What is reattribution?

What reattribution records

A user installs an app, uses it, drifts away, then sees a re-engagement ad and comes back. There is no new install — the app was already there, or is being reinstalled by someone already counted.

Reattribution is the event that records this. It credits the returning session to the campaign that prompted it, and it is measured against a reattribution window that defines how long the credit remains eligible.

Why it is easy to over-claim

Re-engagement targets people who already know the product. A meaningful share of them were going to return regardless — they always did, on their own cycle — and any ad served to them in the meantime will collect the credit.

This is the same structural problem as retargeting, in a more concentrated form, because the targeted population is defined by prior engagement. Reported reattribution performance is therefore reliably better than actual causal lift, and the gap is often large.

The check is a holdout: withhold re-engagement from a random slice of the lapsed population and compare return rates. Most teams that run this test for the first time revise their re-engagement budget.

Distinguishing it from a new install

• Never blend the two. A reattributed user cost less to reach and is worth a different amount than a genuinely new one.
• Set the inactivity period deliberately. Too short and you reattribute normal usage gaps; too long and real reactivations are missed.
• Report re-engagement separately in cost-per-acquisition figures, or blended CPA will flatter the account.
• Hold out. The only number that separates prompted returns from spontaneous ones.