Attribution Models in PPC: First Click, Last Click, or Data-Driven?

If you are running Google Ads in 2026, the honest answer is that your choice has narrowed to two options: last click or data-driven attribution. Google retired first click, linear, time decay and position-based models inside Google Ads back in 2023, and Google Analytics dropped most of them from reporting the same year. That does not make first click irrelevant, it just means you now read it as a diagnostic rather than set it as a rule.

The model you pick decides which keywords look profitable, which campaigns get budget, and what Smart Bidding chases. Get it wrong and you defund the top-of-funnel terms that started the whole purchase.

What an Attribution Model Actually Decides

An attribution model is a rule for splitting credit for one conversion across the multiple ad clicks that preceded it. A single sale might follow a generic research click, a comparison click, a branded search and a retargeting click, all from the same person over eleven days.

Someone has to decide who gets the sale. Attribution models are that decision, written as math instead of opinion.

• Last click gives 100% of the credit to the final ad click before the conversion.
• First click gives 100% to the click that opened the journey.
• Linear splits credit evenly across every touchpoint.
• Time decay weights clicks closer to the conversion more heavily, usually on a seven-day half-life.
• Position-based gives 40% to the first click, 40% to the last, and spreads the remaining 20% across the middle.
• Data-driven attribution uses your own account’s conversion paths to calculate how much each touchpoint actually moved the needle.

Google Quietly Retired Half the Menu

Between April and September 2023, Google removed first click, linear, time decay and position-based as selectable options for conversion actions in Google Ads, leaving last click and data-driven. Google Analytics 4 followed, and its reports now centre on cross-channel data-driven and last click, plus a paid-channels-only last click view. You can read the current options in Google’s own attribution documentation.

Last Click Attribution

Last click is simple, stable and easy to explain to a finance director. It is also structurally biased toward branded search and retargeting, because those are usually the closest clicks to the checkout.

Run last click for six months on a considered purchase and you will slowly starve the non-brand terms that introduced people to you in the first place. The reports stay green while the pipeline thins.

Data-Driven Attribution

Data-driven attribution compares converting and non-converting paths in your account and assigns fractional credit based on observed lift. If people who saw a particular non-brand keyword convert at a meaningfully higher rate later, that keyword earns a share of the conversion, often something like 0.34 instead of a whole number.

Google removed the old eligibility thresholds in 2021 (at one point you needed roughly 15,000 clicks and 600 conversions in 30 days), so most accounts can now use it. Very low-volume accounts may still see the model fall back toward last-click-like behaviour because there simply are not enough paths to learn from.

First Click, Time Decay and Position-Based

These still exist outside Google Ads: in Meta’s comparison reporting, in most CRM attribution tools, and in any custom model you build yourself. First click remains useful as a lens for answering a specific question, which is which campaigns are actually generating new demand rather than harvesting it.

Time decay is the sensible compromise for businesses with a 30 to 90 day sales cycle, because it credits the nurture touches without pretending a six-week-old display click deserves equal billing with the demo request.

A Worked Example: Four Touchpoints, One $400 Sale

Say a customer buys a $400 product after four paid clicks: a broad non-brand search, a YouTube in-stream click, a Shopping click, and finally a branded search click. Here is how each model books that revenue.

• Last click: branded search takes the full $400, the other three record zero.
• First click: the non-brand search takes all $400.
• Linear: each touchpoint books $100.
• Time decay: roughly $180 to branded, $120 to Shopping, $60 to YouTube, $40 to non-brand.
• Data-driven: maybe $150 to non-brand, $110 to Shopping, $90 to branded, $50 to YouTube, depending on what your account’s paths show.

Nothing about the customer changed. Only the bookkeeping did, and that bookkeeping is what your ROAS targets are built on.

Attribution Changes Your Bidding, Not Just Your Reports

This is the part most write-ups skip. Your attribution model feeds Smart Bidding directly, so switching models rewrites the training signal that Target CPA and Target ROAS optimise against.

Under last click, the algorithm learns that branded queries convert and bids them up aggressively. Under data-driven, upper-funnel keywords start showing partial conversions, so the system is willing to pay more for them. Expect CPCs on non-brand terms to rise after a switch, and expect that to be the point rather than a problem.

If your conversion tracking is shaky to begin with, no model will save you. Sort that first, ideally alongside offline conversion and call tracking, because a lead that closes three weeks later by phone is invisible to every attribution model until you feed it back in.

When Last Click Is Still the Right Call

Data-driven is the default recommendation, but it is not automatically correct for every account. Last click holds up in a few specific situations.

