Guide
How to tell which marketing channel brings paying customers
Every analytics tool can tell you which channel sent the most people. The ranking that decides where your next month goes is a different one, and almost nobody has it on screen.
Updated
The two rankings almost never match
Sort your channels by visitors and you get one order. Sort the same month by the money those visitors paid you and you get another, and the top of the second list is usually something that was fourth or fifth in the first.
This is not a quirk. Channels differ in intent far more than they differ in volume. A link in a newsletter somebody chose to subscribe to and a link from an aggregator that pushed your post to a front page can send the same thousand people, and one of those thousands buys at ten times the rate of the other. Ranked by traffic they look alike. Ranked by revenue they are not the same business.
So the question is not "where did people come from". It is "where did the people who paid come from", and answering it means carrying the source of a visit all the way to a charge in a payment processor that has never heard of referrers.
The useful ranking is revenue per visitor by channel. It is the only one that tells you what another thousand visits from that source is worth.
Why the join is the hard part
A visit lives in your analytics. A payment lives in Stripe, Polar, Creem, Dodo or whatever else takes the card. Nothing connects them by default, because the browser that read your article and the server that received the webhook are talking about the same person in two vocabularies that have no word in common.
The usual workarounds are worse than they look. Asking "how did you hear about us" at checkout collects the answer people remember rather than the one that happened. Reading the referrer at the moment of purchase credits the checkout page, or Google, or nothing, because by then the tweet that actually sold it was three weeks ago. Conversion pixels tell an ad platform what it wants to hear and cannot be read back out for a channel that is not an ad.
The join that works carries an identifier the visitor already has into the checkout, so the payment arrives knowing which visit it belongs to. Everything else is reconstruction after the fact.
What this product does about it
A visitor gets an id on their first pageview. When they start a checkout, that id travels with it, and when the payment webhook arrives it carries the id back. The payment is written against the visit that produced it, and inherits that visit's source, campaign, landing page and country.
When the id does not make it through, which happens with payment links somebody pastes into an email and with checkouts that strip metadata, there are four more signals before giving up: the return page the buyer lands on after paying, an earlier payment from the same customer, an email that matches somebody you identified, and finally the timing of who clicked a checkout link shortly before the charge. The last one only ever picks a visitor when the evidence points at exactly one person.
Credit goes to the last visit that was not direct. Somebody who read a post from Reddit, came back by typing your domain three weeks later and paid is credited to Reddit, because that is what produced the sale. If you would rather credit the first touch, that is a per site switch, and the panel says which model produced the number it is showing.
- Revenue, customers and revenue per visitor beside every channel, referrer, campaign and keyword
- The same split for landing pages, countries, devices and search queries
- A report that says why each payment is attached to what it is attached to, and which change would have matched the ones that are not
What it cannot tell you, said plainly
No attribution model is correct. A person who saw a tweet, read a comparison, asked an assistant and then bought was influenced by all four, and any single number that credits one of them is a simplification you are choosing on purpose. Last touch is the one most people mean when they ask this question, which is why it is the default here, not because it is true.
Revenue per search query is an apportionment, not a measurement: Search Console reports queries and never ties one to a person, so a query is credited with a share of the revenue of the visits that arrived on its page from Google. It is marked as an estimate everywhere it appears, and it is useful for ranking queries against each other rather than for adding up.
The dashboard this is describing
Live demo, real queries, generated traffic. No sign-up.
- Organic Social2,412$3,120
- Direct1,072$1,240
- Referral771$960
- Organic Search720$580
- AI246$310
- 🇺🇸United States885$2,410
- 🇮🇳India887$140
- 🇬🇧United Kingdom427$980
- 🇩🇪Germany427$760
- 🇧🇷Brazil451$210
Questions
Can I do this with Google Analytics?
Partly, and only with work. GA4 can receive a purchase event with a value, so a channel report with revenue in it is possible if you send one, and many sites do. What it will not give you is the payment-level history: which person paid, what they read before they did, whether they refunded last week, and what a renewal three months later is worth to the channel that brought them. Those are the questions this product is built around, and they need the payment itself rather than an event that says a payment happened.
What about people who block analytics?
They are not counted, here or anywhere. The honest consequence is that every channel is undercounted, and the ones with a technical audience are undercounted most. Serving the script from your own domain, which takes one rewrite rule, recovers a large part of it. We say the same thing on every page that quotes a number: this is what was measured, not what happened.
How long before I can act on the numbers?
As long as it takes you to get enough payments for a channel to mean something. Five payments from a source is an anecdote. Twenty is a decision. The traffic side is useful from the first day, the revenue side needs a month of ordinary selling for most sites.
