GA4 Attribution Models Explained: The Setting Most Small Sites Get Wrong

A site owner checks their GA4 conversion paths report and finds that a channel they barely think about — say, organic social — is getting credit for sales they were sure came entirely from search. Nothing about their tracking setup changed. What changed is which attribution model GA4 is using behind the scenes to decide which touchpoint in a customer’s journey gets the credit, and for most small WordPress sites, that setting isn’t what they think it is.

This isn’t a reporting bug. It’s a specific, well-documented mechanic in GA4 that most explanations skip over — including the part where GA4 can quietly downgrade your attribution model without ever telling you.

What Attribution Actually Means

Most conversions aren’t the result of one visit. Someone finds a post through a Google search, leaves, comes back a week later after clicking a link in an email, then converts three days after that from a direct visit typed straight into the address bar. Three touchpoints, one conversion. Attribution is the set of rules GA4 uses to decide how much credit each of those three touchpoints gets.

Get the model wrong for your situation and the distortion compounds: a channel that’s actually driving early-stage discovery looks worthless in the reports, while whichever channel happens to close the deal last gets credit it didn’t fully earn. That’s the whole reason attribution modeling exists as its own topic rather than just a footnote in reading GA4 reports generally.

The Two Models GA4 Actually Offers

Older guides still describe five or six attribution models — first-click, linear, time-decay, position-based, and so on — because that’s what Universal Analytics offered. Google removed first-click, linear, time-decay, and position-based from GA4 in November 2023. They’re gone, not just deprecated in some settings menu; they no longer exist as selectable options.

What’s left is two real choices:

Data-driven attribution uses GA4’s own machine-learning model to distribute credit across every touchpoint in a conversion path, weighted by how much each one actually appears to influence the outcome based on your site’s own historical data. This is GA4’s default and, when it’s genuinely running, the most accurate option available.

Last-click attribution gives 100% of the credit to the final touchpoint before conversion, full stop, ignoring everything that happened earlier in the journey. It’s simple to understand and it’s what most people picture when they think about attribution, but it systematically undervalues top-of-funnel channels like content and social that rarely close a sale on the same visit they generate.

The Setting Most Small Sites Get Wrong

Here’s the part that rarely makes it into a beginner explanation. Data-driven attribution isn’t available to every property just because it’s selected as the default model. GA4 requires a property to have at least 400 conversions for the specific key event and 20,000 total conversions across all key events within the current lookback window before it will actually run the data-driven model.

Miss either threshold and GA4 falls back to last-click attribution automatically — and the interface does not tell you this happened. The reporting model still shows as “data-driven” in your settings, because that’s what you selected. What’s actually being calculated behind that label, for a site that hasn’t crossed both thresholds, is last-click. A new blog, a freelancer’s portfolio site, or any WordPress site that hasn’t yet built up real conversion tracking volume is very likely sitting in this exact situation without knowing it.

This matters most for anyone drawing conclusions about which marketing channel is “working.” A site that thinks it’s getting a nuanced, multi-touch view of its traffic sources may actually be looking at last-click numbers with a more sophisticated-sounding label on them — and making channel-investment decisions based on a model that isn’t the one they think they configured.

How to Check and Change Your Attribution Model

Google relocated this setting in the third quarter of 2025, so guides written before then point to the wrong menu. As of now, it lives at Admin > Data display > Events > Attribution settings. You’ll need Marketer-level access or higher on the property to change anything here — Viewer and Analyst roles can see the setting but can’t edit it.

Inside Attribution settings, two things are actually adjustable: the reporting attribution model itself (data-driven or last-click), and the lookback window — how far back GA4 looks when deciding which earlier touchpoints get credit for a conversion. Most key events default to a 90-day lookback window, adjustable down to 30 or 60 days; acquisition events like first_visit use a separate, shorter window defaulting to 30 days, adjustable to 7. A shorter window is more conservative about crediting old touchpoints, which matters more for sites with long consideration periods (anything involving a quote, a consultation, or a considered purchase) than it does for an impulse-buy storefront.

None of this replaces properly configured event tracking or clean UTM tagging on campaign links — attribution modeling only distributes credit sensibly across data that’s actually being captured correctly in the first place. A model can’t fix a channel GA4 never recorded a visit from.

Does This Actually Change What You Should Do

For a site that’s already well past both thresholds, this is mostly good news: data-driven attribution is genuinely running, and it’s the more accurate default without needing to touch a setting. The practical move there is just confirming the lookback window matches how long your actual sales cycle runs.

For a newer or lower-traffic site, the honest takeaway is different: treat channel-level conversion credit in GA4 with real skepticism until conversion volume is high enough to trust it, and don’t restructure a marketing budget around attribution numbers from a property that’s quietly running on last-click. Check the actual thresholds against your own property’s conversion counts before trusting the “data-driven” label at face value — it’s a five-minute check that changes how much weight those numbers deserve.