Incrementality Testing: The Only Attribution Method That Proves Causation

Why Most Advertisers Are Measuring the Wrong Thing

Here is a scenario that plays out constantly in performance marketing: You are running Google Search ads and seeing a 5x ROAS. You pause the campaign for two weeks as a test. Revenue barely moves.

What happened? Those Search clicks were largely from people who were going to buy anyway. You were not driving new demand — you were capturing existing intent. And you were paying for it.

Incrementality testing answers the question every other measurement approach cannot: Would this customer have converted without the ad?

What Incrementality Is (And Is Not)

Incrementality = Conversions caused by your advertising

Not attributed to it. Not correlated with it. Actually caused by it.

Every standard attribution model measures which ad touchpoints occurred before a conversion. None prove that the touchpoint caused the conversion. The only way to prove causation is a controlled experiment.

The Three Main Methods

Method 1: Geographic Holdout Test

Divide your market into test and control geographic groups. Continue ads in test regions; suppress them in control. Compare conversion rates.

Best for: Brand awareness, display, TikTok — channels where attribution is weakest.

How long: 4–6 weeks minimum for statistical significance.

Method 2: Platform-Native Lift Studies

Meta Conversion Lift Studies:

• Meta randomly divides audience into test (sees ads) and holdout (does not)
• Reports true incremental conversions and cost per incremental result
• Minimum spend: ~$30,000/month for statistical significance

Google Lift Studies:

• Available for YouTube and Display
• Measures branded search lift and conversion lift
• Lower spend requirements than Meta

Method 3: On/Off Testing

Pause the channel for 2 weeks, measure business impact. Simplest but most disruptive.

Compare total business revenue in both periods, controlling for seasonality. The difference = estimated incremental value.

How to Interpret Results

After your test, you will have an incremental lift rate — the percentage of conversions genuinely caused by your advertising.

Common Findings That Surprise Advertisers

Branded Search is often 20–40% incremental. Most branded clicks happen because the user already intended to visit your site. You are paying for navigation, not demand.

TikTok often has 60–80% incrementality. Discovery-based platform — almost all TikTok conversions are truly new demand.

Retargeting: 30–50% incremental. Some retargeted users would have returned anyway.

The Quarterly Testing Calendar

Reallocating 20–30% of budget based on incrementality insights typically improves total revenue while reducing total spend.

The bottom line: Every other measurement approach tells you what happened. Incrementality testing tells you what you caused. Run at least one lift test per quarter. The findings will consistently surprise you.