Marketing Attribution Explained: Which Channel Really Gets the Credit?

Someone sees your ad on Monday, reads a blog post on Wednesday, gets a nudge from your email on Friday, and finally buys on Sunday. Come reporting time, which of those four gets the credit for the sale?

That question is the whole of marketing attribution. It sounds like accounting trivia, but it quietly decides where you spend your next budget — and if you get it wrong, you’ll cut the channel that was actually doing the work.

Here you’ll get the core problem in plain terms, the main attribution models and when each makes sense, why any single-touch answer is a bit of a lie, how UTMs feed the whole thing, and a pragmatic recommendation you can actually use.

The problem: many touches, one sale

Almost nobody buys the first time they meet you. They bump into you a few times — a post here, a search there, an email later — and somewhere in that string of touches they decide. That messy string is the customer journey.

The trouble is that a sale is one event, but it had many parents. Attribution is how you split the credit for that one sale across the touches that led to it. Get it right and you fund what works; get it wrong and you starve it.

Attribution isn’t about finding the one thing that made the sale. It’s about being honest that several things did.

This matters because the obvious answer is usually wrong. The last click before purchase looks like the hero, but it might just be the channel that happened to be standing there when a decision made weeks earlier finally landed.

The main attribution models

An attribution model is just a rule for splitting credit across touches. There’s no perfect one — each tells a different story, and knowing what each flatters is half the skill.

1. First-touch

All the credit goes to whatever introduced someone to you. It answers “what got them in the door?” — great for understanding awareness, but it ignores everything that did the convincing afterward.

2. Last-touch

All the credit goes to the final step before the sale. It’s the default in a lot of tools because it’s easy, and it answers “what closed the deal?” — but it robs every earlier touch that set the close up.

3. Linear

Every touch gets an equal share. Fair in spirit and simple to grasp, but it politely pretends the ad someone forgot mattered as much as the email that finally moved them. Equal isn’t always accurate.

4. Time-decay

Touches closer to the sale get more credit, earlier ones get less. This suits longer journeys where recent nudges genuinely matter more — though it can undervalue the awareness work that started everything.

5. Position-based

Also called U-shaped: the first and last touches get the biggest slices (often around 40% each), and the middle shares the rest. It’s a sensible compromise when discovery and closing both clearly matter and you don’t want to ignore either end.

The first two are single-touch — one winner takes all. The rest are multi-touch, spreading credit across the journey. That distinction is where most of the confusion, and most of the truth, lives.

Same journey, five verdicts

The fastest way to feel why the model matters is to run one journey through all of them. Take our four-touch buyer from the top: an Instagram ad introduces them, an organic blog post pulls them deeper, an email nudges them, and a Google search for your brand name is the last click before they buy 100 of product.

Watch how differently each model hands out that 100 of credit:

• First-touch: Instagram takes all 100 — it got them in the door, everything else is invisible
• Last-touch: the Google search takes all 100 — even though they only searched because the earlier touches already sold them
• Linear: 25 each to Instagram, the blog, the email, and search — tidy, but it pretends the forgotten ad mattered as much as the closing email
• Time-decay: search and email get the fat slices, Instagram the thin one — recent nudges win
• Position-based: Instagram and search get around 40 each for opening and closing, the blog and email split the middle 20

Same four touches, wildly different stories. Under last-touch you’d pour everything into brand search and quietly gut the Instagram ads that started it all. Under first-touch you’d do the reverse. Neither is “true” — they’re different lenses, and the whole skill is knowing which distortion you’re looking through.

View-through vs click

There’s a wrinkle the models above skip: not every touch is a click. Someone can see your ad, not click it, and still buy later because it planted the idea. Crediting that is called view-through attribution — giving an impression some credit for a sale even though nobody tapped.

It’s real, and it matters most for awareness-heavy channels like display or video, where the whole point is being seen, not clicked. But it’s also the easiest place to fool yourself — count every eyeball that later bought and suddenly every channel looks like a hero. Treat view-through as a soft signal, not proof. A click is someone choosing to act; a view is just a maybe. Weight them accordingly, and be suspicious of any report where view-through credit dwarfs the clicks.

Why single-touch models lie

First-touch and last-touch are seductive because they give you one clean answer. But a clean answer to a messy question is usually a wrong one.

Say you run ads for awareness and email to close. Judge everything by last-touch and email looks like a genius while the ads look like a waste — so you cut the ads. But the ads were filling the top of the journey; kill them and the emails soon have nobody to close. You optimised your way into a smaller business. Single-touch attribution doesn’t just simplify the story — it hands you a plan to defund the channel that was quietly doing the hardest work.

