First, last, linear, position, time decay: what each attribution model answers

Budget meetings often include an argument like this:

• The content lead: "Most customers first heard of us through a blog post they found in search."
• The paid media lead: "But the last thing they clicked before buying was our ad."

They may both be right. They're looking at the same data through different attribution models.

Three journeys

Say three customers paid last month (example):

Three conversions, three units of credit. Here is how five models share them out.

First touch: who brought them in

All the credit goes to the first visit before the conversion.

It answers: where did new customers first come across us? Good for judging brand, content and SEO — channels that bring people in. Its weakness: it ignores everything that nudged them later.

Last touch: who closed the deal

All the credit goes to the last visit before the conversion.

It answers: what brought them back right before they converted? Many ad platforms work this way by default, which is why paid teams' numbers lean towards last touch. The problem is obvious: content and social get nothing, yet without them A and B might never have shown up.

Linear: everyone gets a share

Credit is split equally across every visit.

• A: organic search, email and paid ad ⅓ each
• B: social and paid ad ½ each
• C: paid ad 1

It answers: which channels keep turning up along the way? Fair, but it treats an email opened in passing the same as the visit that made up someone's mind.

Position: the ends matter most, the middle counts

One touch gets everything; two share half each; with three or more, the first gets 40%, the last 40%, and those in between split the remaining 20%.

• A: organic search 0.4, email 0.2, paid ad 0.4
• B: social 0.5, paid ad 0.5
• C: paid ad 1

It accepts that bringing people in and closing are both crucial, while leaving something for the nurturing in between. For most businesses with a real decision cycle, it's a sensible place to start.

Time decay: the closer, the more it counts

Visits closer to the conversion weigh more, halving with each half-life. With a 7-day half-life, a visit 7 days before converting weighs half as much as one on the day; 14 days before, a quarter.

It answers: what's driving conversions right now? Suits promotions and short decision cycles. In long B2B sales it badly undervalues the channel that brought in the lead.

So which one?

Our advice: don't pick one. Look at two or three side by side.

A practical setup:

1. First touch to judge acquisition channels — content, SEO, brand campaigns.
2. Last touch to judge conversion channels — retargeting, promotional email, brand-term search ads.
3. Position as the reference for overall budget split.

A channel near the top in all three is very probably good. A channel that only looks good under last touch, and barely registers under first touch, looks more like it's harvesting demand other channels created. That doesn't make it useless — but cut the other channels and its numbers will fall too.

Two settings people overlook

Lookback window

Only visits within so many days before the conversion count. 30 days by default, adjustable from 1 to 90. Consumer products with short decisions can go shorter; B2B can go to 60 or 90. A conversion with no visit inside the window is "unattributed" — we don't guess a source for you.

Which dimension a touch uses

Channel by default, or UTM source, UTM campaign, referring domain or landing page. Attribution by campaign tells you something more specific than by channel: not "is paid social good?" but "which campaign was good?"

Finally: attribution isn't causation

Every attribution model shares out credit; none proves cause. "Paid ads earn 3 units under last touch" doesn't mean switching the ads off loses three customers — some of them might have come back through a brand search anyway.

To learn how much a channel really adds, you need an experiment — pausing spend in some regions and comparing, for instance. Attribution models are fine for day-to-day direction; for a large budget decision, run that experiment once.

Related articles

Marketing attribution·Sep 4, 2026·4 min read

A UTM naming convention your whole team will follow

Three rows called facebook, Facebook and fb aren't a tool problem, they're a naming problem. Five rules and one sheet that make everyone's links come out the same.

Read article

Product notes·Oct 16, 2026·3 min read

Why bot traffic gets its own report

By default we take identified bots out of your human reports rather than dropping them or mixing them in. The three reasons behind that, and what it means for your numbers.

Read article

Playbooks·Oct 14, 2026·5 min read

Checkout drop-off: a seven-step investigation from data to fix

Plenty of add-to-carts, not many payments. A checklist to work through in order — rule out technical problems first, then find which step, which visitors and what cause.

Read article