weight in budget conversations than attribution numbers alone. Haus' own Laws of Incrementality also make a point worth remembering: Incrementality is unique to your business, which means someone else's benchmark isn't a substitute for testing your own channels.

Where incrementality fits with Causal MMM and Causal Attribution

Incrementality experiments aren't a one-off exercise — they're the causal foundation for the rest of a measurement stack. Haus' Causal MMM translates complex, cross-channel marketing data into actionable budget recommendations, modeling every channel in one unified framework. Unlike traditional MMMs built on historical correlations, Causal MMM is anchored on causal proof from your incrementality experiments, so your data works together instead of sending you conflicting signals.

Causal Attribution extends that same causal foundation to daily decisions. It uses a daily view of incrementality to guide spend decisions for every ad, ad set, and campaign, debiasing attribution data with your own seasonally adjusted experiments and Haus' Incrementality Index. The result surfaces business-critical metrics like ROAS, CPA, and revenue, but calibrated to what's actually incremental rather than what a platform simply took credit for.

Whether you're running your first geo experiment or building out a full measurement program, incrementality testing is the throughline. It's how Haus, the causal marketing platform businesses trust to optimize billions in ad spend worldwide, helps teams stop guessing and start proving what their marketing actually does.

If you want to go deeper on methodology, Haus' guides on incrementality testing fundamentals and understanding incrementality testing build on the definitions here.

Frequently asked questions

What's the real difference between incrementality testing and attribution?

Attribution assigns credit to the touchpoints a customer interacted with before converting, which is a correlation-based read. Incrementality testing measures the actual change in outcomes caused by a tactic, using a treatment group and a holdout group. Attribution can tell you a customer saw an ad before buying; incrementality testing tells you whether that ad is why they bought.

How is incrementality testing different from A/B testing?

An A/B test picks a winner between two variants of the same tactic, like two ad creatives. An incrementality test asks whether the tactic or channel is driving any real lift at all compared to not running it. Haus breaks this contrast down further in how incrementality experiments differ from A/B experiments.

How long does an incrementality test take?

Active testing often runs 2–4 weeks, plus a post-treatment window to capture delayed conversions. Exact duration depends on the channel and funnel: Demand-capture tactics are often about 2–3 weeks total, Meta conversion campaigns 3–4, video like YouTube or TikTok 4–6, and brand campaigns can run months. Larger holdouts or lower-volume channels generally need more time to produce a statistically reliable result.

What can I actually test with incrementality testing?

Common use cases include channel effectiveness, halo effects between channels, efficiency and diminishing returns at higher spend, upper-funnel channels that are hard to attribute, long-term impact, full media mix testing, secondary metrics like predicted lifetime value, and the lift from promotions and seasonal campaigns.

Is incrementality testing the same thing as marketing mix modeling?

No. Traditional marketing mix modeling relies on historical correlations and doesn't use a holdout group, which leaves it exposed to multicollinearity between channels. Incrementality testing uses an actual holdout to answer the counterfactual question directly. Haus' Causal MMM is built to close that gap by anchoring the model on causal proof from incrementality experiments rather than correlations alone.

What metrics come out of an incrementality test?

The two metrics that matter most for budget decisions are incremental ROAS (iROAS), which is incremental revenue divided by total spend, and cost per incremental acquisition (CPIA), which is total spend divided by incremental conversions. Both isolate the return you're actually getting, rather than the return a platform or attribution model claims credit for.

Where should I start if I've never run an incrementality test?

Start small and specific: Pick one channel or campaign where the budget decision is genuinely in question, define a treatment and holdout group, and run the test for at least a few weeks. Haus' getting started with incrementality testing guide walks through the setup in more detail.