• Single-touch businesses: emergency plumbers, locksmiths and same-day services where the path is one click and a phone call.
• Accounts under 15 conversions a month: the model has too little path data to find real patterns.
• Audit and reconciliation work: when you need whole-number conversions that tie back to the CRM without decimals.
• Brand-only campaigns: where there is no upper funnel to give credit to anyway.

Local service accounts in particular often see almost no difference between the two models, which is one reason working with a local agency that understands short-path buying behaviour tends to beat a generic enterprise playbook.

Where Data-Driven Attribution Breaks Down

Data-driven attribution is a black box. Google does not publish the weights, so you cannot fully audit why a keyword earned 0.27 of a conversion, which makes some stakeholders uncomfortable.

It also only sees Google’s own surfaces. Organic search, email, a podcast mention and a Meta retargeting click are all outside the model, which is why the sum of your platform-reported conversions usually exceeds your actual order count by 20% to 40%.

Invalid traffic muddies it further. Fake clicks in the path do not convert, but they do pollute the data the model learns from, so protecting your budget from bot clicks is part of keeping attribution honest. Privacy changes matter too: with consent mode and modelled conversions, a growing share of the paths the model sees are statistical estimates rather than observed clicks, as Google Analytics documents in its attribution reference.

Switching Models Without Panicking Your Stakeholders

Changing the model is a two-click job in the conversion action settings, but the fallout lands in the reports for weeks. Plan the change rather than flipping it on a Friday.

1. Screenshot your baselines. Record 30 days of conversions, CPA and ROAS by campaign before you touch anything.
2. Run the model comparison report first. Google Ads shows what your numbers would have looked like under each model, so you know the size of the shift in advance.
3. Change one conversion action at a time if you track several, starting with your primary revenue action.
4. Leave bid targets alone for 14 days. Smart Bidding needs a learning window, and moving targets mid-transition makes the cause of any change impossible to isolate.
5. Warn everyone that historical data is not restated. Reporting reflects the model in place at the time, so month-over-month comparisons will be uneven for one cycle.

Build a dashboard that holds both views side by side. Pulling platform data into a custom Looker Studio report alongside CRM closed revenue is the cleanest way to show what changed in the ledger versus what changed in the business.

Reading Attribution Across Google, Meta and Everything Else

Every platform grades its own homework on a different curve. Google Ads counts conversions on a click-based window you set (commonly 30 days), while Meta defaults to 7-day click and 1-day view, meaning an ad someone scrolled past yesterday can claim today’s sale.

Those windows are not comparable, which is why channel budgets should be argued from a single source of truth such as your CRM or a blended CAC number. Reliable Meta Pixel and Conversions API setup at least makes the social side of the path measurable before you try to reconcile anything.

A practical rule used by most media buyers: use platform attribution for in-platform optimisation decisions, and use blended numbers for budget allocation between platforms.

Frequently Asked Questions

What Is the Difference Between Last Click and Data-Driven Attribution Models?

Last click assigns 100% of a conversion to the final ad click, while data-driven attribution splits that same conversion into fractions across every click in the path based on measured contribution. A conversion that last click records as 1.0 on a branded keyword might appear as 0.4 on brand, 0.35 on a non-brand term and 0.25 on Shopping under data-driven. The total conversion count stays the same; only the distribution changes.

What Is a Data-Driven Attribution Model?

A data-driven attribution model uses machine learning on your own account’s converting and non-converting paths to calculate how much each touchpoint increased the chance of a conversion. It has been the default for new conversion actions in Google Ads since 2021, when the previous minimum volume requirements were removed. Because it is calculated per account, two advertisers in the same industry can get quite different credit splits.

What Is the Difference Between Last Click Attribution and Multi-Touch Attribution?

Last click is a single-touch model that credits one interaction, while multi-touch attribution distributes credit across two or more interactions in the path. Linear, time decay, position-based and data-driven are all multi-touch approaches. Multi-touch generally produces a fairer view of assisted conversions, at the cost of fractional numbers that are harder to reconcile against invoices.

What Is the Difference Between First Touch and Last-Touch Attribution?

First touch credits the interaction that started the journey, and last touch credits the one that ended it, so they answer opposite questions. First touch tells you which campaigns generate demand, which matters for prospecting and audience work. Last touch tells you which campaigns close, which matters for retargeting and brand defence, and most mature accounts look at both before reallocating budget.

Get a Second Opinion on Your Attribution Setup

If your Google Ads reports and your bank account keep telling different stories, the attribution model is usually one of the first places to look. SEO Quirk can review your conversion actions, tracking and model choice, then show you what the numbers look like once the credit is assigned properly.