Measurement

Incrementality resultsyou can take to finance.

Randomised holdouts show the conversions your campaigns caused, next to the attributed results from your MMP. Report spend and outcomes down to the individual app.

Why measure incrementality

Why attribution alone overstates retargeting.

Retargeting reaches users who already know your app, and many return without an ad. Last-touch attribution still credits the final impression.

A holdout separates the returns your ads caused from the returns that would have happened anyway.

How it works

How holdout testing works.

  1. Randomise

    Eligible users are assigned by chance to an exposed group or a holdout group before delivery starts.

  2. Run

    The exposed group sees your ads. The holdout group is eligible too, and sees none from the campaign.

  3. Compare

    The gap in conversions between the groups is incremental lift. Spend divided by incremental conversions gives iCPA.

MMP attribution

Attributed and incremental results side by side.

Your MMP remains the system of record for installs and events. We read its results and show them beside the holdout comparison.

  • Attributed follows your MMP's rules and lookback windows
  • Incremental comes from the randomised comparison
  • The gap shows how much of the attributed total the campaign caused
Attribution counts every return, and incrementality counts only the returns the ads caused.

iOS

iOS measurement with SKAN and AdAttributionKit.

SKAdNetwork and AdAttributionKit postbacks are aggregate, delayed and capped by Apple. We plan campaign structure around those limits and run holdouts where signals allow. See retargeting after ATT.

Reporting

Report down to app level.

See spend and conversions for every app where your ads ran. Cut results by format, creative, audience and campaign, and export them for your own analysis.

Compare spend against conversions for each app, then shift budget to the best performers.

Questions

What teams ask about measurement.

How large should the holdout be?

The share is agreed with you in advance. Larger audiences can detect smaller lifts.

Why do attributed and incremental results differ?

Attribution asks which ad came last. Incrementality asks what the ads caused, and it is usually the smaller number.

Can we export app-level reports?

Yes. Export details are confirmed during onboarding.

Get started

Ready to bring your users back?