How credit is shared between the visits
7 minute read
One order, five visits and £84.00 to share between them. Follow first click, last click, linear, time decay and position based on one journey, and see why the platforms' own numbers never add up.
One order, many visits
Nobody buys on their first visit, most of the time. A shopper sees an Instagram ad on Monday, clicks a Google ad on Wednesday, opens an email on Friday, follows a link from ChatGPT on Saturday, and orders on Sunday by typing your address straight in. That's five contacts and one order of £84.00.
Attribution is the rule that shares the £84.00 between those contacts. Different rules give different answers, and none of them is the truth. They're five reasonable ways to look at the same journey. Pick a model in the picture and watch the credit move.
Every visit shares the credit equally.
Instagram ad
Day 1
Google ad click
Day 3
Day 5
ChatGPT link
Day 6
Order placed
Day 7, direct visit
What each platform says it earned
£84.00
£84.00
£84.00
£252.00 claimed for a £84.00 sale
The £84.00 never changes. Only the split does. Forge Mark credits every order this way, and it never gives out more than the store took.
The five models
First click
All £84.00 goes to the visit that started the journey: the Instagram ad. It answers "what introduces people to us?" It ignores everything that happened after.
Last click
All £84.00 goes to the last visit before the order: the ChatGPT link. It answers "what closes sales?" It ignores everything that made the shopper ready to buy.
Linear
Every visit gets an equal share: £21.00 each. It's the fairest starting point because it makes no guess about which visit mattered. It also flatters weak visits and shortchanges strong ones.
Time decay
Visits closer to the order earn more. In the sample the four visits earn £10.08, £15.96, £25.20 and £31.92. It suits a shopper who decides quickly, where the last few contacts did the persuading.
Position based
The first and last visits earn 40% each, £33.60, and the two in the middle share 20%, £8.40 each. It treats the introduction and the close as the two moments that count.
When the models roughly agree, you can trust the picture. When they disagree a lot, that tells you something too: a channel that wins on first click and loses on last click is an introducer. One that wins on last click and loses on first is a closer.
Why platforms disagree
Meta, Google Ads and Klaviyo each report the sales they touched, using their own rules. In the sample journey all three would claim the whole £84.00. Add up their reports and you have £252.00 of sales for one £84.00 order.
Nobody is lying. Each platform sees only its own contact, and each is built to show you what it earned. But if you cut a budget on the strength of one platform's number, you're cutting on a figure that was never meant to add up with the others.
Forge Mark puts each platform's claim in the column beside its own credit. When two platforms claim the same order, you can see it. Its own total stays at what the store took.
Direct visits
A direct visit is someone who types your address, uses a bookmark or opens a link with no source. In the sample, the order comes from one. Under a plain last-click rule the direct visit would take all of the credit, and the four contacts before it would get nothing.
That's rarely fair. People often arrive "direct" because an ad or an email introduced them and they came back by themselves. So in Forge Mark a direct visit only gets credit when nothing else in the journey could. In the sample, the direct visit gets none, because four visits with a source came before it.
Consent
Attribution needs to know which visits belong to the same person. That's only fair when the visitor has agreed to it, so Forge Mark's pixel works in two tiers.
- Analytics consent lets the pixel record behaviour: pages seen, scrolling, clicks, form steps.
- Marketing consent lets it read the ad click ID, which is what ties a visit to the ad that paid for it.
In the UK and EU a visitor has to say yes to each. In Canada outside Quebec the pixel tracks visitors until they say no, or their browser sends Global Privacy Control. In Quebec, and everywhere else, it needs a yes. A visit the visitor didn't allow isn't recorded, guessed or fingerprinted.
The result is honest but incomplete. Some orders will have a short journey, and some none at all. That's the price of counting only what people agreed to, and it's why attribution shouldn't be your only lens.
What it can't tell you
- Whether an ad caused the sale. Attribution shows who was there. A shopper who'd have bought anyway still gives credit to the ad they happened to click. Only a test can separate those, and the incrementality guide shows how.
- Contacts it never saw. A friend's recommendation, a podcast, a shop window. None of it leaves a click.
- Visits nobody allowed. A visitor who declined consent leaves no trail to credit.
- The best split. The five models are rules, not measurements. They disagree because they ask different questions.
That's why Forge Mark sets attribution beside a marketing mix model and holdout tests, and lets you pick the lens.
Choosing a model
Start with linear. It treats every visit alike and makes the fewest assumptions. Then look at last click to see what closes sales, and first click to see what starts them.
If the three tell the same story, you're done. If they disagree a lot, don't argue about which model is right. Take the question to a model that reads totals, or to a test. Attribution is good at telling you where to look. It isn't good at settling whether an ad earned its spend.
Whatever you pick, keep it for a while. Switching models every week makes every channel look as if it moved when only the rule did. Compare like with like: the same model, the same window, this period against the last.
Frequently asked questions
Which attribution model should I use?
Start with linear, which treats every visit alike. Check last click to see what closes sales and first click to see what starts them. If they disagree a lot, take the question to a marketing mix model or a holdout test.
Why does Meta show more sales than Forge Mark?
Each platform counts every sale it touched, so one order can be claimed by two or three. Forge Mark shares one order's value across the visits that led to it, so the total never exceeds what your store took.
What counts as a direct visit?
A visit with no source: a typed address or a bookmark. It only earns credit when nothing else in the journey could.
Does attribution need cookies?
Forge Mark's pixel sends nothing in the UK or EU until a visitor accepts analytics cookies, and reads ad click IDs only with marketing consent. Canada outside Quebec is tracked until the visitor says no.
See your own orders credited five ways.
Connect Shopify, Google Ads and Meta in a few minutes, read only, and compare every platform's claim with Forge Mark's.