GA4 Attribution Models: What Is Left and How to Choose

GA4 attribution models now come down to three choices: data-driven, paid and organic last click, and Google paid channels last click. First click, linear, time decay and position-based are gone, so the real decision is whether you trust Google's machine-learning model or a last-click rule, and which channels you allow to earn credit.

The choice matters because the reporting model decides which source, medium and campaign gets credit for each key event in your reports. Pick the wrong one and a channel that starts journeys can look worthless, or a branded search campaign can look like it does all the work. This guide covers what each remaining model does, what was removed, the reports that let you compare models, how lookback windows work, and why GA4 and Google Ads rarely show the same conversion count.

Which attribution models GA4 still offers

GA4 lists three reporting attribution models. Two of them apply when you let both paid and organic channels receive credit, and one applies when you limit credit to Google paid channels.

Data-driven attribution

Data-driven attribution is the default and the option Google labels as recommended. It does not follow a fixed rule. Instead, it uses your own property's data, separately for each key event, and applies machine learning to estimate how much each ad interaction contributed to the outcome. Where your data sharing settings allow it, the model can also draw on aggregate data. Two properties with the same channels can therefore end up with different credit splits, because each model is built per advertiser and per key event.

For YouTube, the model counts engaged views as well as clicks. Google defines an engaged view as watching 30 seconds of an ad (or the whole ad if it is shorter), or interacting with elements such as a teaser card, companion banner, call-to-action or end screen. One practical point to know: data-driven attribution can reattribute a conversion for up to 7 days after it happens, so very recent days in your reports may still shift slightly.

Paid and organic last click

This model gives 100% of the credit to the last channel the user clicked through from before the key event (or the last engaged view, for YouTube). Direct traffic is ignored. If someone found you through organic search, came back two days later by typing your URL, and then bought, organic search gets the credit. Google's documentation also calls this approach last non-direct click.

Google paid channels last click

This model gives 100% of the credit to the last Google Ads channel the user clicked through from. If the path to the key event contains no Google Ads click at all, the model falls back to paid and organic last click, so the event is still credited somewhere. It suits teams that want GA4's view of conversions to line up more closely with how Google Ads sees its own clicks.

What was removed in 2023 and why it matters

GA4 used to offer four rule-based models alongside last click: first click, linear, time decay and position-based. Google's GA4 Help pages state that these four models are no longer available as of November 2023. Properties that relied on them now use one of the three remaining options.

Each removed model encoded a fixed opinion about the journey. First click credited the channel that started it. Linear split credit evenly across touchpoints. Time decay leaned towards touchpoints close to the conversion. Position-based weighted the first and last touchpoints and spread the rest across the middle. None of them learned from the data, and they could disagree sharply with each other on the same journey. With them gone, the question is no longer "which rule do we believe" but "do we accept a modelled answer, or do we want a simple last-click answer that is easy to explain".

If old reports or decks still show first-click or linear numbers, treat them as historical and do not try to recreate them in current GA4. Comparing a new data-driven report against an old linear export will mislead more than it informs.

The model comparison and conversion paths reports

GA4 keeps its attribution reports in the Advertising section of the left-hand navigation. Two of them are worth knowing well.

Key event attribution models report

Go to Advertising, then Attribution, then Attribution models. The report puts key event metrics for two models side by side, broken out by primary channel group, with key events, revenue and a percentage change column. Choose the models from the drop-downs at the top of each column. This is the quickest way to see whether switching from last click to data-driven would move credit towards, say, paid social or display, before you change anything in Admin.

Two details are easy to miss. Attribution reports only include data from 14 June 2021 onwards. The report also uses event time by default, meaning key events are counted when they happened. You can switch to ad interaction time, which shows the touchpoints that fell inside your date range, even if the key events they led to came later.

Key event attribution paths report

Under Advertising, open Key event attribution paths. This report shows the sequences of touchpoints that led to key events, along with key events, purchase revenue, days to key event and touchpoints to key event. Its chart splits each path into three segments:

• Early touchpoints: the first 25% of touchpoints in a path, rounded to the nearest whole number. Empty when the path has only one touchpoint.
• Mid touchpoints: the middle 50%. Empty when the path has fewer than three touchpoints.
• Late touchpoints: the last 25%, rounded the same way. When a path has a single touchpoint, the late segment receives all the credit.

Use the paths report to answer questions a single model cannot. If most paths to a lead form have one touchpoint, the model choice barely matters.

How GA4 lookback windows work

The lookback window sets how far back from a key event a touchpoint can still earn credit. GA4 has separate windows for two kinds of key event:

• Acquisition key events (first_open and first_visit): 30 days by default, with 7 days as the alternative.
• All other key events: 90 days by default, with 30 or 60 days as alternatives.

Engaged-view key events sit outside these settings and have a default of 3 days. Changes to the lookback window apply going forward only. A shorter window gives less credit to channels that start long consideration journeys, so shorten it only when your buying cycle is genuinely short.

