GA4 Custom Conversion Windows: Match Attribution to the Buying Cycle
Updated: September 2026
GA4 custom conversion windows let properties use integer lookback periods instead of a few fixed presets. Choose windows from observed buying-cycle evidence, not the setting that produces the most conversions. A longer window assigns more historical touchpoints credit; it does not create more demand.
What changed in GA4
Google’s August 11, 2026 release added custom integer windows for attribution settings. Properties can configure engaged-view conversions from 1 to 30 days and click-through conversions from 1 to 90 days. The flexibility is useful for businesses whose consideration cycles do not fit standard buckets.
It also creates a new comparability problem. If teams change windows without annotation, historical dashboards, campaign evaluations, and channel contribution can appear to shift even when customer behavior did not.
Start with the buying cycle
Measure delay from first meaningful visit to conversion
Use user-level or aggregated journey analysis to understand the distribution, not just the average. Look at median, 75th percentile, and long-tail delays. Segment by conversion type, product line, new versus returning customer, and purchase value. A demo request and an ecommerce purchase may require different interpretation even when they share a property.
Distinguish clicks from engaged views
A click represents an explicit navigation action. An engaged view reflects meaningful video exposure followed by conversion without a click. They should not automatically receive identical windows. Set each according to the behavior and evidence you can defend.
| Business pattern | Diagnostic question | Window implication |
|---|---|---|
| Fast repeat purchase | Do most orders follow within a few days? | Shorter click window may reduce distant credit |
| Considered lead generation | How long from first visit to qualified opportunity? | Longer click window may capture research |
| Video-led discovery | Is engaged-view exposure associated with later quality? | Test EVC window separately |
| Seasonal promotion | Does the sale compress normal decision time? | Do not generalize the promotional window |
Use a controlled change process
- Export the baseline. Save channel, campaign, conversion, and model-comparison views before the change.
- Write the hypothesis. Example: “A 21-day click window better represents qualified lead consideration than 30 days.”
- Change one window family at a time. Avoid simultaneously changing attribution model, conversion definitions, and data imports.
- Annotate the effective date. Update dashboard notes and stakeholder documentation.
- Compare reporting and business outcomes. Examine how credit shifts across channels, but validate against CRM stages, revenue, and experiments.
Avoid three common mistakes
Do not optimize for more attributed conversions
Extending the window will often increase eligible historical interactions. That is a reporting choice, not evidence of incremental performance.
Do not force one window onto every decision
Executive planning may need a broader view of demand creation, while daily bid management needs a window aligned with actionable feedback and conversion lag. Document which view supports which decision.
Do not compare platforms without normalizing definitions
Meta, Google Ads, and GA4 can use different click, view, and engaged-view rules. Build a comparison table of window length, interaction type, identity, and attribution model before reconciling totals.
Pair window governance with our guide to GA4 aggregate identifier diagnostics. A theoretically correct lookback setting cannot compensate for missing identifiers or broken event collection.
Frequently asked questions
What is the best GA4 conversion window?
There is no universal best. Use the shortest window that represents the decision cycle for the conversion and interaction type, then validate it against business outcomes.
Will changing the window change historical results?
Attribution-setting behavior and reporting effects can depend on the report and processing rules. Preserve a dated baseline and review Google’s current property documentation before making trend claims.
Should engaged-view and click-through windows match?
Not automatically. They describe different interactions and can have different evidence thresholds, so configure and evaluate them separately.