Attribution is the question of which marketing activity gets credit for a sale. It sounds administrative. It decides where budget goes, which makes it one of the most consequential things measurement does.
Why it is hard
A typical B2B buying journey might run: an organic search, a LinkedIn post two weeks later, a retargeting ad, a webinar, a direct visit, and finally a branded search before the form is submitted.
Six touchpoints. One sale. Who gets the credit? Every attribution model is a different answer to that question, and none of them is objectively correct.
First-touch attribution
All credit to the first interaction. In the example above, organic search takes everything.
Useful for: understanding which channels create awareness and introduce you to new audiences.
Blind spot: it ignores everything that actually persuaded the buyer. Optimise on it alone and you will overinvest in top-of-funnel and starve the channels that close.
Last-touch attribution
All credit to the final interaction — here, branded search.
Useful for: simplicity, and it remains the default mental model in most organisations.
Blind spot: it systematically over-credits the bottom of the funnel. Branded search looks extraordinary because people who already decided to buy searched your name. Cut the activity that made them aware of you and branded search quietly dries up.
Linear attribution
Credit split evenly across all touchpoints — here, one sixth each.
Useful for: a fair first look at which channels are involved at all.
Blind spot: it treats a passive impression and a demo request as equally influential, which they plainly are not.
Time-decay attribution
More credit to touchpoints closer to conversion.
Useful for: shorter sales cycles where recent interactions genuinely matter more.
Blind spot: still under-values the early activity that started the journey.
Data-driven attribution
GA4’s default. It uses your own conversion data to model how much each touchpoint actually changed the probability of conversion.
Useful for: being the most defensible model available when you have sufficient volume.
Blind spot: it is a black box, it needs meaningful conversion volume to work, and it is difficult to explain to a finance director.
Which to use
The practical answer is more than one, for different decisions:
- Judging a new awareness channel? Look at first-touch.
- Optimising a bottom-funnel campaign? Last-touch is reasonable.
- Allocating the annual budget? Data-driven, supported by incrementality testing.
Compare models against each other. If a channel looks strong on first-touch and invisible on last-touch, that tells you what job it is doing — and that is more useful than any single number.
What attribution cannot do
No model proves causation. The only way to know whether a channel is genuinely incremental is to turn it off in a controlled way and watch what happens. Attribution narrows the question; holdout tests answer it.
For businesses with long sales cycles, attribution also has to reach into the CRM. Modelling website conversions alone tells you which channels produce form fills, not which produce customers.
Where to go from here
Treat attribution models as different lenses on the same journey rather than competing claims about truth. If you want a second pair of eyes on your setup, our analytics and tracking team runs a free tracking audit that covers GA4, Google Tag Manager, conversion events, CRM integration and attribution.