What is an attribution window?
An attribution window is the period of time after someone clicks or sees an ad in which a later conversion will still be credited to that ad. If a campaign runs a 30-day click window, a purchase made 29 days after the click is counted; a purchase made 31 days after it is not. The window is a cut-off, and everything past it is invisible to the report regardless of whether the marketing influenced it.
Windows are usually expressed as a pair: a click-through window and a view-through window. Click windows are longer because a click is a stronger signal of intent than an impression. According to Google Ads Help, the click-through conversion window for Search and Display campaigns can be set anywhere from 1 to 90 days and defaults to 30 days, while the view-through window can be set from 1 to 30 days and defaults to 1 day. Other platforms use their own ranges and defaults, and those defaults change from time to time, so the setting is worth checking rather than assuming.
Why the window changes the numbers
Two things happen when a window is widened. More conversions get counted, so reported volume and ROAS go up. At the same time, more of what gets counted would have happened anyway, so the reported figure drifts further from the incremental one. Narrowing the window does the reverse: the numbers get smaller and, usually, more conservative.
This is why comparing channels on their own reported numbers is unreliable. A platform running a 30-day click and 1-day view window will report a very different contribution from one running a 7-day click window on the same underlying behaviour, and both will happily claim the same order. Reconciling those claims to one revenue number is the practical reason teams run independent attribution alongside the platform reports, a point covered in why attribution matters.
Choosing a window that matches your buying cycle
The right window is a description of your customers, not a preference. The test is simple: how long does it actually take someone to buy after their first serious interaction?
- Impulse and low-consideration purchases (under $100, decided in a sitting) are well served by short windows of 1 to 7 days. A longer window mostly adds coincidence.
- Considered consumer purchases (furniture, travel, higher-priced apparel) typically need 14 to 30 days, because people compare, abandon a cart, and come back.
- B2B and high-ticket sales routinely run past 90 days, which is longer than most ad platforms will measure. This is the case where a CRM holding the real close date matters more than the ad platform’s own window, since the deal closes long after the platform has stopped looking.
Most analytics tools can show the actual time-to-conversion distribution for your business. Setting the window to cover roughly the 90th percentile of that distribution is a defensible rule of thumb: long enough to capture almost all genuine influence, short enough that you are not crediting a click from a different quarter.
Windows and multi-touch models
The window and the model are separate decisions that interact. The window determines which touchpoints are eligible for credit; the attribution model determines how credit is split among the eligible ones. A 7-day window with a linear model can only ever share credit among touchpoints from the last week, so it will look far more last-click-like than the model name suggests.
Longer windows make multi-touch models meaningful, because there are more touchpoints in scope to distribute credit across. Very short windows quietly collapse most models toward last-click whatever is selected in the interface.
Practical rules for handling windows
- Change one thing at a time. Widening the window mid-quarter and then celebrating a performance improvement is a self-inflicted measurement error. Note the change date and compare like with like on either side of it.
- Record the setting alongside the number. A ROAS figure without its window is not comparable to anything, including your own results from last year.
- Keep the window stable for trend reporting. Trends are only readable when the measurement rule holds still.
- Do not tune the window to flatter a channel. Widening a window until retargeting looks good is fitting the ruler to the result.
An attribution window is a modelling assumption wearing the costume of a technical setting. Choosing one deliberately, documenting it, and leaving it alone is most of what good attribution hygiene requires.