Attribution models are rules or algorithms that decide how much credit each touchpoint, such as an ad click, an email or an organic visit, gets for a conversion. As of October 2026, Google Analytics 4 and Google Ads only offer data-driven and last click models, because Google removed first click, linear, time decay and position-based attribution in 2023.
That change still confuses people, because many guides list models you can no longer select in Google's tools. This guide explains the types of attribution models, what GA4 and Google Ads support today, and how to choose a model for B2B lead generation, where the sales cycle is long and the real conversion happens in the CRM. It is part of our analytics, tracking and privacy hub.
What are attribution models?
An attribution model is the logic that splits credit for one conversion across the touchpoints that came before it. The same conversion can be credited to paid search, to LinkedIn or to email, depending on the model you pick.
Take a buyer who clicks a LinkedIn ad, later finds you through organic search, and finally requests a demo after clicking a Google search ad. Last click gives the paid search ad 100% of the credit. First click gives it all to LinkedIn. A linear model gives each a third. None of these is "true". Each model answers a slightly different question, and the choice changes which campaigns look profitable and how automated bidding behaves.
Types of attribution models
Attribution models fall into three groups: single-touch, rule-based multi-touch and data-driven. The table compares the common ones and shows where you can still use them.
| Model | How credit is split | Good for | Available in GA4 or Google Ads? |
|---|---|---|---|
| Last click | 100% to the last click before the conversion | Simple reporting, short buying cycles | Yes, both |
| First click | 100% to the first recorded touch | Seeing which channels start journeys | No, removed in 2023 |
| Linear | Equal credit to every touch | A neutral view of all touches | No, removed in 2023 |
| Time decay | More credit to touches closer to the conversion | Short campaigns and promotions | No, removed in 2023 |
| Position-based (U-shaped) | Most credit to the first and the lead creation or last touch | Lead generation funnels | No, but common in CRM and attribution tools |
| W-shaped | Large shares to first touch, lead creation and opportunity creation | B2B funnels with a sales stage | No, CRM and attribution tools only |
| Data-driven | Credit based on how touches change conversion probability in your data | Accounts with enough conversion volume | Yes, both (default in Google Ads) |
Single-touch models are easy to explain but ignore everything except one interaction. Rule-based multi-touch attribution models spread credit by a fixed formula you can explain to anyone, but the formula is a guess. Data-driven models replace the guess with a statistical estimate from your own data.
B2B tools often add funnel milestones. Adobe Marketo Measure, for example, documents a U-shaped model that gives 50% each to the first touch and the lead creation touch, a W-shaped model that gives 30% to first touch, lead creation and opportunity creation with 10% spread over touches in between, and a full path model that adds the closed-won touch.
Attribution models in Google Analytics 4
GA4 offers three attribution models: data-driven attribution, paid and organic last click, and Google paid channels last click. According to Google's Analytics Help, first click, linear, time decay and position-based models have not been available since November 2023.
Paid and organic last click is the same model as last non-direct click: it ignores direct visits and gives 100% of the credit to the last channel the user clicked through. Google paid channels last click gives all credit to the last Google Ads click and falls back to paid and organic last click when there is no Google Ads click in the path. All GA4 models exclude direct visits unless the whole path is direct. AI assistant visits that arrive without a referrer or tags count as direct and get no credit, which is why classifying Gemini referral traffic and other AI sources correctly matters.
You set the model under Admin, Data display, Events, Attribution settings. You need the Marketer role or higher. Two details matter:
- Changing the reporting attribution model applies to historical and future data, and it only affects event-scoped traffic dimensions such as Source, Medium and Campaign. Session- and user-scoped dimensions, such as Session source, do not change.
- The key event lookback window defaults to 30 days for acquisition key events (with 7 days as an option) and 90 days for other key events (with 30 or 60 days as options). Lookback changes only apply going forward.
To compare models side by side, use the attribution models report under Advertising, Attribution. Google notes one trap: for Google Ads conversions based on GA4 key events, Analytics uses last click, so only key events where Google Ads was the last non-direct click become conversions in Google Ads, whatever model you select there.
