Incrementality testing has moved from being a specialist measurement technique used by a small number of sophisticated advertisers to being a mainstream measurement practice across most large advertising programs. The May 2026 picture is more grounded than the early enthusiasm suggested. Worth a working read of where the practice sits.

The headline shift.

Most large Australian advertisers in 2026 run some form of incrementality testing as a regular part of their measurement practice. The frequency varies — some advertisers run continuous tests, some run quarterly, some run on a specific channel-by-channel cadence. The integration of incrementality results into media planning and budget allocation has become more disciplined than it was three years ago.

The shift away from sole reliance on last-click and multi-touch attribution as the basis for budget decisions has been the most consequential change in the measurement practice through 2024–2026. The combination of platform attribution data, multi-touch attribution where applied, marketing mix modelling, and incrementality testing as the four-corner stool has become the default operating model for serious advertisers.

The methods that are working.

The platform-led conversion uplift studies have been the entry point for many advertisers. Meta’s Conversion Lift, Google’s Conversion Lift, TikTok’s Conversion Lift studies all give advertisers a way to run incrementality measurement directly inside the platform. The methodology is well-developed, the operational set-up is reasonably straightforward, and the results have credible statistical properties when the test is properly powered.

The geo-based holdout tests have been a major part of the incrementality testing toolkit. The advertiser holds advertising spend in a set of geographic regions for a defined period and compares the conversion outcomes in the holdout regions against control regions. The geo-test approach works well for advertisers with national reach who can afford the opportunity cost of the holdout.

The audience-based holdout tests work for advertisers with the data infrastructure to randomise the audience exposure. The intervention and control groups are defined at the audience level rather than the geographic level. The approach is more precise than geo-tests but requires more sophisticated implementation.

The structural variation analysis — comparing periods of significant media intervention against periods without — has been used as a lighter-weight approach. The methodology is less rigorous than experimental approaches but provides useful signal when experimental approaches are not feasible.

The lessons from running tests at scale.

The most common operational failure mode is under-powered tests. The advertiser runs a test that does not have enough sample size, enough run time, or enough conversion volume to produce statistically meaningful results. The output is then mis-interpreted as a finding when it is actually noise. The discipline of pre-test power analysis has been the single most important practice improvement.

The second most common failure is test contamination. The holdout group receives exposure that it should not have, the geo regions have cross-region exposure, the test runs alongside an organic event that confounds the result. The test design needs to account for these risks.

The third common failure is misinterpretation of test results that are noisier than the advertiser expects. The output of a single test is rarely a clean answer. The accumulation of test results across multiple channels, multiple time periods, and multiple audiences is where the useful learning emerges. The advertiser who treats one test as the answer is going to be misled.

The integration of incrementality results into operational planning has been an area of meaningful operational learning. The findings have to be translated into actionable budget and bidding decisions. The teams that have built the right operational interfaces between the measurement team and the media planning team are getting more lift from the testing program than the teams that have not.

A few practical observations.

The channels that consistently show high incrementality across most advertisers in 2026 include established direct mail programs for the right audiences, podcast advertising in the right contexts, and well-targeted CTV and digital video. The channels that often show lower-than-expected incrementality include brand-search advertising for established brand searches, retargeting display advertising for already-engaged audiences, and broad-targeting social campaigns without strong creative.

The branded search incrementality picture has been the most uncomfortable finding for many advertisers. The branded search spend that platforms continue to recommend often has incrementality lower than the platform attribution data implies. The advertisers that have adjusted their branded search spend on the basis of incrementality testing have generally seen total program performance hold or improve.

The retargeting incrementality picture has been similarly uncomfortable. The retargeting spend on already-engaged users often has lower incrementality than the platform attribution credits. The reallocation of retargeting budget to prospecting has been a common adjustment.

The brand-building media incrementality picture has been more positive than some skeptical observers expected. The well-targeted brand campaigns running in the right environments often show meaningful incrementality on conversion outcomes within reasonable time horizons. The brand-building media is doing real work when it is well-executed.

The outlook for incrementality testing through the rest of 2026.

The practice is becoming more democratic — the tools, the platform support, and the analytical capabilities required to run incrementality testing have all become more accessible. The advertisers who have not yet built a regular incrementality testing program are at a measurement disadvantage relative to the advertisers who have.

The integration with marketing mix modelling is the area of continued development. The combination of MMM and incrementality testing produces a measurement picture neither approach can produce alone. The advertisers building the integrated measurement framework are operating with better decision inputs than the advertisers using either approach in isolation.

The next twelve months should see continued operational maturation of incrementality testing as a mainstream measurement practice. The advertisers running it well will continue to refine their programs and accumulate institutional knowledge. The advertisers not running it will increasingly be at a measurement disadvantage in budget allocation decisions.