What three guides still teachsix models to pick from
first click · linear · time decay · position-based
Google's help page now lists two for a website conversion action, so a reader of a stale guide finds four options missing.
What the default doesdata-driven, since October 2021
credit split across clicks → 0.4 of a conversion
Credit is split whenever a lead followed more than one ad click, so the column can hold fractions. Keep that in mind when you divide spend by it.
Google began withdrawing first click, linear, time decay and position-based in 2023. The default since October 2021, data-driven, can leave fractions in the Conversions column, and none of the seven guides says so.
The retired models are the smaller problem: Google's help page lists only two for the setting, so a reader of a stale guide finds four options missing. The advice that can cost a small advertiser something is about eligibility. And none of the seven says what the default model does to the number you divide spend by. Data-driven attribution, which Google began rolling out as the default for new conversion actions in October 2021, can leave fractions in the Conversions column whenever a lead followed more than one ad click, and the dashboard I sell rounds them away.
What I checked
On 21 September 2026 I ran three phrasings through a US web search tool and took its first page for each: google ads attribution models, attribution model google ads and google ads attribution. That tool is not Google's own results page, which may order things differently. Together the three return thirteen results. One is a YouTube video and WordStream refuses this location, so eleven pages were read. Three of those are Google's own and one is a Wikipedia article about a 2017 product called Google Attribution, which leaves seven advertiser guides. Every count below is out of those seven, and each verdict rests on a sentence quoted in the sources.
The seventh is Usermaven, a vendor of attribution software. It walks through all six as options Google Ads offers and, further down, says "Google Ads provides two main attribution models: last-click attribution and data-driven attribution." I count it in neither of the first two rows.
The split follows the date on the page. The three that get it right all print a 2026 date: Growth Minded Marketing updated in March, AdNabu in April and Cometly in May. Usermaven carries an April 2025 byline and DataFeedWatch puts 2025 in its title. Grow My Ads and KlientBoost print no date on the article at all. Grow My Ads' metadata says it was published on 25 February 2023, six weeks before Google's announcement, when all six models still existed; a later edit in November 2023 left the list in place. KlientBoost's metadata gives only a modified date in September 2023; its text still calls last click the default for every new AdWords campaign and its images were uploaded in January 2018, so that date most likely records an edit rather than a rewrite.
Only two of the three correct pages say when it happened. AdNabu gives each retired model a status line, "Google no longer supports this attribution model" for first click, and never says in which year.
What happened to the other four Google Ads attribution models
Google's announcement is dated 6 April 2023 and describes two steps. First: "Starting in June 2023, we'll remove the ability to select first click, linear, time decay, and position-based attribution models for conversion actions in Google Ads that do not already use one of these models." The second: "Then, starting in September 2023, we'll switch any conversion actions still using these models to data-driven attribution."
The dates shift depending on where you read them, so here is each one as Google wrote it. The Google Ads API post of 20 April 2023 carries an update that the first step "will now begin mid-July instead of June". Google Analytics 4 gives its own end point: the four models "are no longer available as of November 2023". I found no completion date for Google Ads itself. Its help page says the four "are no longer supported by Google" and that "Conversion actions that used the deprecated attribution models have been upgraded to use data-driven attribution."
The reason Google gave was usage: "less than 3% of Google Ads web conversions are attributed using first click, linear, time decay, or position-based models." The same announcement says "Data-driven attribution, last click, and external attribution won't be impacted." External is for fractional credit you calculate in your own or a third-party attribution system and import, so for a conversion Google measures itself the choice is data-driven or last click.
Google's documentation disagrees with itself
The stale guides have company on Google's side, and blaming them alone would misrepresent what I found.
I also asked the API. GoogleAdsFieldService, which describes the schema and returns no account data, lists nine values for a conversion action's attribution model in v25, the current version, and the four retired models are among them. Someone writing a guide from the reference finds six models described, down to the one week half-life of time decay, with no line saying four are gone.
The thresholds are where a guide can mislead, because they tell a small advertiser the default model is out of reach. Usermaven says data-driven attribution "Requires at least 3,000 ad clicks and 300 conversions in the last 30 days" and calls it inaccessible to many small and mid-sized advertisers. KlientBoost prints 15,000 clicks and 600 conversions, which was Google's own AdWords guideline in 2017 and has since been dropped. Growth Minded Marketing and DataFeedWatch give 3,000 and 300 with a qualifier about conversion action types, close to Google's Switch to DDA page. Google's September 2021 announcement said "we're removing the data requirements", and its data-driven page recommends 200 conversions and 2,000 interactions in 30 days. It also warns: "Depending on data availability, the last click and data-driven attribution models can have the same results in certain situations."
Data-driven attribution turns leads into decimals
Google announced in September 2021: "We'll roll out data-driven attribution as the default model for all new conversion actions starting in October". Its help page on conversion data explains the side effect: "This is because data-driven attribution (DDA) and other attribution models attribute fractional credit for each conversion across multiple clicks. These fractions are represented as decimals like 0.33 or 0.50." Its page on changing models warns that you will "find decimals in your "Conversions" and "All conv." columns for the first time when you switch to a non-last-click model". Linear, time decay and position-based split credit too, so the decimals are older than the 2023 change.
