How it works
At first open the measurement platform collects every engagement that falls inside its attribution window. Clicks normally outrank impressions: if there is any eligible click, the most recent click wins, even when an impression came later. Only when there is no click does the most recent impression get the credit. Earlier engagements are recorded as assists and receive no credit.
Why it matters for an app team
The model is simple, the same for every network and easy to audit, which is why networks accept it for billing. Its weakness is that it rewards whoever shows up last. Channels that create demand early in the journey look worse than they are, and fraudsters have a reason to fake a late click.
In practice
Example: a user watches a video ad from network A on Monday, clicks a banner from network B on Tuesday and a search ad from network C on Thursday, then opens the app on Friday. Network C gets the install. A and B appear only as assists.
Common confusion
Last touch is not the same as last click: an impression can win when no click exists. It is also not multi-touch attribution, which splits credit across several engagements, and it says nothing about whether the ad caused the install.