Attribution Models in Plain English

One buyer, one sale, three touchpoints. She sees your Instagram ad on Tuesday and scrolls past, but it registers. Thursday she searches for the product, clicks your Google ad, pokes around, leaves. Sunday morning your newsletter lands, she clicks the link, and buys.

So which channel gets the credit?

Ask your ad platform, your email tool, and your analytics dashboard, and you'll get three confident, contradictory answers. Instagram says it started everything. Google says it captured the intent. The newsletter says it closed the deal. None of them are lying. They're just using different rules for handing out credit.

An attribution model isn't a measurement. It's a policy, and that's the part most explanations bury. There is no objective fact about which channel "caused" the sale, the same way there's no objective fact about which ingredient made the soup good. You pick a rule for splitting credit, apply it consistently, and hope the rule matches the questions you're actually asking.

Which means every model is a story about the same event. First-click tells the discovery story. Last-click tells the closing story. Multi-touch tries to tell the whole story and pays for it in complexity. We'll walk that one Sunday sale through each model and watch the numbers change while the reality doesn't, which is honestly the fastest way to understand what your dashboard is really telling you.

Attribution sits on top of the plumbing, so if you're still fuzzy on how a click even gets recorded, start with our complete link tracking guide and come back. The models will make a lot more sense.

Last-click: the default everyone argues with

Last-click gives all the credit to the final touch before the sale. In our Tuesday-Thursday-Sunday journey, the newsletter gets 100%. Instagram and Google get nothing. As far as your report is concerned, they were never in the room.

Why is this the default almost everywhere? Because the last click is the one thing you know for sure. It's the click that actually landed on the checkout page, in the same session as the purchase, with no guesswork about whether it "influenced" anything. Every other touch requires you to have tracked that same person days earlier and connected the dots.

The last click connects itself.

When analytics tools had to pick one model that works with zero setup, this was the obvious choice, and it stuck. And to be fair, it does answer the question "what closes?" honestly. If you want to know which channel is best at catching people who are already ready to buy, last-click will tell you.

What it quietly punishes is discovery. The channels that introduce you to people who've never heard of you almost never get the last click. Nobody sees your Instagram ad and buys in the same breath. They see it, forget about it, get nudged twice more, and buy later from whatever channel happened to be nearby at the moment of decision. Under last-click, that nearby channel looks like a hero and the ad that started everything looks like a money pit.

Which is how teams end up cutting the top of their funnel to fund the bottom of it. Run last-click long enough, shift budget toward whatever it rewards, and your "winning" channels slowly run out of new people to close. The report stays green right up until it doesn't.

Last-click isn't wrong. It's just answering one question while pretending to answer all of them.

First-click: crediting discovery

First-click is last-click's mirror image. Run the whole journey back to the beginning and hand all the credit to the first touch. In our buyer's case, that Tuesday Instagram ad gets the entire sale. The Google ad and the newsletter did real work, and they get nothing.

That sounds just as unfair as last-click, and it is. But it's unfair in the opposite direction, which makes it useful.

First-click is good at answering where new people actually come from. If you're trying to figure out which channels introduce your brand to strangers, last-click is nearly useless, because strangers rarely buy on the first touch. First-click shines a light on exactly the channels last-click leaves in the dark. That paid social campaign that "never converts anyone"? Under first-click, it might turn out to be the thing starting half your journeys. Same data, different scoreboard.

In everyday terms, first-click credits whoever made the introduction at the party, and last-click credits whoever was standing there when the handshake happened. Both matter. Neither is the whole story.

Where first-click misleads is judging the channels that close. Your newsletter could be quietly sealing every deal and it would score zero, because by the time someone's on your list, some other channel already claimed them. Optimize purely on first-click and you'll starve the bottom of your funnel while pouring money into the top.

So treat first-click as a discovery lens, not a performance lens. Use it to ask "what fills the funnel," and never to decide which closing channel to cut.

Multi-touch: splitting the credit

Multi-touch models start from a simple observation. If three channels touched the sale, giving all the credit to one of them is a choice, not an insight. So instead of picking a winner, you split the credit. The question becomes how.

Linear is the fair-split model. Every touch gets an equal share. Three touches, a third each. It's the "everyone chipped in for the pizza" approach, and its bias is exactly that fairness, because it treats a half-remembered Instagram ad and the newsletter link that closed the sale as identical contributions.

Time-decay says the closer a touch is to the purchase, the more it mattered. Sunday's newsletter click gets the biggest slice, Tuesday's Instagram ad the smallest. This is the model for teams who believe momentum matters more than introductions, which is a defensible opinion, not a law of nature.

Position-based (often called U-shaped) plays favorites at both ends, giving typically 40% to the first touch, 40% to the last, and spreading the remaining 20% across everything in between. The logic is that finding the customer and closing the customer are the hard parts, and the middle is supporting cast.

Here's our buyer from the top of this guide scored under each:

Model	Instagram ad (Tue)	Google ad (Thu)	Newsletter (Sun)
Linear	33%	33%	33%
Time-decay	15%	30%	55%
Position-based	40%	20%	40%

Same sale, same three clicks, three different reports. Under time-decay, the Instagram ad looks nearly worthless. Under position-based, it's tied for MVP. If your team is deciding budget off one of these tables without knowing which model produced it, you're arguing about numbers that were opinions before they were numbers.

Every model in that table also assumes you actually saw all three touches. Miss one and the math doesn't degrade gracefully — the credit just flows silently to the touches you did capture, and your linear model quietly becomes a "whatever we happened to track" model. Multi-touch is only as good as your touch coverage.

Which is where the plumbing comes in. You need click IDs that connect touches to the same person across days, so Tuesday's ad and Sunday's newsletter click are recognized as one journey instead of three strangers.

And you need consistent UTM tagging on every link you send out, so each touch arrives labeled with a channel name your reports can group on. One untagged newsletter link and Sunday's touch shows up as "direct," and suddenly your closing channel is invisible in every model you run.

Multi-touch isn't smarter than single-touch. It's greedier about data. Feed it complete journeys and it tells you a richer story; feed it gaps and it makes them up.

Picking a model without overthinking it

Teams agonize over this choice like it's permanent. It isn't. The model you pick matters far less than using the same one everywhere and knowing which direction it lies.

Every model lies in a predictable direction. Last-click flatters your closers, first-click flatters your discovery channels, multi-touch spreads flattery around evenly. None of that is a problem as long as you know the tilt. It's like a bathroom scale that reads two pounds heavy. Useless for your absolute weight, perfectly fine for tracking whether the number is going up or down. Attribution works the same way. Compare campaigns under one consistent model and the bias mostly cancels out.

If you're a small team running two or three channels, use last-click and move on. Your journeys are short, your data is sparse, and multi-touch on sparse data invents stories. Just remember your top-of-funnel work is being undercounted, and don't kill an awareness channel because last-click says it "never converts."

If you're heavy on paid acquisition across several channels, add first-click as a second lens. Not a replacement, a comparison. When a channel looks great under first-click and terrible under last-click, that's a discovery channel doing its actual job.

If you're running ten-plus campaigns with long consideration cycles, multi-touch starts earning its complexity. But only if your tagging is airtight, because a multi-touch model with holes in it is worse than a simple model without them.

The thing that actually breaks attribution at every team size isn't the model. It's inconsistency, half the links tagged newsletter, half tagged email, a few tagged nothing. That's why we built campaign grouping with UTM management into Acturity, so every link in a campaign carries the same tags and campaigns are comparable side by side without a spreadsheet cleanup first.

Pick a model, write down its bias somewhere your team will see it, and spend the energy you saved on consistent tagging. That trade wins every time.