GA4 Attribution Settings Most Teams Get Wrong

The GA4 attribution settings most teams get wrong are the acquisition lookback window left at 30 days, key event counting stuck on "once per event," the reporting identity left on a default that changes how users are counted, and channel groupings that dump paid traffic into unassigned or organic buckets. Each setting silently reshapes the numbers you report, and because GA4 applies most changes going forward, teams compare distorted data without realizing the ground shifted underneath them.

Attribution model: you have less choice than you think

Since Google removed the first-click, linear, time-decay, and position-based options, GA4 credits paid and organic channels with data-driven attribution by default, leaving last-click variants as the only alternatives. Many teams never noticed the switch, which alone explains sudden shifts in how much credit each channel appears to earn.

The bigger trap is inconsistency between reports. Traffic acquisition reports use session-scoped, last-click-style logic, while the advertising and attribution reports use your event-scoped model. Comparing the two side by side and calling the difference an insight is one of the most common GA4 mistakes, and it produces confident conclusions built on mismatched math.

Lookback windows: the silent data deleter

GA4 lets you set the acquisition lookback window to 7 or 30 days and the key event lookback to 30, 60, or 90 days. Teams selling considered purchases who leave the key event window short will structurally erase early-funnel touchpoints from credit, then wonder why top-of-funnel channels look weak.

Two facts matter here. First, changes apply going forward only, so your history stays on the old setting and any trend comparison across the change breaks quietly. Second, if your sales cycle runs longer than 90 days, no GA4 setting fixes it. The window simply cannot stretch to match a journey that outlives it, and that is a hard ceiling, not a configuration you missed.

Key event counting: once per event vs once per session

Each key event counts either every occurrence or once per session. Leave a lead form on "once per event" and one enthusiastic visitor who submits three times becomes three conversions. Leave a purchase event on "once per session" and a customer who buys twice in a single session collapses into one.

The rule of thumb: transactional events like purchases should count once per event, while lead and signup events usually belong on once per session. Audit this per event, because the default is once per event for everything, and the default is wrong for a surprising share of the events you actually care about.

Reporting identity and data thresholding

Reporting identity controls how GA4 stitches users together: blended, observed, or device-based. Blended leans on modeled data and Google signals, which can trigger data thresholding that hides rows in reports with low user counts. Teams watch data disappear and blame their tracking setup, when the culprit is often the identity setting interacting with privacy thresholds.

Device-based gives the most literal counts; blended gives the most complete modeled picture. Neither is objectively correct. Pick one deliberately, document why, and stop toggling between them mid-quarter, because every switch quietly rewrites the user counts your dashboards depend on.

Channel groupings and the unassigned problem

GA4 assigns traffic to default channels using source, medium, and campaign rules. Untagged or mistagged links land in unassigned or direct, and paid social without proper UTMs often reports as organic social, flattering your organic team while starving your paid team of the credit it earned.

The fix is unglamorous but durable

Write a UTM convention document and enforce it on every campaign link, then build custom channel groups wherever the defaults misclassify your mix. Remember that custom channel groups apply retroactively in reports, but the underlying source and medium data cannot be repaired after the fact, so tagging discipline at the moment of the click is the only real prevention.

The settings worth checking this week

1. Your attribution model, and whether every report you compare uses the same one.
2. The key event lookback window versus your real purchase cycle.
3. The counting method for every key event, not just the important ones.
4. Reporting identity, chosen deliberately and left alone.
5. UTM discipline and the share of traffic sitting in unassigned.
6. Google Ads linking, so click IDs flow and platform discrepancies shrink.

None of these are advanced. They are the quiet defaults that decide whether your reporting reflects reality or a series of accidental choices no one signed off on.

Where a measurement layer helps

GA4 remains a strong free baseline, but it was never designed to be the final word on budget decisions. It cannot see impression-driven channels well, it caps out at a 90-day window, and it has no account-level view. Admira sits above tools like GA4, combining multi-touch attribution, marketing mix modeling, and incrementality testing with cookieless tracking, and pulling every channel together through prebuilt integrations across your ad platforms, GA4, and CRM. If GA4 settings archaeology is eating your week, book a demo and see what a unified measurement layer reports that GA4 structurally cannot.

FAQ

Why does GA4 show fewer conversions than Google Ads?

Google Ads counts click and view conversions with its own attribution logic and windows, while GA4 deduplicates credit across every channel. Neither is wrong; they answer different questions. Platform-reported conversions summed across channels will almost always exceed what a single deduplicating system like GA4 shows, so treat the gap as expected rather than a bug.

Should I use data-driven attribution or last click in GA4?

Data-driven attribution is the better default because it spreads credit using observed conversion patterns instead of a fixed rule. Keep a last-click comparison view if your team needs continuity with older reports, but never average the two models into one narrative. Pick one as your source of truth and annotate the date you switched.

Can GA4 attribution handle long B2B sales cycles?

Only partially. The maximum 90-day key event lookback cannot cover journeys that run six months or more, and GA4 has no account-level view of a buying committee. B2B teams usually pair GA4 with CRM-connected attribution or a dedicated measurement platform that extends windows and rolls touchpoints up to the account.

Do these GA4 settings change my historical data?

Mostly no. Lookback window and conversion-counting changes apply from the change date forward, so your history stays on the old setting and trend comparisons across the change quietly break. Annotate the date you changed anything. Custom channel groups are the exception: they apply retroactively in reporting, though the underlying source data cannot be repaired after the fact.