Why GA4 and Google Ads conversion numbers differ

A gap between GA4 and Google Ads is normal and usually explainable. Google's own Help pages list several causes:

• Date of click versus date of conversion. Google Ads reports a conversion against the date and time of the ad click. GA4 reports it on the day the conversion happened. Google's example: a click on 19 July and a purchase on 20 July shows on 19 July in Google Ads and on 20 July in GA4. Daily and monthly totals drift apart around every date boundary.
• Counting method. Google Ads can count "Every" conversion or only "One" per click or interaction. Two lead form submissions in separate sessions can show as 2 in GA4 but 1 in Google Ads when the action is set to One.
• Windows and cookies. A Google Ads conversion window can be set between 1 and 90 days, and Google Ads cookies expire 90 days after a click. GA4 cookies can last up to 2 years, so a late conversion may appear only in GA4.
• Import delay. GA4 conversion data is imported into Google Ads up to 24 hours after it occurs.
• Different settings for key events and conversions. GA4 now separates key events, which drive its behaviour and acquisition reports, from conversions, which have their own attribution settings under Advertising, then Conversion management. The two can use different models, channel settings, windows and counting methods. Because conversions are event-scoped, they can use data-driven attribution in place of last non-direct click.
• Modelling and identity signals. Reports can include modelled conversions where Google cannot observe every conversion directly, and consent settings affect which identity signals are available.

To compare fairly, line up the same model, the same channels, the same window and the same counting method, and use the Google Ads "All conv. (by conv. time)" column when you compare against GA4, which dates conversions by when they happened.

Changing the reporting attribution model safely

The change itself takes a minute. In Admin, under Data display, click Events, then Key events attribution. Choose a model in the reporting attribution model section and click Save. You need the Marketer role or above at the property level.

The consequences deserve more thought than the clicks:

1. The model change is retroactive. Switching the reporting attribution model rewrites historical and future data in the affected reports. Export the current figures first if anyone relies on them.
2. Only event-scoped dimensions move. The model affects key events reports and explorations that use event-scoped traffic dimensions such as Source, Medium, Campaign and Default channel group, and the metrics key events, total revenue, purchase revenue and total ad revenue. Session-scoped and user-scoped dimensions such as Session source or First user medium stay as they are. This is why two reports in the same property can credit the same channel differently.
3. Window and channel changes are not retroactive. Unlike the model, changes to the lookback window and to which channels can receive credit apply from the moment you save onwards.
4. Check the comparison report first. Run the key event attribution models report with your current and proposed models and look at the percentage change by channel. If one channel swings by a large margin, warn its owner before the switch.
5. Write the change down. Note the date, the old and new model and who approved it, so future readers know why the trend line moved.

The channels setting deserves one more note. Choosing paid and organic channels applies only to web conversions. App conversions always use Google paid channels, so a mixed web and app property can show two different logics inside the same reports.

What this means in the Thai market

When a purchase or lead journey continues off the website, in a LINE Official Account, a phone call or a chat, keep one GA4 rule in mind: GA4 can only credit a channel for a key event it actually records, so if a LINE button click is not tracked as a key event, the touchpoints before it earn nothing under any model. In that case the first job is tracking, not model choice.

Second, because app conversions always use Google paid channels, the web and app sides of the same property can tell different stories about organic search and social. Read them separately.

Third, if a bilingual site runs campaign tags in a mix of Thai and English, inconsistent source and medium values split one campaign into several rows in the model comparison report, which makes any model look noisy. Agree one naming convention before you judge a model. If you are unsure how your property is set up, a GA4 setup and migration review is the place to start, and a broader SEO audit will show whether organic traffic is being tracked consistently enough to receive fair credit.

Frequently Asked Questions

Which attribution models are available in GA4 today?

GA4 offers data-driven attribution, paid and organic last click, and Google paid channels last click. Data-driven is the default and is labelled as recommended in the attribution settings.

When were first click, linear, time decay and position-based removed?

Google's GA4 Help pages state these four models are no longer available as of November 2023. Properties that used them now report with one of the three remaining models.

Does changing the attribution model change old data?

Yes, changing the reporting attribution model applies to historical and future data in the affected reports. Changes to the lookback window and to which channels can receive credit only apply going forward.

Why does Google Ads show more or fewer conversions than GA4?

The most common reasons are that Google Ads dates conversions by the ad click while GA4 dates them by when they happened, and that the two often use different models, windows and counting methods. Imported GA4 conversions can also take up to 24 hours to appear in Google Ads.

Which GA4 attribution model should most businesses use?

Data-driven attribution is the sensible default for most properties, because it learns from your own key events rather than following a fixed rule. Paid and organic last click is the simpler choice when you need numbers that are easy to explain, and the model comparison report shows how far apart the two land for your data.

If your attribution numbers feed bidding and budget decisions, the settings in GA4 and in your ad account need to agree. Our Google Ads management team and Relevant Search for Google Ads can help line up conversion tracking, attribution and reporting so the numbers you act on mean the same thing in both places.