Marketing Attribution Models: Which One to Use

Olamide Sule, founder of Dolphin Analytics: a digital analytics expert based in London delivering solutions for agency and in-house clients.

Attribution is one of the questions we field most weeks, usually in the form “why does every platform claim the same sale?” This is the explainer we give clients before we touch their accounts: what each attribution model actually does, where it misleads, and why the model is rarely the real problem.

What is a marketing attribution model?

A marketing attribution model is a rule for splitting credit for a conversion across the touchpoints that led to it. A buyer might click a Google ad, read a blog post from organic search, then return by typing the URL directly a week later. The attribution model decides which of those touchpoints gets the credit for the sale, and how much. Change the model and the same journey produces a different winner, even though nothing about the customer changed.

That is why two dashboards can report the same revenue against completely different channels. They are not wrong; they are applying different rules to the same events.

The main marketing attribution models, compared

Six models cover almost everything you will meet in GA4, Google Ads and the attribution tools built on top of them. Here is how they split credit before we get into each one.

Model	How it splits credit	Best fit	Main weakness
First-touch	All credit to the first interaction	Judging top-of-funnel awareness	Ignores everything that closed the sale
Last-touch	All credit to the last interaction	Short, direct buying paths	Ignores everything that started the journey
Linear	Split evenly across all touchpoints	A simple, fair-feeling baseline	Treats a throwaway click like a decisive one
Time-decay	More credit to touchpoints nearer the conversion	Longer sales cycles with a clear close	Under-credits early awareness work
Position-based (U-shaped)	Most to first and last, rest shared	Considered purchases with long journeys	The split is still an assumption you did not choose
Data-driven	Weighted from your own conversion data	Accounts with high conversion volume	Needs volume, and the logic is a black box

First-touch and last-touch attribution

First-touch (also called first-click) gives all the credit to the first interaction. Last-touch (last-click) gives it all to the final one. Both are single-touch models, which makes them easy to read and easy to misread. First-touch flatters awareness channels; last-touch flatters whatever sits closest to the sale, usually branded search or direct. Neither sees the middle of the journey, so both overstate one channel and hide the rest.

Last-touch is still the most common default in ad platforms, which is part of why paid channels look so strong in their own dashboards.

Linear, time-decay and position-based attribution

These are multi-touch models, and they try to fix the single-touch blind spot by sharing credit. Linear splits it evenly. Time-decay gives more weight to touchpoints closer to the conversion, on the logic that recent interactions did more of the closing. Position-based, often called U-shaped, hands the biggest shares to the first and last touch and spreads what remains thinly across the middle.

Multi-touch models read more fairly, but every split is a guess baked into the tool. Position-based decides that the first and last touch matter most before it has seen your data. That is a reasonable assumption for a long B2B journey and a poor one for an impulse purchase.

Data-driven attribution

Data-driven attribution (DDA) drops the fixed rules and works out each touchpoint’s weight from your own conversion data, comparing the paths that converted against the paths that did not. It is the model GA4 now uses by default. In 2023 Google removed first-click, linear, time-decay and position-based from GA4 and Google Ads, leaving data-driven and last-click as the only options, per Google’s own attribution documentation.

DDA is the most accurate model on this list when you have the conversion volume to train it. The trade-offs are real: it needs a meaningful number of conversions to produce stable weights, and the model is a black box, so you cannot fully explain why a channel’s credit shifted month to month.

Which marketing attribution model should you use?

Match the model to the journey, not to the dashboard that flatters you most. Use data-driven attribution if you have the conversion volume, because it weights touchpoints from your real data instead of a rule someone else chose. Below that volume, use position-based for considered purchases with long research journeys, and last-touch for short, direct paths where one channel genuinely does most of the work.

For top-down channel-mix questions across everything, including offline and brand spend, attribution models are the wrong tool entirely. That is the job of marketing mix modelling, which works from aggregate data and needs no user-level tracking. The two are complements: MMM for the big picture, attribution models for the campaign-level detail.

What marketing attribution software and tools actually do

Marketing attribution software applies these models to your data and joins touchpoints across channels that would otherwise sit in separate silos. GA4 and Google Ads include attribution for the Google stack; each ad platform reports on its own conversions; dedicated tools add the cross-channel join and a single view. The category is crowded, and most of it competes on how many sources it can stitch together.

The catch is that the tool only reasons over the data it receives. If consent blocks a chunk of sessions, if a conversion tag stopped firing after a checkout redesign, or if the ad platform is counting view-through conversions the visitor never clicked, the smartest model on the market still gives you a confident, wrong answer. Buying more attribution software does not fix missing or duplicated data underneath it; it just renders the bad data more persuasively.

Why the model matters less than whether your tracking can see the touchpoints

The uncomfortable truth is that most attribution disputes are tracking problems wearing an attribution costume. Before any model can split credit fairly, every touchpoint has to be captured, deduplicated and joined to the same customer. When that foundation is broken, changing the model just reshuffles credit between channels the data can already see, while the touchpoints it cannot see stay invisible.

This is the work we do inside client accounts most weeks: reconciling what the platforms claim against a single source of truth before trusting any model’s output. On one private aviation engagement, we mapped the full enquiry-to-deal flow across three websites and the client’s HubSpot CRM. Once the sources were joined, organic search turned out to drive 50.9% of the deal contacts that had any web attribution at all, the largest single source, ahead of direct traffic. No off-the-shelf attribution model would have surfaced that, because the touchpoints were scattered across systems that never spoke to each other.

Cross-platform overclaim is the same story from the ad side: every platform counts the sale under its own model, so the totals never add up. We break that down in our guide to why Meta reports more conversions than your sales data, and it sits at the centre of the attribution reconciliation work on our Insight pillar.

If your attribution numbers already look off, and you are not sure whether it is the model or the tracking underneath it, start with the tell us what’s broken, or book a call.