Attribution Window

How the window actually determines credit

When someone clicks or views an ad, that interaction is logged with a timestamp. If a conversion happens later, the platform checks whether it falls inside the attribution window measured from that interaction. If it does, the conversion is credited to the ad and reported against the date of the original click or view, not the date the conversion actually happened - which is why a conversion can appear in a report for a week that has already closed.

Different conversion actions can carry different windows in the same account, and click-through and view-through interactions are frequently given different window lengths from each other, since a view is a weaker signal of intent than a click and platforms typically credit it for a shorter period.

What goes wrong with attribution windows in practice

The most disorienting failure is not a misconfiguration at all - it is simply not understanding that recent data is provisional. A business with a long sales cycle, where someone clicks an ad and books a job two or three weeks later, will always see its most recent two or three weeks under-reported relative to how they will eventually look once the window closes. Read too early, this looks like a sudden drop in performance that resolves itself with no action taken, and repeated overreaction to it erodes trust in the reporting generally.

The second is changing a conversion action's window and then comparing before-and-after periods as if performance changed. Widening a window from thirty to ninety days will typically increase reported conversions for past periods retroactively, making the change look like an improvement in the campaigns rather than a change in what is being counted.

The third, more subtle problem is a window mismatch between the ad platform and a business's actual sales cycle. A window set shorter than the real time-to-close for the product being sold will systematically undercount conversions and mislead Smart Bidding into thinking certain traffic performs worse than it actually does, since the bidding algorithm only ever sees the conversions that fall inside the window.

How attribution windows relate to the rest of the account

The window is set on the conversion action itself, so it inherits all the same audit considerations - it is easy to change, easy to forget was changed, and it reshapes reported history quietly when it is touched.

It is one of the two inputs, alongside cost, behind return on ad spend: a ROAS figure calculated on a window too short for the sales cycle will understate real performance, and widening the window is one of the most common reasons a ROAS number appears to improve with no change in actual sales.

It also directly affects cost per acquisition in the same way, since CPA is simply cost divided by the conversions the window allows to be counted - a longer window generally produces a lower, more complete CPA figure for a business with a longer path to close.

Why do my conversion numbers for last week keep going up days after the week ended?

This is normal and expected. Conversions are attributed back to the date of the original click or view, so a conversion that happens several days after someone clicked gets added to that earlier date once it occurs, as long as it falls inside the attribution window. Recent periods will always look incomplete until the window has fully closed.

What attribution window should a home services business use?

The window should roughly match how long it actually takes a typical customer to move from clicking an ad to booking or buying. A business with same-day calls can use a shorter window; one with a multi-week decision process, like larger installations, needs a longer window or it will systematically undercount conversions and mislead its own bidding.

Does changing the attribution window affect past reporting?

Yes. Changing the window on a conversion action typically causes the platform to reprocess attribution for past periods as well as future ones, which can make historical numbers shift even though nothing about the campaigns themselves changed. This is worth checking before assuming a reported improvement or decline reflects real performance.