Attribution modelling is how you decide which touchpoints get credit for a conversion when a customer took several steps before buying, rather than handing all the credit to the last click.
Attribution modelling is how you decide which touchpoints get credit for a conversion when a customer took several steps before buying, rather than handing all the credit to the last click. Someone might find you through a Google search, come back a week later via a Facebook ad, then finally convert by typing your name in directly. Three touchpoints, one sale, so who earned it?
An attribution model is the rule you use to split that credit. Last-click gives it all to the final touchpoint. First-click, linear, position-based and time-decay were the other classic ways to split it, first to the opening touch, evenly across every step, weighted toward the first and last, or weighted toward the touchpoints nearest the sale, but Google retired all four from GA4 and Google Ads in 2023. GA4's default is a data-driven model, which uses your own conversion patterns to work out how much each step actually contributed rather than applying a fixed rule.
This is not academic. If you judge everything on last-click, the channels that introduce you to customers, the blog post, the first ad they ever saw, look worthless, because something else usually gets the final click. Cut them, and the enquiries quietly dry up. Attribution modelling is what stops you defunding the top of your funnel by mistake, and it changes how your ROAS reads depending on which model you trust.
The honest truth is that no attribution model is 'correct', they are all different lenses on the same messy reality, and anyone presenting one as the definitive answer is overselling it. Data-driven models are more sophisticated but harder to interrogate, they are something of a black box. And none of them sees everything: a customer who spotted your billboard, asked a friend, or clicked an ad on a device you never tied together leaves gaps no model can fill. I use attribution to make better decisions, not to pretend I have perfect knowledge of every journey.
Key points
- Attribution modelling decides which touchpoints get credit for a conversion, not just the last click.
- GA4 and Google Ads now offer two live models, last-click and data-driven; first-click, linear, position-based and time-decay were retired in 2023 but still explain how attribution thinking works.
- GA4 defaults to a data-driven model built from your own conversion patterns.
- Judging on last-click alone undervalues the channels that first introduce you to customers.
- The model you choose changes how your ROAS and channel performance read.
- No model is the truth; each is a different lens on the same messy reality.
Frequently asked questions
Common questions about attribution modelling.
For most businesses on GA4, the data-driven default is a sensible starting point, it adapts to your own patterns rather than applying a rigid rule. But you should still understand last-click, because many ad platforms report on it and it is the one most likely to mislead you. The right answer depends on how long your sales cycle is and how many channels you run. I walk through how these budget calls play out in my guide to what Google Ads management costs in Australia.
Because it rewards the finish and ignores the build-up. The last click is often someone typing your name in or clicking a brand search, which is the easy part, and the channels that introduced them and did the persuading get nothing. Judge your marketing purely on last-click and you will starve the very activities that fill the top of your funnel, then wonder why the enquiries thin out.
No, and be wary of anyone who says otherwise. Every model is an estimate, and none can see the whole journey, the friend's recommendation, the billboard, the research done on a device you never linked together. Attribution is a tool for making better spending decisions with imperfect information, not a perfect ledger of who influenced whom. Used with that in mind it is genuinely useful; treated as gospel, it misleads.
Want to know what is actually working?
Tell me what you are trying to measure and I will show you how to capture it.
Start a conversation