1. Why two reports disagree
The most common conversion question I get is why GA4 and Google Ads show different counts for the same action. Part of the answer has always been that they count differently. Until recently GA4 applied one attribution setting to every conversion in a property, and Google's 2026 changes allow finer control. Two things in Google's "What's new" notes matter here: per-conversion attribution settings, announced on January 16, 2026, and custom lookback windows, which Google lists as generally available from August 11, 2026.
2. Per-conversion settings are a beta
Google describes per-conversion attribution as a beta that is not available to every property, with eligibility confirmed through support. If you do have it, you can set attribution for each conversion on its own instead of one rule for all, which is useful when a purchase and a newsletter signup should not be credited the same way. If you do not have it yet, plan for the global setting and do not write process around a feature you cannot open. The same notes list a Conversion Attribution Analysis report, also in beta, with an assisted-conversions view.
3. Lookback windows change the denominator of the story
The lookback window decides how far back GA4 will look for a touchpoint to credit. Google's notes say click-through conversions can now use a custom window from 1 to 90 days, and engaged-view conversions from 1 to 30 days, set under Advertising, then Conversion management, then Settings. A longer window credits more conversions to earlier touchpoints; a shorter one credits fewer. Neither is wrong, but the two give different numbers for the same behavior, which is a hidden cause of the discrepancies the event quality scorecards article keeps asking you to explain before blaming tracking.
4. Write the setting into the metric definition
A conversion count is only comparable between two periods if the settings were the same in both. The first time I got this wrong, a month-over-month "improvement" in paid conversions turned out to be a lookback change made mid-month that nobody had recorded. My habit now is to add the attribution model and window to the metric definition in the tracking plan and to log any change with its date, the same discipline as a north star metric mapping. When you compare periods across a change, say so on the slide.
5. Reconcile before you escalate
When GA4 and Google Ads disagree, work down a short list: the attribution model on each side, the lookback windows, the conversion definition (is it the same event?), and the date and time zone alignment. Only then suspect tagging, using the measurement QA checklist. My take: most gaps are definitional, not broken, and the fix is a written definition rather than a code change. A limitation worth stating: parts of this are beta and availability varies by property, so check what your own account shows rather than trusting this article's description of the options.
How different teams plug in
Attribution settings are easy to change and hard to notice, so ownership should be explicit:
- Marketing owns the attribution choices for paid campaigns and tells analytics when they change.
- Analytics records the model and window with each conversion definition and annotates changes.
- Finance uses the documented definition when reconciling reported conversions against revenue.
- Leadership asks which settings a number was computed under before comparing two periods.