GA4 Attribution Reporting: A Practical Guide

At some point, GA4 quietly switched everyone over to data-driven attribution as the default reporting model. Most people never chose that — and most people also don't know about the one requirement that determines whether it's even working the way it's supposed to.

GA4's default model, and what it actually does

By default, GA4 uses data-driven attribution for its reporting — a model that uses machine learning to distribute fractional conversion credit across every touchpoint that contributed, weighted by how much each one actually seemed to matter, based on your own account's conversion patterns rather than a fixed rule. It considers up to 20 touchpoints across channels in the 90 days before a conversion.

This matters a lot for smaller accounts. If your GA4 property is under that conversion threshold, the "data-driven" label in your reports may not mean what you think it means for at least some of your conversion events.

Where GA4's attribution reporting actually lives

• Advertising > Attribution > Conversion paths — the sequence of channels that led to conversions, closest to a raw customer journey view.
• Advertising > Attribution > Model comparison — lets you see how conversion credit shifts across different models (data-driven, last-click, first-click) for the same underlying data.
• Multi-channel funnel-style reporting — covered in more depth in multi-channel funnel reports.

Reading the model comparison report correctly

The most common mistake here is picking whichever model makes a favorite channel look best and reporting that one. The report is more useful read the other way: look at where a channel's credit changes the most between models. A channel whose credit drops sharply under last-click but stays high under data-driven is very likely playing an assist role — it shows up early and often, but rarely closes.

Example reading

If Paid Social's share of conversions is 24% under data-driven but only 9% under last-click, that gap is the story — it's a channel doing real work earlier in journeys that a last-click-only view would make you underfund.

What GA4 attribution still can't tell you

• It only sees what it was tagged to see — untagged campaigns and broken UTMs show up as direct/(none) traffic, invisible to any attribution model.
• It's bounded to a single property's tracked data — offline conversions, phone sales, and CRM outcomes need to be explicitly imported to be part of the picture (see offline conversion tracking).
• Cross-device journeys are only stitched together as well as GA4's own identity resolution allows — see cross-device tracking.

For the full comparison of what data-driven attribution is actually doing relative to simpler models, see marketing attribution models compared.