Ad fraud

How it works

App advertisers often pay per install. Whoever is credited with the install gets paid, and credit normally goes to the last recorded click before the first open. Fraud targets that rule from two directions.

• Stealing credit. A fraudster sends clicks that no person made, in bulk or at the exact moment an install is under way. When a real person later opens the app, the fake click is the most recent one on record and takes the credit.
• Inventing users. Bots, emulators or rooms full of phones install and open the app, or forged messages report installs that never took place. The advertiser pays for people who are not there.

When it matters

The risk grows with the number of partners you pay per install and shrinks when you judge results by purchases. A fake user does not subscribe, so a paid event is harder to forge than an install. When the risk applies to you, and what to do about it, is covered in ad fraud prevention basics.

How it relates to attribution

Fraud is an attack on attribution itself, because the fake click enters the same record of touchpoints that the MMP uses to assign credit. That is why detection sits inside the MMP: it sees every partner's clicks next to the installs. Fraud tooling is not in every plan. It is one of the two jobs with a less clear boundary in what an MMP does not do. Airbridge, for example, includes fraud detection in Growth and not in Core.^[1]