All seven guides describe data-driven attribution sharing credit between clicks. AdNabu comes closest to the consequence: it says the model "affects how conversions are counted in the "Conversions" and "All conversions" columns" and that CPA and ROAS may change. None of the seven says the column will show decimals.
Say an enquiry arrives after three clicks: a generic search ad, a Performance Max ad, then a search for your business name. Under last click the brand campaign gets 1.0 and the other two get nothing. Under data-driven attribution the credit is shared, say 0.4, 0.2 and 0.4. The model decides the weights and Google does not publish them, so those three are invented for the example. The generic campaign spent $50 this month and that 0.4 is all it holds. Its cost per lead is $125, and there is nobody in it you could phone.
Two more things follow for a business that counts leads.
The model is set per conversion action. Google: "The attribution model you choose only affects the conversion action to which it's applied." One Conversions column can therefore add together actions counted under different models, and nothing in the column tells you which.
Credit lands on the day of each click. Google: "Google Ads reports conversions according to the date of the ad interaction." Under data-driven attribution it follows that a conversion whose clicks straddle the end of a month is shared between two months, so a monthly total can end in a decimal too.
Three things to do
Check the model on each conversion action
Goals, then Conversions, then Summary, and open each action's settings. The Conversions column adds your actions together whatever their model, so knowing which ones are data-driven tells you which rows can hold fractions.
Compare periods under one model
When you change a model, Google's own columns restate the history: "If you want to check how your historic conversions data would look with the attribution model you've just selected, you can add the "current model" columns". Use Conversions (current model) and Cost / conv. (current model) for any comparison that crosses the change.
Sum first, round last, and count people somewhere else
A campaign's 0.4 is real credit and belongs in the total, and rounding each row before adding loses it. For how many people enquired, use your booking system or CRM. The Conversions column answers a different question: which clicks earned the credit.
Where this sits in our own tool
Google Ads Vitals reads the Conversions column from the campaign report (or All conv., if that column comes first in the export) and calls it leads, which sits uneasily with a product page that promises the cost of one enquiry. The report has no column naming the attribution model, so the file cannot say which model produced the number.
Three parts of its display handle this badly, and none is fixed yet. Lead counts are printed rounded to whole numbers while cost per lead uses the unrounded total, so a campaign with $50 of spend and 0.4 of a conversion shows 0 leads next to a cost per lead of $125; I checked that by running the file's own formatting functions. The Leads tooltip says "Fractions appear because Google models some conversions.", a cause Google's paragraph on decimals does not give. And the monthly Leads chart says conversions are "counted in the month they were recorded by Google Ads", where Google reports them by the date of the ad interaction. The decimals survive in the arithmetic only when the export uses a decimal point: a CSV written with decimal commas has 0,40 read as 40.
Related reading
- Before you change how Google Ads counts your leads, read your repeat rate is the other setting that decides what the Conversions column counts
- Google Ads reporting: the number is still moving when you send the report, because credit lands on the date of each click
- Breakeven ROAS: the one field no calculator asks for, for re-basing a target after a model change
- Performance Max reports arrived without their labels
Questions people actually ask
For website and Google Analytics conversion actions, two: data-driven, which Google calls the default for most conversion actions, and last click. Google also keeps external attribution for credit calculated in another tool and imported. Google began withdrawing first click, linear, time decay and position-based in 2023.
Google announced it on 6 April 2023. The plan was to stop new selections from June 2023 (an update to Google's API post moved that to mid-July) and, from September 2023, to switch any conversion action still on one of the four to data-driven attribution. Google Analytics 4 says the four have been unavailable there since November 2023. I found no completion date for Google Ads itself.
Because of the attribution model. In Google's words, data-driven attribution and other attribution models "attribute fractional credit for each conversion across multiple clicks", so one enquiry can show as 0.4 in one campaign and 0.6 in another. This is not new: Google began rolling data-driven attribution out as the default for new conversion actions in October 2021, and the retired rules-based models split credit too. Under last click each conversion goes whole to one click, though Google still prints two decimal places.
Google's pages disagree. The data-driven attribution page says all conversion actions are eligible regardless of volume and recommends 200 conversions and 2,000 ad interactions in 30 days. The Switch to DDA page says some types of conversion action need 3,000 interactions and 300 conversions in 30 days to become eligible, and 2,000 and 200 to stay eligible. The 15,000 clicks and 600 conversions that one guide prints was Google's AdWords guideline in 2017, since dropped.
Google warns that with little data the two "can have the same results in certain situations", so on a small account the choice may change less than the guides suggest. What matters more is keeping one model long enough to compare periods, and moving your target cost per lead when you do change it.
It stays as it was. Google says a change affects how conversions are counted going forward. To see history under the new model, add the current model columns, such as Conversions (current model), which show how past data would have looked under the model you just selected.
WRITTEN BY
Olha · clinic data analyst
I build the reporting our managers open every morning at a multi-branch medical clinic — and package it so other practices don't have to start from scratch.
Published on 22 September 2026; the pages were read on 21 September 2026. Three limits worth stating outside the body text. The sample is seven guides from three phrasings on one day, and one site refused this location, so rerun the method rather than trust the counts. I did not open a Google Ads account to look at the attribution setting: what the interface offers is taken from Google's help pages and API reference, which this article also shows lagging in places. And the display problems I describe in my own $39 file are gaps I could close and have not; writing it here is cheaper than letting a buyer find it.