Skip to main content
Revenue7 min read

First-touch vs last-touch attribution for a small SaaS

First touch credits the source that found the customer, last touch the one that was there when they paid. What each gets wrong, what renewals and long trials do to both, and the rule Vero uses.

Vicente TorresFounder of Vero

Every revenue-by-source table rests on a rule that decides which touch gets the sale. Most tools pick one for you and never say which. For a small SaaS the choice matters more than it looks: you have few sales, trials that run for weeks and renewals that arrive with no visit at all, and each model reads those differently. This post explains what first touch and last touch credit, where each one misleads, what changes once subscriptions renew and trials run long, and which rule Vero uses.

What does first-touch attribution credit?

First touch gives the whole sale to the source of the customer’s first recorded visit. A developer finds you through a forum thread in March, comes back four times through search and buys in May: the forum thread gets the sale.

It answers a discovery question: which channels bring people who did not know you existed. For a small SaaS that is usually the question that costs the most to get wrong, because finding new buyers is the expensive part.

It needs a memory. To know the first visit of a May buyer, something has to tie March to May: a cookie, a login, an email address. Without one, first can only mean the first visit the tool can still connect to the sale. The overview of revenue attribution covers how that connection is made at all.

What does last-touch attribution credit?

Last touch gives the whole sale to the source of the visit right before the purchase. In the same story, search gets the sale, because the customer came back through search on the day they paid.

It answers a closing question: which channels are there when people decide. It is also the easier one to compute, because it only needs the visit that ended in the payment.

Taken literally, it breaks on any site with a checkout. The last page before the payment is your own pricing page, and its referrer is your own site. Read that way, nearly every sale goes to Direct. A useful last-touch model has to skip your own pages and look at how the last visit began.

Where each model misleads

First touch over-credits discovery. A channel that brings many curious visitors scores well even when a different channel did the convincing. It also ages badly: a sale today is judged by a source from months ago, under a campaign you may have stopped.

Last touch over-credits the channels people use to come back. A search for your name, a bookmark, your newsletter to people who already signed up and the link in a trial reminder email all sit close to the payment, and none of them found the customer. Cut the forum thread because last touch never credits it, and a few months later fewer people search for you by name.

Neither model splits credit. A journey through four sources ends with one of them holding the whole sale. Multi-touch models spread it, but they need every visit of the journey tied to one person, and with a few dozen sales a month the split rests on very few journeys.

What changes with renewals?

A renewal has no touch at all. Stripe bills month six without anyone visiting your site, so the payment arrives with no visit. Last touch has nothing to credit, and counted that way, recurring revenue piles up under no source or under Direct. The channels that bring customers who stay for years then look exactly like the ones that bring customers who leave after a month.

For a subscription, the rule that holds up is that a renewal keeps the source of the customer’s first payment. The source that won the customer owns the money the customer keeps paying, and a refund of a renewal comes off that same source.

In the Vero demo, a fictional SaaS with simulated numbers, 6 of the week’s 187 sales, worth €412, are automatic renewals paid with no visit. On a subscription business that has run for a while, renewals can be most of what you bill, so the rule for renewals ends up deciding more of the table than the rule for first purchases.

What changes with long trials?

A trial splits the decision from the payment. The customer starts the trial in one visit and pays when it ends, often with no visit that day, because the first paid invoice is billed on its own. Last touch credits that payment to nothing. First touch credits it correctly only if something tied the trial’s visit to the later invoice.

With Stripe, the tie is the subscription’s metadata. Put the visit id in subscription_data.metadata when the trial checkout is created, and every invoice of that subscription carries it, the first paid one included. How to pass UTM parameters and the visit to Stripe Checkout shows the fields.

Trials also stretch the wait you are measuring. Time to pay, a Pro card in Vero, shows the typical wait from the first touch of the visit that brought a customer, or of the trial their checkout points to, to their first payment. It names a typical wait once the period has 5 such sales. A long wait tells you that one visit rarely holds the whole decision for your product.

Which attribution model does Vero use?

Vero credits each sale to the first touch of the visit in which the customer paid. The payment carries a session id, that id names a visitor, and the visitor plus the gap of inactivity names the visit. The first event of that visit carries the source: its utm_source, else the host of its referrer, else direct. A referrer on your own site counts as none, so the checkout page never takes the credit. It is the same rule the traffic reports use, so money and traffic agree on what a source is.

Across visits, that makes it a model of the closing visit. Vero calls it first touch because it reads the first event of the visit that paid. A customer who arrived from a forum thread in the morning and came back through search in the afternoon to pay is credited to search in revenue by source.

Sales without a visit id, renewals among them, inherit the source of the customer’s first attributed payment. A trial whose checkout carried the visit is credited to that visit when it converts. Each sale stores its source when it is received, so it keeps it whatever date range you read.

The finding side has its own card. Who found the customer and who closed the sale, on Pro, puts the source that found each customer beside the one that closed the sale. A first sale goes back to the first visit of the same UTC day from that visitor, and a later sale of the same customer goes back to their first sale. It shows once the period has 5 such sales. In the fictional demo, 22% of the sales were found by one source and closed by another, and 68% were paid on the first visit.

The limit is the one a cookie-free design brings. The visitor hash rotates daily, so a visitor who came back on another day before buying cannot be linked, and the card says so: it never shows the full journey. How to attribute revenue without cookies walks through the mechanics and the limits.

Which model should a small SaaS use?

Pick the model whose mistakes you can live with, then read it knowing them. With few sales, a single-touch model you can explain beats a fractional one you cannot check.

If most of your sales are paid on the first visit, first touch and last touch give the same answer and the debate does not matter. If a large share is found by one source and closed by another, read revenue by source as the closer and the found-and-closed card as the finder before you cut a channel. If your time to pay runs to weeks, expect channels that open journeys to read low, and judge them over a month, not a launch.

Whatever the model, tag every link you place yourself with a utm_source, so a return visit from your newsletter reads as the newsletter and not as Direct. Then subtract what each source cost before you move money between them.

Unsure of a term? Read the glossary

Vicente Torres
Vero is built by one person in Spain. I read every waitlist entry and pick the beta accounts myself.
Vicente Torres, Founder of Vero LinkedIn
Hand-picked beta, waitlist open

Ask for a place in the beta.

Anyone can join the list. The accounts we pick get Pro free for the whole beta.

By joining you agree to receive launch updates and accept our Privacy Policy.

See the demo