Incrementality is showing up in more measurement conversations and more of your toolkit. 

Google’s recent global launch of Meridian GeoX is one example, giving you another way to run geo experiments and use the results to calibrate marketing mix modeling (MMM).

So let’s talk about it.

As incrementality becomes more accessible, having a test at your fingertips is one thing. Knowing how to use it well is another.

Many marketers are still working through where incrementality fits within their broader measurement strategy, what makes a test trustworthy, and what good incrementality actually looks like. 

You need to understand how that result was produced, how uncertainty was handled, and how well the methodology has been validated.

A quick refresher: What are you actually testing?

Incrementality measures the additional sales or conversions your advertising caused, not the ones you would have earned anyway.

Even though there is a fixed calculation (iROAS = incremental revenue / media spend), there are various ways you could test incrementality.

GeoLift testing is one way to estimate that impact. You change advertising spend in selected regions and use comparable regions to estimate what would have happened without the change. The difference between your actual results and that estimate is your incremental lift.

Want a closer look? Here’s how GeoLift testing works.

Why the methodology behind your lift matters

An incrementality test will tell you, “Your campaign drove 10% lift.” Before you scale spend, it’s worth understanding how your testing tool reached that conclusion.

Every geo test needs to estimate what would have happened without the advertising change. The methodology determines how that baseline is built, how normal sales fluctuations are separated from advertising impact, and how uncertainty is calculated.

Different approaches make different assumptions. Even with the same data, one method might find evidence of lift while another finds the result inconclusive.

A trustworthy method needs to do both: detect meaningful impact when it’s there and avoid reporting lift when it isn’t. Sometimes, the most useful answer is that your test hasn’t produced enough evidence to justify a change.

A Recast simulation study illustrates these trade-offs. It compared Meta GeoLift, Google Matched Markets, CausalPy, and Google CausalImpact across 32,000 model fits. Under the study’s simulated conditions, the tools differed substantially in their ability to detect real lift and avoid false positives.

The study highlights a bigger issue with broad stroke incrementality solutions: the methodology behind the test you choose can affect both your lift estimate and, most importantly, how confidently you can act on it.

That’s why you should look beyond whether a tool measures incrementality. When evaluating an incrementality solution, ask how its methodology has been validated and whether it holds up under conditions like yours.

What’s behind Triple Whale’s approach?

At Triple Whale, evaluating different methodologies against real customer data has shaped how we approach incrementality.

When building our incrementality solution, we evaluated Meta’s GeoLift, then explored alternatives including CausalPy and CausalImpact to understand how these approaches performed on the real customer data we were working with.

Commerce data presents challenges for incrementality testing. Seasonality, promotions, and uneven sales volumes can make it harder to distinguish advertising impact from other changes in demand. A methodology needs to account for that complexity while giving marketers a clear view of the uncertainty around their results.

Across our evaluations, we encountered limitations in how the existing implementations handled the customer data we were working with. And that led us to develop our own Bayesian methodology.

That’s where scale becomes one of Triple Whale’s advantages. With access to real commerce data from 65,000+ businesses, we can validate and continuously refine our methodology against the seasonality, demand shifts, and sales patterns that commerce businesses actually experience.

Simulations are valuable because researchers know the true effect. However, real-world validation adds another essential perspective: how the test behaves across the messiness of real commerce data.

That same approach carries into Compass, where incrementality sits alongside attribution and MMM as part of one measurement system. 

Attribution helps teams understand daily performance, MMM supports weekly budget planning and total channel contribution, and incrementality validates causal lift. 

Together, the signals calibrate and strengthen recommendations, giving teams a more complete body of evidence for knowing where to invest next.

What to ask before you act on lift

You don’t need to become a marketing scientist to evaluate an incrementality result. Start with three questions:

• Could this test detect a change that matters to your commerce business? Your budget, duration, and geographic design should support the question you’re asking.
• How certain is the result? Read the range around the estimate, not just the headline percentage. Inconclusive doesn’t mean zero impact.
• How was the methodology validated? Ask about simulations, real-world testing, and the conditions where the approach becomes less reliable.

Bottom line

Incrementality testing tools give you another way to connect experimentation with marketing planning. That’s good news if you want a clearer understanding of what your advertising contributes.

As your options grow, look beyond the promise of “proving lift.” You need to understand how a test works, where its limits are, and whether the evidence is strong enough to support your next move.

Triple Whale’s approach is built and continuously validated against real commerce data at scale, giving you incrementality measurement designed for the complexity of modern commerce.

A lift number starts the conversation. The methodology helps you decide what to actively do about it. Ready to test what’s truly working in your marketing? Schedule a demo.