What is temporary attribution?

Why provisional credit exists

Privacy frameworks report late by design. A SKAdNetwork postback arrives after a randomised delay, and the conversion value may be updated during the measurement period before it is sent.

That leaves a gap between something happening and the reporting settling. Temporary attribution fills it with a best current estimate, so campaigns are not flying entirely blind while they wait.

The risk it creates

Provisional numbers get treated as final ones, particularly when they arrive in the same dashboard column as settled data.

Two failure modes follow. Optimising on provisional data means reacting to figures that will be revised, sometimes substantially. And comparing a recent period against an older one compares partly-settled data against fully-settled data, which manufactures a trend that does not exist.

The fix is presentational as much as analytical: mark provisional periods explicitly, exclude them from trend comparisons, and set an internal rule for how old data must be before it is treated as final.

Practical handling

• Label it. Provisional and settled figures should never appear in the same column unmarked.
• Set a settlement age — a number of days after which a period is treated as final — and apply it consistently.
• Exclude unsettled periods from optimisation decisions wherever the decision can wait.
• Track revision size. If provisional figures routinely move a long way, that is itself information about how much weight they deserve.