Three touchpoints, one sale. Which one closed it?

There are two ways to answer "which link made the sale." The first is to look at the last click before payment and stop there. It's fast, it's wrong half the time, and it's what most dashboards do.

The second is to admit that a sale has multiple parents, choose a sharing rule, and stick with it. That rule has a name: an attribution model. This article explains the three that matter, what each answers, and — most importantly — where each goes wrong.

The problem isn't the click, it's the handoff

A short link counts clicks. Every tool has done that for twenty years. The problem comes after.

Between the click and the dollar, there are three links in the chain, and each one breaks for a different reason:

• The click. You need one short link per placement, not one per campaign. The story link is not the bio link. Without that distinction, no attribution model will ever tell you anything: the two touchpoints are merged from the start.
• The touchpoint. The click stores an identifier. When the person provides their email, that identifier needs to travel with them. This is where the chain breaks most often, because the form submits without it: the sale then arrives without an origin, and the report shows "direct." That's the whole subject of sales tracking.
• The sale. Payment comes back, signed, from the tool that collected it. Only at that moment — not before — can you write a dollar amount next to a placement.

Break the first link and you're measuring campaigns instead of placements. Break the second, you're measuring traffic. Break the third, you're measuring clicks — and you're back to talking about engagement rates with your client because there's nothing else to show.

The three models, on the same sale

Take a $120 sale with three touchpoints before it: a LinkedIn post, a newsletter, a story. The same data, read three times — the three models are calculated simultaneously, you don't choose among them in advance.

Three columns of numbers, none of them wrong. They simply answer three different questions — which is why an agency that switches models mid-month ends up with a report that can't be compared to the previous one.

First click rewards discovery

It attributes everything to the placement that introduced someone to the brand. It's the model that values top-of-funnel content — the article, the long video, the podcast — and it's also the one that will make you over-invest in awareness if you look at it alone.

Its blind spot is mechanical: the longer the sales cycle, the more likely the first touchpoint was unremarkable. A link clicked in January for a June purchase didn't "sell" much, but the model gives it everything.

Last click is the easiest to explain

It's the default model in almost every tool, and for good reason: a client can understand it in one sentence. "That's the link that was clicked right before payment."

Its blind spot is more insidious. The last click is very often a bottom-of-funnel placement — a reminder, a follow-up, a retargeting ad — which means a placement that only reaches people already convinced by something else. Judging it by what it brings in is like congratulating the cashier for the sale.

Linear is the only one that admits reality

It splits equally among all known touchpoints. It doesn't claim that one touchpoint matters more than another — it just says that none of them did nothing.

It's the most honest model for a multi-step journey, and the most uncomfortable to present to a client: nobody wants to read that their star placement brought in a third of what they thought. It's also the only one that, over a six-month cycle, will mislead you less than the other two.

The detail that decides trust: the totals must balance

Here's the trap nobody sees coming. You split 100 cents across three touchpoints. Three times 33 makes 99. Your report then shows three lines totaling $0.99 under a total of $1.00.

One cent. Except that on one sale a day for a year, the gap shows — and the day a client adds up the lines in their report with a calculator, it's not a cent you lose, it's the credibility of the whole document.

That's why, with us, splitting happens in whole cents, never floating-point numbers, and the remainder goes to the first shares:

100 cents across 3 shares  →  [34, 33, 33]

The guarantee that follows states it in one line: whatever model you choose, the sum of the shares of a sale equals exactly the amount collected. "Attributed + unattributed = total" balances to the cent.

Same logic on currencies: revenue in euros divided by spend in dollars gives a number that looks like a result without being one. When the two don't match, the figure stays blank. A gap is visible; a false number isn't.

The attribution window: the setting nobody chooses

A touchpoint can't claim a sale indefinitely. The duration during which it keeps that right is called the attribution window, and it's the most consequential setting in the entire setup.

Too short, you lose long cycles: the sale arrives after expiration, the journey is forgotten, and your report files under "direct" what your newsletter actually sold. Too long, you credit a January post for a June purchase.

Our default is 30 days, adjustable per workspace in the tool settings. It's not a truth — it's a reasonable starting point for a product bought over a few weeks. If your cycle is six months, this setting needs to change before you look at a single number — and it's exactly the kind of setting you inherit from a tool without ever opening it.

What this looks like in practice

On our demo workspace, open to everyone without sign-up, the report shows $18,875.84 in revenue attributed to a specific placement over thirty days, spread across 121 sales, in first-click model. Without attribution, that amount exists too — but nobody knows where it came from.

The interesting point isn't the total. It's that the same screen recalculates all three models on the same data, and the placement rankings change. The link that tops the last-click ranking isn't the one that tops the linear ranking. You can check for yourself: the demo workspace is open, no account needed.

And if you'd rather query the raw data than a screen, the model is an API parameter:

GET /api/v1/analytics?model=linear

Where to start, concretely

1. One link per placement. Not per campaign, not per platform. Tuesday's story and the bio link are two different links — otherwise there's nothing to compare later.
2. Check that your form carries the click identifier. That's the link that breaks silently. The most common tools connect from their own interface, without writing code: sales tracking explains which one does what.
3. Connect the tool that collects payment. Without the sale, you have leads, not dollars.
4. Choose a model, and don't change it mid-month. Last click to start, because it's explainable; linear as soon as you have enough volume to see multi-step journeys.
5. Set your window to your actual sales cycle, not the default.

Questions that come up

"Google Analytics already does this, and it's free." It attributes sessions, not dollars. It can tell you how many people came from LinkedIn; it can't tell you which ones paid, or how much they left.

"We use UTMs and track in a spreadsheet." The most common solution, and it works — until three things happen. The spreadsheet doesn't get updated the week things are busy; UTMs disappear the moment someone copies the link without them; and nobody ever connects the "sale" row to the "click" row, because it takes half a day each month. That's actually why our UTM generator is free: that's not where the problem is.

"Without third-party cookies, attribution is dead." Cross-site advertising attribution is dead. This kind isn't: it relies on your own links, your own domain, and your own server. That's the subject of the next article.

"How long before I have usable data?" As long as one full sales cycle, no less. For a product bought in a few days, two to three weeks is enough to see trends. For a six-month cycle, the first complete numbers arrive at six months. Any faster promise is talking about sessions, not sales.

The full method, with all three models side by side and a timeline of how credit is split, is detailed on the attribution marketing page.