GA4 Attribution — How to Interpret Attribution Models

An attribution model determines how GA4 assigns conversion credit across the touchpoints in a user's journey. Since 2023, Google narrowed the choice down to just two models — the rest (linear, time decay, position-based, first click) were sunset. Which model you pick can shift your reported numbers by dozens of percent, so it's worth understanding the difference.

Data-Driven vs Cross-Channel Last Click — the Only Options Left

Data-driven (default) — a machine-learning model. GA4 analyzes all conversion paths (both successful and unsuccessful) and assigns credit to each touchpoint based on how much it actually increased the probability of conversion. It requires a minimum data volume (300+ conversions and 3,000+ ad clicks within 30 days per channel) — below that threshold, GA4 silently falls back to a different model.

Cross-channel last click — 100% of conversion value goes to the last click before conversion (ignoring direct traffic if an earlier touchpoint exists). Simple and predictable, but it systematically favors "closing" channels — brand search, remarketing — at the expense of channels that build awareness at the top of the funnel (social, display).

Where to Check and Compare

Advertising → Attribution → Model comparison. Pick two models side by side and see how conversions and revenue shift between channels. For fashion e-commerce, a classic pattern: Paid Social loses under last-click, gains under data-driven — because social is rarely the last click, but often the first touchpoint in the path.

Three Interpretation Pitfalls

1. Don't compare pre- and post-model-change data in standard reports. Changing the model in property settings retroactively recalculates conversion reports — if you're analyzing a trend, a mid-period model change introduces a jump that has nothing to do with an actual business change.

2. Data-driven ≠ objective truth. It's still a statistical model, not a measurement of reality — with small data volumes it can be unstable week to week.

3. Attribution in GA4 isn't the same as attribution in Google Ads. Google Ads has its own conversion model (often data-driven at the campaign level), which can differ from what GA4 property reports show — don't be surprised by discrepancies between the two panels.

When to Manually Verify

If revenue attributed to one channel jumps by more than 30% after a model change — check your data volume (are you above the data-driven thresholds), check for duplicate purchase events, and cross-check against an independent source (e.g., revenue from your order system) before making budget decisions.

TL;DR

• Since 2023, GA4 has only 2 models: Data-driven (default, ML-based) and Cross-channel last click
• Data-driven needs 300+ conversions / 3,000+ clicks per channel in 30 days — below that it silently switches models
• Last-click favors closing channels (brand search, remarketing) over top-of-funnel channels
• Don't compare trends over time across a model change — data is recalculated retroactively
• GA4 and Google Ads can show different conversion counts — that's different attribution models, not a tracking bug