GA4 Custom Attribution Windows: Pick the Right Days

As of August 11, 2026, GA4 attribution windows are no longer a short menu of presets — they are a number you choose. Google made both conversion lookback windows fully configurable: the click-through conversion window now accepts any whole number from 1 to 90 days, replacing the old six-value set of 1, 7, 14, 30, 60 and 90, and the engaged-view conversion window — previously fixed at 3 days for every property — can now be set to any integer from 1 to 30. Google's release notes frame the change as letting analytics "align conversion attribution lookback windows more closely to your unique business cycles."

For SaaS teams this is a small setting with an outsized effect, because the window silently decides which marketing touches get credit for a conversion and which are discarded. A window set too short amputates the long tail of a considered B2B purchase; one set too long credits clicks that had nothing to do with the sale. This post covers what actually changed, how to pick the exact click-through and engaged-view integers by measuring your own cycle, why the change is a reporting event rather than a silent tweak, and how to keep GA4 and Google Ads reconciled once you set a custom number.

What actually changed on August 11, 2026

The change has two parts and it is worth separating them. First, the click-through conversion (CTC) window — the lookback that credits a conversion to a prior ad click — moved from six fixed presets to any integer between 1 and 90 days. Second, the engaged-view conversion (EVC) window, which credits a qualifying video view that did not end in a click, moved from a hard-coded 3-day value to any integer between 1 and 30 days. Both are found under Advertising, then Conversion management, then the settings for a given conversion, and the same controls surface in the linked Google Ads conversion management interface.

The practical upshot is that the rounding you used to be forced into is gone. A SaaS product with a measured 21-day median trial no longer has to choose between a 14-day window that clips the tail and a 30-day window that overshoots — it can set 21, or 24, or whatever the data supports. Search Engine Land notes the update also makes the previously immovable 3-day engaged-view window editable, which matters for anyone running YouTube or demand-gen video whose assists were being cut off after 72 hours. This is a genuine new lever; the rest of the work is deciding where to set it, which Google pointedly does not tell you. If you want the conceptual grounding first, our guide to lookback versus conversion windows explains what each window governs before you start tuning the numbers.

Measure your cycle before you pick a number

The right window is a property of your customers, not a best practice you can copy. Pull the distribution of elapsed days between first ad click and conversion for people who actually converted, and look at where the mass sits. Most SaaS conversion-lag distributions are heavily right-skewed: a large cluster converts quickly and a long thin tail drags out for weeks. The average is misleading because the tail pulls it upward; the percentiles are what you want. Set the click-through window near the 80th to 90th percentile of that lag so you capture the realistic long tail while excluding clicks so old they are almost certainly coincidental to the purchase.

Concretely, if 85% of your conversions happen within 34 days of the first click, a 35-day window captures nearly all real influence without paying credit to month-old clicks. Before August 2026 you would have rounded that to 30 and lost the tail, or to 60 and inflated older channels; now you set 35. The one trap is letting the window quietly exceed the cycle it is meant to measure — if your sales cycle runs longer than your window, conversions get orphaned from the touches that drove them, a failure mode we cover in when your sales cycle outruns the attribution window. Measure first, then set the integer to fit.

Tuning the engaged-view window without inflating video

The engaged-view window deserves separate, more conservative treatment than the click-through window. View-based credit is inherently softer evidence than a click — the user did not act, they merely watched — so widening the EVC window mechanically increases the volume of conversions credited to video without any change in actual performance. That is not automatically wrong; the old 3-day cap genuinely discarded real YouTube assists for products with longer consideration. But moving it from 3 to 30 will make video look dramatically more effective overnight, and most of that lift is a settings artifact, not incremental demand.

The disciplined approach is to widen the EVC window in steps and validate against something outside GA4's own crediting — a holdout test, a geo experiment, or a media-mix read — before you trust the larger number for budget decisions. If a 7-day EVC window doubles credited video conversions but a proper incrementality test shows the same true lift as before, the extra credit is double-counting that will distort your channel comparison. Treat view-through credit as a hypothesis to confirm, not a result to bank, and keep it reconciled with the harder click-based evidence in your multi-touch attribution model.

Changing the window is a reporting event, not a silent tweak

The moment you change an attribution window, your conversion counts and per- channel credit can shift — including for historical dates in reprocessed reporting — because the same underlying clicks are now attributed under different rules. This means a window change produces a step change in your dashboards that has nothing to do with real marketing performance. If nobody records that the setting changed, the next person to compare this month against last month will read a definitional shift as a trend, and chase a "drop" or celebrate a "lift" that is purely an artifact of the new number.

Protect against this the way you would any definition change: annotate the exact date the window changed directly in GA4, note the old and new values, and warn anyone doing period-over-period analysis that the two sides of that date are not comparable on a like-for-like basis. This is the same hazard as metric-definition drift: when the meaning of a number moves, the number moves with it, and undocumented changes turn your own history into a source of false signals. A single annotation saves an afternoon of investigating a ghost.

Keep GA4 and Google Ads reconciled

A frequent reason GA4 and Google Ads disagree on conversion counts is that they are using different attribution windows for the same campaigns. If GA4 looks back 30 days and Google Ads looks back 90, each system credits a different set of conversions, and the two reports will never tie out no matter how carefully you map channels. Before August 2026 exact alignment was often impossible because the preset menus differed; now that GA4 takes any integer, you can set both platforms to the same window and remove that entire class of discrepancy.

Alignment will not make the numbers identical — consent-driven modeling, cross-device stitching, and each platform's own attribution logic still create gaps — but it removes the window mismatch as a suspect, which is usually the largest and most fixable one. Set both to the same integer, document it as the canonical window, and treat any residual difference as a modeling or consent question rather than a settings bug. When two tools report different numbers, knowing which differences are expected is half the battle; our guide on why analytics tools show different numbers walks through the rest.

A practical setup checklist

Turn the change into a short, repeatable procedure rather than a one-off guess. First, export the click-to-conversion lag distribution for your paying customers and read the 80th–90th percentile — that percentile, rounded to a whole day, is your candidate click-through window. Second, decide the engaged-view window separately and conservatively, starting near your typical consideration length and only widening it once an incrementality check justifies the extra video credit. Third, set the same click-through integer in Google Ads so the platforms reconcile. Fourth, annotate the change date everywhere your team reads conversion data.

Then revisit the setting on a schedule, because the right window is not static: as your product adds a sales-assisted tier, lengthens a trial, or shifts channel mix, the conversion-lag distribution moves and the window should follow. The new integer control makes that ongoing calibration cheap — you are adjusting a number to fit measured behavior, not choosing from a menu that never quite fit. Anchor the whole exercise to a window you can defend from your own data, keep it aligned across platforms, and document every change, and GA4's new flexibility becomes a sharper measurement instead of one more setting to misconfigure. For teams rethinking how conversion windows interact with funnel definitions, our piece on open versus closed conversion windows pairs naturally with this decision.