Attribution Models Compared for Ecommerce in 2026: First Click, Last Click, Linear, Time Decay, Position-Based

An attribution model is the rule that decides how much credit each marketing touchpoint gets for a sale. In 2026 the interesting question is no longer which of the five classic models is “best.”

It is which ones you can still use, since Google has removed four of them from Ads and GA4, Shopify has added five to its marketing reports, and Meta never offered a choice at all.

This post compares the models with a worked example, covers what each platform now supports, and gets to the part most comparisons skip: the model only redistributes credit among the touchpoints your tool saw, and that set matters more than the rule.

What is an attribution model?

An attribution model assigns conversion credit to the touchpoints in a customer’s path according to a fixed rule or a statistical estimate.

Single-touch models give all the credit to one interaction.
Multi-touch models split it across several.
The model changes how a sale is divided among channels; it does not change whether the sale happened or which touchpoints were recorded.

The five classic attribution models, with one order

The fastest way to understand the models is to run the same order through all five. Dataslayer’s comparison uses a $200 fashion order with five touches: a Meta prospecting ad, a Google Shopping click, an email, a Meta retargeting ad, and a branded search. Here is where the $200 goes.

First click

100% to the first recorded touchpoint. The Meta prospecting ad gets $200. This is the model for measuring what introduces new customers, and it says nothing about what closes them.

Last click

100% to the last touchpoint before purchase. Branded search gets $200. It rewards whatever sits closest to checkout, which is usually branded search, email, or retargeting, and it is the default almost everywhere because it is the easiest to compute.

Linear

Equal split. Each of the five touches gets $40. Fair on its face, and useless for deciding anything, because it assumes every interaction mattered the same amount.

Time decay

More credit the closer the touch is to purchase. Branded search gets around $80, the Meta prospecting ad around $8. A softened version of last click.

Position-based (U-shaped)

40% to the first touch, 40% to the last, 20 percent shared across the middle. Meta prospecting and branded search get $80 each. This is the compromise model, built to reward both acquisition and conversion.

Same order, same five touches, five different verdicts on whether Meta prospecting was worth $8 or $200. That is the whole point. The model is a lens, and the lens you pick decides which channel looks like a hero.

What changed about attribution models in 2026?

The platforms moved in opposite directions, and the practical result is that the model you can pick depends on where you are looking.

Google Ads and GA4 dropped four models. First click, linear, time decay, and position-based are gone. Google’s stated reason was that fewer than 3 percent of conversion actions still used them, as Search Engine Journal reported. What remains is data-driven attribution, which uses Google’s own model to spread fractional credit, and last click. Data-driven attribution needs volume to work; we’ve written about the threshold problem, where smaller accounts effectively land on last click whether they chose it or not.

GA4 is down to three options. Data-driven (the default since November 2023), paid and organic last click, and Google paid channels last click, per WeltPixel’s GA4 attribution guide. Switching between them recalculates how recorded credit is split; it does not reprocess what GA4 saw in the first place.

Shopify went the other way. Its marketing reports now offer five selectable models per report (last non-direct click, last click, first click, any click, linear), per WeltPixel’s guide, over a 30-day window. Two people looking at the same orders can see different channel credit depending on the dropdown.

Meta doesn’t offer a model. Meta credits the last interaction inside its attribution window, and its windows are the lever: 7-day click, 1-day engage, 1-day view. There is no first-click view of Meta data inside Ads Manager.

So a media buyer in 2026 has data-driven or last click on Google, last touch with windows on Meta, five rule-based options on Shopify, and none of them agreeing about the same order.

Which attribution model should an ecommerce brand use?

Use two, for two different decisions, and stop looking for the one that is “right.”

Last click (or last non-direct click) is the defensible model for spend decisions, because it credits the touch that demonstrably preceded the purchase.

First click is the model for understanding acquisition, because it shows what brought the customer in.

Reading both on the same period tells you which channels open the funnel and which close it, and the gap between the two views is where your budget questions live.

That is also, not coincidentally, how AdBeacon customers describe using the platform in their G2 reviews: flipping between first click, mid-funnel, and last click to see where paid social really sits.

Linear, time decay, and position-based are fine for a slide and poor for a decision, because their weightings are arbitrary.

Data-driven attribution is a reasonable idea with two problems for ecommerce: it needs more conversions than most stores have, and the model lives inside the platform whose ads it is grading.

Our post on last click’s blind spots covers where the default fails; our multi-touch attribution guide covers what the multi-touch models add.

The part the model can’t fix: which touchpoints are in the set

Every attribution model divides credit among the touchpoints the tool recorded. It cannot credit an interaction it never saw, and it cannot un-credit one it shouldn’t have counted. That makes the eligibility rule more important than the split.

On Meta, the eligible set includes 1-day view-through impressions and 5-second video engagements, so any model run on Meta data starts from a set that includes unverifiable touches.
On Google, the set is whatever the Google tag and enhanced conversions captured.
On Shopify, it is the last UTM the shopper landed with.

None of these sets are the same, which is why no model makes the platforms agree.

First-party, click-only attribution defines the set differently: a touchpoint is eligible only if it was a captured ad click, joined to the order through your own session record.

Then you run first click, last click, or any model you like, over that set.

The model becomes a choice you make on data you trust, instead of a choice the platform makes on data you can’t check. It undercounts some upper-funnel influence, and it never credits an impression on a platform’s word. Where MMM and incrementality fit around that is a separate question.

What to actually do

Pick last click for spend, first click for acquisition, and report both. Put them side by side for the same month. The channels that move most between the two views are the ones to investigate.
Know which model each platform is showing you. Google Ads: data-driven or last click, check per conversion action. GA4: check the property setting. Shopify: check the dropdown on every report before you compare. Meta: check the attribution window.
Stop comparing model outputs across platforms. A first-click number from Shopify and a data-driven number from Google are not two measurements of one thing. They are two things.
Audit the eligible set before the model. For any attribution reporting you rely on, ask what counts as a touchpoint. If the answer includes impressions, the model on top is decorating a guess.
Run your models on a click-only, first-party set. That is the setup where switching models tells you something about your customers rather than about the platform’s defaults.

Attribution models are lenses, and 2026 took several of them away on Google while adding them on Shopify. The lens matters less than what it is pointed at. If you want to see first click and last click run over a click-only, first-party record of your own orders, side by side with what Meta and Google report, book a live AdBeacon demo.

—-

FAQ

What are the main attribution models?

First click, last click, linear, time decay, and position-based are the classic rule-based models. Data-driven attribution uses a statistical model instead of a fixed rule. Last click and last non-direct click are the most common defaults.

Which attribution models did Google remove?

First click, linear, time decay, and position-based were removed from Google Ads and GA4, leaving data-driven attribution and last click. Google said fewer than 3 percent of conversion actions used the removed models.

Which attribution model does Shopify use?

Last non-direct click over a 30-day window by default, with last click, first click, any click, and linear selectable per report in the marketing channel performance reports.

What is the best attribution model for ecommerce?

There isn’t one. Use last click for spend decisions and first click for acquisition insight, read both for the same period, and make sure the touchpoints being credited are verified clicks rather than impressions.

What is the difference between an attribution model and multi touch attribution?

Multi touch attribution is any approach that splits credit across several touchpoints. Linear, time decay, position-based, and data-driven are all multi touch attribution models. First click and last click are single-touch.