First click, linear, time decay and position-based were withdrawn in November 2023. What remains in the GA4 attribution reports is data-driven, paid and organic last click, and Google paid channels last click — and one of the three is still the one arguments end on.
GA4 offers three attribution models in its attribution reports: data-driven attribution, paid and organic last click, and Google paid channels last click. First click, linear, time decay and position-based were withdrawn in November 2023 and are no longer available. All three surviving models exclude direct visits from receiving credit, unless the path to the key event consists entirely of direct visits, which is the rule that explains most of the results that look wrong at first reading.
| Model | What it credits | Path: Display, Social, Paid Search, Organic Search |
|---|---|---|
| Data-driven attribution | Credit distributed across touchpoints according to a model of their contribution | Split across the touchpoints in the path |
| Paid and organic last click | All of the value to the last channel clicked through, ignoring direct | All of it to Organic Search |
| Google paid channels last click | All of the value to the last Google Ads channel clicked through | All of it to Paid Search |
Paid and organic last click has a second name that causes confusion in reports and in meetings: last non-direct click. Google states plainly that the two are names for the same model, so a disagreement about which of them a report is using is a disagreement about vocabulary rather than about numbers.
Google paid channels last click gives all of the credit to the last Google Ads channel the customer clicked before converting, and it has a documented fallback: when the path contains no Google Ads click at all, the model falls back to paid and organic last click. A path of display, social, email and then a conversion therefore credits email under both models, which is why two reports can agree exactly on some conversions and diverge on others. The divergence is not noise; it is the presence or absence of a Google Ads click in the path.
The direct exclusion works the same way in both last-click models. A path that ends in a direct visit credits the previous clicked channel instead
A signup that starts on your marketing site and finishes in your app is two users in GA4 unless the domains are configured together. Here is what the setting does and what it cannot fix.
The Data API answers questions you can express in its dimensions. The BigQuery export hands over the events themselves, with the costs and caveats that come with owning them.
Market4 turns one release note into a changelog page, a blog post, a mail-out and a week of social posts — and then tells you which of them brought anyone back.
Last click keeps winning arguments because it is the model a person can recompute by hand. Given a path, the rule is one sentence long, and two people applying it to the same path get the same answer. Data-driven attribution is built from path data covering both converting and non-converting users, comparing the probability of a key event for users exposed to an ad against similar users in a holdback group. That is a better description of reality and a worse instrument for settling a disagreement, because nobody in the room can reproduce the number from the rows in front of them.
The practical resolution is to name the model on the report rather than to pick a favourite. A revenue figure without an attribution model attached is not comparable with anything, and two teams quoting different totals for the same campaign are usually quoting different models rather than different data.
A stable campaign taxonomy is the part a small team controls completely, and it is worth more than the choice of model. Every attribution model reads the same source and medium values, so an inconsistent tag produces a wrong answer under all three at once.
Three: data-driven attribution, paid and organic last click, and Google paid channels last click. Four earlier models — first click, linear, time decay and position-based — were withdrawn in November 2023 and cannot be selected. Reports and exports built before that change may still refer to the withdrawn models by name, which is worth checking before comparing a historical figure with a current one.
Nothing. Google states that paid and organic last click and last non-direct click are two names for the same attribution model. It ignores direct traffic and assigns all of the key event value to the last channel the customer clicked through, or engaged-view-through in the case of YouTube. If a path consists entirely of direct visits, direct traffic receives the credit.
Use one, name it on every report, and keep it fixed for comparisons. Data-driven attribution describes contribution across a path more faithfully; the last-click models are reproducible by hand, which makes them easier to defend in a discussion. Switching between them mid-analysis produces two numbers for the same period and no way to tell which change came from the traffic and which from the model.
All three models exclude direct visits from receiving attribution credit, unless the path to the key event consists entirely of direct visits. The reasoning is that a direct session usually follows an earlier discovery that was itself attributable, so crediting it would hide the channel that did the work. The visible effect is that a conversion whose last session was direct is credited to the previous clicked channel.
Realtime covers the last 30 minutes and never settles. Standard reports cover a date range in the property's time zone and keep changing until processing finishes. Neither is the corrected version of the other.