Attribution

What attribution means

Attribution is the set of rules that decides which marketing touchpoint gets credit when someone converts. It matters because the model you pick silently reallocates budget: change the rules and the same conversions suddenly "belong" to different channels, and whoever reads the report draws different conclusions about what's working.

A typical journey: someone finds your article via organic search on Monday, clicks a retargeting ad Wednesday, and converts Friday from a newsletter link. One conversion, three touchpoints. Attribution decides whether organic, paid, or email gets the win — or a fraction each.

Where GA4 stands today

GA4's default is data-driven attribution (DDA): a machine-learning model that distributes credit across touchpoints based on observed conversion patterns in your own property. Google retired the old rules-based multi-touch models — first-click, linear, time-decay, and position-based were removed from GA4 in 2023. What remains selectable is data-driven, paid-and-organic last click, and Google-paid-channels last click.

Be honest about DDA's limits. It's a black box — you cannot inspect why it credited a channel. It needs conversion volume to model against, so on low-traffic sites it behaves close to last-click anyway. And it only sees what GA4 sees: cookie-consent rejections, Safari's ITP cookie caps, and cross-device journeys all leave holes the model papers over with modeled estimates. Treat attribution reports as directional evidence, not a court ruling.

Attribution models compared

Why SEOs should care

Organic search is structurally the channel most likely to open a journey and least likely to close it. Under pure last-click, organic looks like it "assists" while email and branded paid search harvest the conversions organic created. That's how SEO budgets get cut by people reading the wrong report. When you present organic performance, show assisted conversions and the model-comparison view, not just last-click totals — the difference is often the whole argument. We cover the organic-specific angle in depth in Attribution Modeling for Organic Search and the GA4 wiring in Connecting GA4 and Search Console.

Scope: the detail that makes GA4 reports "disagree" with themselves

GA4 exposes the same journey through three different dimension scopes, and mixing them is the top source of "these two reports don't match" tickets. First user source/medium is user-scoped: the channel that acquired the person originally, frozen forever. Session source/medium is session-scoped: whatever started this visit. Source/medium (unprefixed) is event-scoped and attribution-modeled: for key events it reflects DDA's fractional credit. So a purchase can sit under "organic" in one report, "email" in another, and split fractionally in a third — all correct, all answering different questions. Decide which question you're asking before you build the report, and label the scope in the header so the next reader doesn't relitigate it.

How to check your attribution setup

1. In GA4, go to Admin → Attribution settings and note the reporting attribution model and lookback windows (30/60/90 days for acquisition events).
2. Open Advertising → Attribution → Model comparison and put data-driven next to last click. Big swings on a channel mean the model choice is doing heavy lifting.
3. Check the share of "direct" and "unassigned" traffic. High values mean tagging gaps — attribution can only distribute what's labeled. Start with your UTM parameters.
4. Verify key conversions are marked as such (Admin → Events), because attribution reports only run against designated key events.
5. Compare GA4 conversion counts against your backend (orders, signups). A persistent gap tells you how much of reality your attribution never sees.

Common mistakes and fixes

• Treating one model as the truth. Fix: always look at model comparison; use the deltas between models as the insight.
• Judging organic on last-click only. Fix: report assisted conversions and first-touch involvement alongside.
• Ignoring the "direct" bucket. Untagged campaigns, stripped referrers, and app traffic all hide there. Fix: tighten tagging before blaming the model.
• Comparing attribution across tools. GA4, ad platforms, and your CRM each use different rules and windows — the numbers will never match. Fix: pick one system of record per decision.
• Forgetting consent-mode gaps. In consent-heavy markets a chunk of journeys is modeled, not measured. Fix: annotate reports so stakeholders know the error bars exist.

FAQ

Is last-click attribution dead?

No — it's still selectable in GA4 and still what most ad-platform dashboards effectively show. What died is the menu of rules-based multi-touch models. Last click survives because it's simple and auditable, even though it undervalues discovery channels.

Can I see why data-driven attribution credited a channel?

Not really. Google describes the methodology (a counterfactual model comparing converting vs non-converting paths) but exposes no per-conversion reasoning. You can only sanity-check it against model comparison and your own backend data.

What lookback window should I use?

GA4 defaults to 90 days for key events and 30 for acquisition. Match the window to your sales cycle — a 90-day window on a same-day impulse purchase inflates the appearance of multi-touch journeys.

Does attribution work across devices?

Only when the user is signed in to Google or your site (User-ID). Otherwise a phone research session and a desktop purchase look like two unrelated users, and the phone touchpoint gets zero credit.

How does attribution relate to a ranking drop investigation?

Differently — that's traffic attribution, figuring out why organic fell rather than which channel gets conversion credit. We keep a separate framework for that: Diagnosing a Ranking Drop.