Incrementality

Paid Media

Also: Incrementality Testing · Incremental Lift

Incremental lift = Conversions with the channel on minus conversions with it off
What it measuresCausation, not just correlation
How it's testedHoldout groups or geo tests
Watch forConfusing attribution with lift
Common findingSome spend is buying what would happen anyway

Quick definition

Incrementality is the measure of how many conversions a marketing channel actually caused, rather than simply claimed credit for. It's tested by comparing results with the channel switched on against results with it switched off, usually through a holdout group or geo-based experiment.

How it varies across Australia

Incrementality testing is still rare among mid-market Australian advertisers, more common in enterprise retail and finance where media budgets justify the setup cost. Where it has been run, brand and retargeting spend usually shows the widest gap between attributed credit and actual lift.

See acquisition benchmarks across Australian industries

What it actually means

Imagine turning off your retargeting campaign for two weeks and watching what happens to sales. If nothing changes, that campaign wasn't causing conversions. It was catching people who were going to buy regardless and taking the credit for it. That's the entire idea behind incrementality.

Most businesses run on attribution, which divides credit among the touchpoints a customer encountered before converting. Attribution assumes those touchpoints mattered. Incrementality tests whether they actually did, by comparing a group exposed to the channel against a matched group that wasn't.

This matters most for channels with strong intent-capture behaviour. Branded search, retargeting and email to existing customers all tend to look brilliant on CPA and ROAS reports because they're reaching people who were already close to converting. Incrementality testing routinely finds that a meaningful share of that spend is buying conversions that would have happened without it.

The uncomfortable part is that incrementality often contradicts what the attribution model says, which means the two systems answer different questions and neither replaces the other.

Attribution tells you who claimed the sale. Incrementality tells you whether the sale would have happened anyway.

How to calculate it

Incremental lift = Conversions with the channel on minus conversions with it off

Worked example. You run a two-week geo holdout on retargeting. Regions with the campaign live convert at 1,200 sales. Matched regions with the campaign switched off convert at 1,050 sales. The incremental lift is 150 sales, meaning the campaign only caused 150 of the 1,200 conversions it was credited with.

The Australian context

Australia's smaller population makes geo-based incrementality tests trickier to run cleanly than in the United States, since there are fewer comparable regions to split into test and control without one market bleeding into another through shared media or word of mouth. Businesses here often lean on holdout audience tests instead of geo splits, particularly for email and retargeting where the audience can be randomised directly rather than split by state.

Where people get this wrong

Treating attribution and incrementality as the same measurement.Attribution divides existing credit across touchpoints. Incrementality tests whether the spend caused new conversions. A channel can dominate attribution and still show close to zero incremental lift.
Running a holdout test for too short a period.Short tests get swamped by normal week-to-week noise in conversion volume. Most reliable tests need several weeks of data before the lift is statistically distinguishable from random variation.
Only testing the channels that look weak.The channels most worth testing are usually the ones with the best attributed numbers, like branded search and retargeting, because that's where inflated credit hides the most wasted spend.

Incrementality vs Attribution

IncrementalityAttribution
What it answersDid this spend cause the conversionWho gets credit for the conversion
MethodHoldout groups or geo experimentsRule-based or modelled credit assignment
Cost to runRequires a controlled test periodRuns continuously from existing data
Best forDeciding whether to keep funding a channelReporting and day-to-day optimisation

Related terms

Common questions

How is incrementality testing different from A/B testing?

A/B testing usually compares two versions of a single asset like an ad or a landing page. Incrementality testing compares having a channel switched on against having it switched off entirely, across a holdout audience or region, to measure whether the channel causes conversions at all.

Why would a high-ROAS channel show low incrementality?

ROAS is calculated from attributed conversions, which credit the channel for sales it simply caught rather than caused. Retargeting and branded search often reach people already intending to buy, so the attributed number looks strong while the actual incremental lift is small.

How long does an incrementality test need to run?

Long enough for normal conversion volatility to settle so the lift is statistically clear, typically several weeks depending on your conversion volume. Lower-traffic businesses need longer tests to get a reliable read than high-volume ones.

Do small businesses need incrementality testing?

Not always. It requires enough conversion volume to detect a meaningful difference between test and control groups. Smaller businesses often get more value from simpler holdout checks on their biggest spend channel rather than building a full testing programme.

Debrief

Get the next one

No spam. No fluff. Just the next article, straight to your inbox.

Keep exploring

About New Rebellion

New Rebellion is a marketing intelligence consultancy. We build tools, score Australian businesses on how their marketing actually performs, and publish Debrief every day. This dictionary is part of how we work in the open.

How we think →