Why incrementality is the only metric that proves marketing’s real impact

Attribution shows who gets credit. Incrementality shows what your marketing truly caused.

In an era of automation and privacy restrictions, understanding the real lift behind your campaigns is the only way to prove what’s working.

This article breaks down what incrementality measures, why it matters, and how to test it across today’s major ad platforms.

The problem with ‘great’ results that don’t actually drive growth

Marketers love big numbers – CTR, impressions, and ROAS all sound great in a deck. 

But what if those results don’t represent real business growth?

For example, a paid search campaign reports a 10x ROAS.

It might sound amazing. But if 90% of those conversions would’ve happened organically without your ads, your true ROAS is much lower.

That’s where incrementality comes in. It measures how many of those conversions happened because of your marketing, not in spite of it. 

It’s the difference between taking credit and creating value.

When eBay paused its brand search ads, a large-scale field experiment found sales were largely unchanged – showing those ads were capturing existing demand, not creating new growth.

Dig deeper: Incrementality testing in advertising: Who are the winners and losers?

What incrementality actually measures

Incrementality quantifies the causal lift from your marketing. It’s a measure of what changed because your campaign existed.

In practice:

• Test group: People or regions exposed to your ads.
• Control group: Similar people or regions not exposed.
• Lift: The difference in outcomes between the two groups.

If your test group produced 1,250 purchases and your control group 1,000, your campaign drove +250 incremental sales (+25% lift) – the part that wouldn’t have happened without you.

Why incrementality matters more than ever

Traditional metrics hint at performance – incrementality proves it. 

• It reveals waste: You can see where ads simply capture organic demand (like branded search for established brands).
• It informs budget: You’ll know which channels actually generate new revenue and which just take credit for it.
• It builds trust: Finance and leadership teams care about what changed, not what was “attributed.”

In short, incrementality aligns marketing metrics with business outcomes.

4 reliable ways to measure incrementality

Each incrementality test asks the same question: What would’ve happened without my ads?

These four methods offer different ways to answer it, depending on how much control and data you have.

1. Randomized holdout (user-level testing)

Also called randomized controlled trial (RCT), this is the cleanest way to measure lift.

You randomly divide your audience. 

• One half sees your ads (test).
• The other half doesn’t (control). 

Your campaign directly causes any difference in conversions or revenue.

Platforms like Meta (Facebook/Instagram) and Google Ads (YouTube, Display) now offer built-in lift tests that handle randomization and reporting automatically.

When to use: Digital campaigns with measurable conversions and sufficient volume.

2. Geo holdout testing

When you can’t randomize individuals, randomize regions.

Choose comparable locations, such as two cities with similar purchase patterns. 

Run your ads in one and pause in the other. The difference in results reveals your incremental lift.

Why it works: Real-world scale, works across offline or mixed channels (e.g., TV, radio, or retail).

Caution: Match regions carefully and allow for enough time to balance out local fluctuations.

Dig deeper: The ROAS illusion: Rethinking what Google Ads success looks like

3. Synthetic control and causal modeling

When experiments aren’t possible – say you ran a national campaign – you can estimate what would’ve happened without your ads using data models.

Tools like Google’s CausalImpact and Meta’s GeoLift build a synthetic “twin” of your audience or region based on past trends. 

Comparing actual results to this modeled baseline reveals your campaign’s incremental effect.

It’s not as airtight as a true experiment, but it’s a strong option for retrospective or large-scale campaigns.

4. Marketing mix modeling (MMM)

MMM uses historical, aggregated data (e.g., spend, impressions, sales) to measure each channel’s contribution over time.

It’s not an experiment, but when calibrated with incrementality studies, it provides a strategic, privacy-safe view of ROI across channels.

MMM answers questions like:

• “What share of sales did Meta vs. Search drive last quarter?”
• “What happens to revenue if we cut TV spend by 20%?”

Think of MMM as the macro view, and lift testing as the ground truth that keeps it accurate.

How ad platforms support incrementality

Major ad platforms now offer built-in tools to help marketers measure lift directly – no manual setup required.

• Meta (Facebook/Instagram): Offers Conversion Lift and Brand Lift studies – randomized tests that directly measure incremental conversions or brand outcomes.
• Google Ads: Provides Conversion Lift for YouTube and Display, with “ghost ads” simulating withheld exposure for the control group. You can also run A/B experiments with Drafts and Experiments for Search.
• TikTok: Recently launched Conversion Lift Studies, showing that a large percentage of conversions measured by lift were exclusive to TikTok – meaning they wouldn’t have occurred through other channels.
• Amazon Ads: Has limited native lift testing; most advertisers use geo-based experiments or work with measurement partners to determine incremental impact.

How to run your first incrementality test

Here’s a straightforward process to get started:

• Choose one campaign and KPI: For example, Facebook campaign targeting add-to-cart conversions.
• Form a hypothesis: “This campaign will increase conversions by at least 10% over baseline.”
• Set up control and test groups: Use a platform lift test or create your own random or geo holdout.
• Run the test for a full conversion cycle: Avoid overlapping changes (like price updates or promotions).
• Collect data and calculate lift:
  • Incremental conversions = Test − Control
  • Lift (%) = (Test − Control) ÷ Control × 100
  • iROAS = Incremental revenue ÷ Spend
• Make decisions: Scale what’s proven incremental. Pause or rethink what isn’t.
• Repeat quarterly: Use learnings to calibrate attribution models and budget plans.

Common pitfalls to avoid

Even well-designed tests can fail if the setup or timing is off. 

Watch out for these common mistakes that can distort your results or hide true lift.

• Running tests that are too small or too short: Without statistical power, you can’t trust the result.
• Contaminating the control group: Make sure control users or regions truly don’t see your ads.
• Testing too many variables at once: Keep it simple – one campaign, one goal.
• Relying solely on attribution: Attribution models show credit, not cause.
• Forgetting to document results: Keep an “incrementality log” test setup, data, and learnings so your team can build institutional knowledge.

Make lift your new baseline

Three major shifts make incrementality indispensable today:

• Privacy restrictions limit what we can track – experiments measure lift without personal data.
• Automated ad systems optimize for conversions, not necessarily incremental ones.
• Economic pressure demands proof of value. When budgets tighten, finance wants to know what happens if you turn ads off.

Attribution shows where conversions came from. Incrementality shows whether marketing caused them at all. 

In a world where every click is already claimed by someone, lift is how you prove your ads aren’t just showing up – they’re driving growth.

Start with one clean test, validate key channels, and make lift your new baseline. Because if your marketing doesn’t create new demand, it’s not really working.

Dig deeper: PPC experimentation vs. PPC testing: A practical breakdown