Attribution models in Google Ads
Google Ads supports two attribution models: data-driven and last click. Data-driven is the default for most conversion actions, and Smart Bidding optimizes on the conversions credited by the model you select.
Google announced the change on April 6, 2023. Starting in June 2023, advertisers could no longer select first click, linear, time decay or position-based models for conversion actions that did not already use them. Starting in September 2023, conversion actions still on those models were switched to data-driven attribution, with last click as the alternative. In the same announcement Google said that less than 3% of Google Ads web conversions were attributed with those four models at the time.
You set the model per conversion action in the conversion settings. The click-through conversion window, which defaults to 30 days, determines how far back a click can get credit. The attribution reports, including Conversion paths, Path metrics, Assisted conversions and Model comparison, sit under Goals, Attribution. Model comparison is the quickest way to see which campaigns gain or lose credit before you switch. For more on bidding and budgets, see our Google Ads hub and the guide to Google Ads cost.
How data-driven attribution works
Data-driven attribution uses machine learning to compare the paths of people who converted with the paths of people who did not, and estimates how much each touch changed the chance of converting. Credit follows that estimated contribution, not a fixed rule.
Google's documentation for GA4 describes a counterfactual approach: the model looks at factors such as time from the key event, device type, number of ad interactions, the order of ad exposure and the type of creative, then compares what happens with and without a given touch. The output is fractional credit, so you may see decimals in key event columns.
The limits are worth stating plainly. Data-driven attribution only sees touches Google can measure, so offline sales calls, events and most dark social stay invisible. It works with the conversions you send it, so if your conversion is a form fill, it optimizes for form fills, not for qualified pipeline. And it is a black box: you cannot audit the weights the way you can audit a linear model.
Attribution models for B2B lead generation
For B2B lead generation, the most useful attribution usually happens in the CRM, not in the ad platform. That is where you can link marketing touches to opportunities and revenue instead of to form submissions.
Ad platform attribution answers "which campaign got the lead". Revenue attribution models in a CRM or a dedicated attribution tool answer "which touches show up in deals that closed". With sales cycles of several months, those answers can point in different directions.
Two practical steps close part of the gap. First, send later funnel stages back to the ad platforms, for example qualified leads and won deals through offline conversion imports, so automated bidding learns what a good lead looks like. Second, report on a milestone model, such as W-shaped, in the CRM, while letting Google Ads use data-driven attribution for bidding. Email and nurture touches matter here too, and our guide to email automation covers how to track them.
How to choose an attribution model
Choose the attribution model based on the decision you need to make, the data you have and where the real conversion happens. There is rarely one model that serves both bidding and reporting.
- Define the conversion that matters. For lead generation, that is a qualified lead or opportunity, not every form fill.
- Use data-driven attribution for bidding. In Google Ads it is the default for most conversion actions. Switch to last click only for a specific reason.
- Compare before you switch. Use the Model comparison report in Google Ads or the attribution models report in GA4 to see which campaigns gain or lose credit.
- Pick one reporting model and document it. Write down which model and lookback window your reports use, so everyone compares like with like.
- Report pipeline in the CRM. Use a milestone or multi-touch model there for budget decisions across channels.
Common mistakes with attribution models
- Following guides that recommend linear or time decay in Google Ads. These models were removed in 2023. Check what your account actually offers before you plan around them.
- Comparing GA4 and Google Ads numbers without matching settings. Different models, lookback windows and conversion definitions make the totals differ. Use paid and organic last click in GA4 when you compare with Google Ads.
- Switching models in the middle of a reporting period. In GA4 the change applies retroactively to reports, so last month's numbers change too. Note the date of every change.
- Optimizing data-driven attribution on the wrong conversion. If the primary conversion is a newsletter signup, bidding chases newsletter signups. Make qualified leads the primary goal where possible.
- Treating any model as the truth. All attribution models are estimates. Back up big budget shifts with tests, such as geo or holdout experiments, where the spend allows it.