This is exactly why multi-touch models exist. They’re not more complicated for the sake of it — they’re trying to stop you from firing a channel for the crime of not being last. If two channels both matter, a model that only ever credits one of them is going to mislead you.

How UTMs feed attribution

None of this works if you can’t see the touches in the first place. That visibility comes from tagging your links, so each visit carries a label saying where it came from and which campaign it belonged to.

Those tags are UTM parameters, and they’re the raw data every attribution model chews on. Sloppy or missing tags mean your model is splitting credit across a journey it can’t fully see — garbage in, confident-looking garbage out. If that part is fuzzy, sort it first with our UTM parameters guide before you trust any attribution report.

It also helps to know where each touch sits in the bigger journey — awareness versus closing behaves very differently, which is the same map we draw in the marketing funnel guide. Attribution and the funnel are really two views of the same customer path.

The privacy reality — and why data is “modeled” now

Time for the uncomfortable truth every attribution guide should include: you can’t see the whole journey anymore, and you never fully could. Attribution leans heavily on cookies and tracking to stitch touches together — and those are getting weaker every year. Browsers block third-party cookies, people say no to tracking prompts, and someone who browses on their phone and buys on their laptop looks like two different strangers.

So there are always gaps, and platforms fill them with modeled data — statistical guesses about the touches they couldn’t directly observe. That’s not a scandal; it’s the honest state of measurement now. But it means you should read any attribution report as a well-informed estimate, not a bank statement. Two tools measuring the same campaign will disagree, sometimes a lot, precisely because they’re each filling the blanks with different assumptions.

• Don’t expect two platforms to agree — each models the gaps differently
• Trust direction and trend over any single exact number
• Keep your own first-party data clean (email sign-ups, purchases) since it’s the sturdiest thing you’ve got
• Be extra sceptical of any channel whose credit leans mostly on view-through or modeled touches

Attribution vs incrementality

One honest caveat that’ll save you from a classic trap. Attribution tells you which touches were present on the way to a sale — but present isn’t the same as responsible. Some of those buyers would’ve bought anyway, with or without that email or ad.

Incrementality asks the sharper question: how many extra sales did this channel actually cause? The cleanest way to find out is a holdout test — show a channel to one group, withhold it from a similar group, and compare. If the two groups buy at the same rate, that channel was taking credit for sales it didn’t create. You don’t need to run these constantly, but it’s worth a proper test on your biggest spend at least once, because attribution alone will happily credit a channel that was just standing near the finish line. Attribution is your everyday map; incrementality is the occasional reality check that keeps the map honest.

A pragmatic recommendation

Here’s the honest advice most tools won’t give you: don’t agonise over the perfect model. For most small and mid-sized businesses, the gap between a bad setup and a good-enough one is enormous, while the gap between two decent models is small.

• Tag every external link consistently first — no tags, no attribution worth trusting
• Start with position-based or time-decay so you value both discovery and closing
• Treat single-touch views as sanity checks, not verdicts
• Look at trends over weeks, not one purchase’s exact split
• When a model tells you to cut a channel, ask what quietly feeds it before you do

The goal isn’t a perfect number — it doesn’t exist. It’s to stop lying to yourself about which channels earn their keep, and to make budget calls with eyes open. Once you can trust the credit, connecting it back to actual returns gets far easier, which is where social media ROI picks up the thread.

How to actually pick and apply one

Choosing a model sounds like a big decision and mostly isn’t — the setup around it matters far more. Here’s the sequence that keeps it simple.

Start by looking at how your customers really buy. If most people find you and buy in the same session, a shorter-journey model like last-touch won’t mislead you much and you can keep it simple. If they typically bump into you several times over weeks — an ad, then content, then an email — reach for position-based or time-decay so the early touches don’t vanish. Set that model as the default in your analytics tool, then leave it alone for a month; flipping models weekly just gives you noise you’ll misread as insight.

Then use it to make one decision at a time. Pick a channel you’re unsure about, look at its credit under your chosen model over several weeks, and ask a plain question: if I halved the spend here, what would break upstream? Apply the model to that single call, watch what happens, and repeat. Attribution is a decision aid, not an oracle — its job is to make your next budget move a little less blind than the last one.

The short version

Marketing attribution is how you split credit for one sale across the many touches that led to it. Single-touch models — first and last — give tidy answers that quietly mislead, which is why multi-touch models like time-decay and position-based exist. Tag your links with UTMs so the data is real, pick a multi-touch model, watch trends rather than single sales, and never cut a channel without asking what it was secretly feeding.