GA4 attribution models explained: what to tell your boss or client

GA4 offers three attribution models today: data-driven (the default), paid and organic last click, and Google paid channels last click. First click, linear, time decay and position-based were removed in November 2023. Most attribution confusion in GA4 comes from something else, though: the model you pick only applies to some reports, so the same sale can sit under two different channels depending on where you look.

This guide covers what each model does, where the settings live, and how to explain the result in two sentences to a boss, a partner or a client who asks “so which channel actually worked?”

Which attribution models does GA4 offer in 2026?

Google lists three models in Get started with attribution:

Model	How it gives credit	When it makes sense
Data-driven (default)	Spreads credit across the clicks on the path, based on your own property’s data	The default for most sites. Fair to channels that start journeys
Paid and organic last click	100% of the credit to the last channel clicked before the key event, ignoring direct	When you need to match a tool or report that counts last click
Google paid channels last click	100% to the last Google Ads click; falls back to paid and organic last click if there was none	Only if Google Ads is the one thing you are judged on

One rule applies to all three: direct visits never get credit unless the whole path was direct. Google notes that paid and organic last click is also called “last non-direct click”.

How data-driven works, in plain words. GA4 looks at paths that converted and paths that did not, and estimates what each touchpoint added. Google’s own illustration: a path of four ad interactions (paid search, social, affiliate, search) gives a 3% chance of a key event, the same path without the fourth gives 2%, so the fourth interaction gets credit for that lift (Get started with attribution). The model also weighs time to the key event, device, the number and order of ad interactions, and creative type. One practical consequence: conversions can be reattributed for up to 7 days after they happen, so last week’s channel split can still move a little when you reopen it.

If you read an older guide that still lists first click, linear, time decay or position-based as GA4 models, it predates the 2023 change. We also covered which model to pick for ROI figures in marketing ROI: how to calculate it.

Why the same sale shows up under different channels

This is the question you will actually get, usually as “your report says organic, the Ads report says paid, which is it?”

GA4 has three families of traffic dimensions, explained in Scopes of traffic-source dimensions:

• User-scoped (First user source, First user medium): where the user came from the first time. Used in the User acquisition report.
• Session-scoped (Session source, Session default channel group): where each session started. Used in the Traffic acquisition report.
• Event-scoped (Source, Medium, Default channel group, without a prefix): who gets credit for a key event. Used in the Advertising reports and anywhere you break key events or revenue down by these dimensions.

The attribution model you choose only touches the third family. For user- and session-scoped dimensions, “Analytics uses the paid and organic channels last click attribution model”, always. Change the reporting attribution model confirms it: Session source and First user medium are unaffected.

A hypothetical path makes it concrete. A buyer clicks an Instagram ad on Monday, comes back from a Google search on Thursday, and types your URL to buy on Friday.

• Traffic acquisition (session-scoped, last non-direct click): the purchase goes to Google organic. The direct Friday visit is ignored because an earlier source exists.
• Advertising reports with data-driven (event-scoped): the purchase can be split between Instagram paid social and Google organic, in proportions GA4 estimates from your data.
• User acquisition: the buyer counts as a user acquired by Instagram, because that was their first visit.

None of these is wrong. They answer three different questions: what started the relationship, what brought this visit, and what deserves credit for the sale.

How lookback windows change the count

The lookback window is how far back GA4 looks for a click that can take credit. The defaults, from Select attribution settings:

Key event type	Default	Other options
Acquisition key events (first_visit, first_open)	30 days	7 days
All other key events	90 days	30 or 60 days

A shorter window removes older touchpoints from the path, so credit tends to shift toward the channels people click last. Keep 90 days unless your sales cycle is clearly shorter. Changes to the window apply going forward only, so a mid-quarter change makes the before and after periods inconsistent. Unlike the model, the lookback window also applies to session attribution, so it can shift Traffic acquisition numbers too.

How to change the reporting attribution model

1. Open Admin. Under Data display, click Events, then Attribution settings (labelled Key events attribution in some properties).
2. Pick the reporting attribution model.
3. Check the channels that can receive credit. Paid and organic channels is the default for GA4 reports and applies to web conversions only (app conversions always use Google paid channels). Changing it applies going forward and to linked Google Ads accounts, and can affect bidding there.
4. Set the key event lookback windows, then click Save.

You need the Marketer role or above on the property (Select attribution settings). The important warning: “Changing the reporting attribution model applies to historical and future data.” Switch models in March and your January numbers change too. If you already shared January’s numbers, the figures in GA4 no longer match them.

Compare models before you argue about them

GA4 has two reports for this, both under Advertising in the left menu.

Attribution models (Key event attribution models report) puts two models side by side, channel by channel, with key events, revenue and the % change between them (Analytics Help). This is the quickest way to answer “is paid social undervalued by last click?” with a number instead of a debate. If the change between models is a few percent, the model choice does not matter for that account and you can stop discussing it.

Attribution paths (Key events attribution paths report) shows the journeys of users who completed a key event, up to 20 touchpoints long, split into early (first 25% of touchpoints), mid (middle 50%) and late (last 25%), with days and touchpoints to key event in the table (Analytics Help). Use it to show a client that their awareness channel shows up early in paths, even when it rarely closes the sale.

Both reports include data from 14 June 2021 onwards, and they only show paths for key events, so they stay empty until the events that matter are marked as key events.

How to explain attribution to a boss, partner or client

What works in practice is less about the model and more about agreeing on rules once:

1. Name the model in the report. One line in the footer: “Revenue by channel uses GA4 data-driven attribution, 90-day lookback.”
2. Use one report per question. Channel trends and traffic come from Traffic acquisition (session-scoped). Credit for sales comes from the Advertising reports (event-scoped). Do not mix the two in one table.
3. Never switch models mid-year without telling the reader first, because history changes with it.
4. Show the range when it matters. If a channel’s value swings a lot between data-driven and last click, show both numbers in an appendix instead of picking the flattering one.
5. Expect GA4 and ad platforms to disagree. Meta, Google Ads and your store each count with their own rules. For store revenue versus GA4, see Shopify Analytics vs GA4.

Which model should you use?

• Most small sites and stores: keep data-driven. It is the default, it gives some credit to channels that start journeys, and you do not have to defend a change.
• You reconcile against a last-click tool: use paid and organic last click so both sides count the same way.
• Google Ads is the only budget you manage: Google paid channels last click, knowing it hides the role of organic and social.

Whatever you choose, attribution answers “who gets credit”. It does not explain why a channel moved this week. When a number drops, start with the GA4 traffic drop checklist, then the step-by-step traffic analysis.

Where Dialytics fits

Dialytics is our product, so read this with that in mind. The Monday digest it posts in Slack reads channels with GA4’s session-scoped default channel group, the same view as the Traffic acquisition report, so its channel numbers do not move when someone changes the attribution model in Admin. Each week it says what moved and the one point worth your attention, and a daily check posts an alert only when a metric is unusual for that weekday. When your boss asks “why is paid search down?”, anyone in the channel can @mention the bot in the thread and get an answer that remembers previous weeks.