First-Touch vs Last-Touch Attribution: What's the Difference?

Marketing attribution is a set of rules for assigning credit for a conversion, or the revenue from it, to the marketing touchpoints that came before it. The rules are the attribution model. Change the rules and the same customer journey produces a different report.

Touchpoint

An interaction your analytics tool recorded, usually a visit with a source attached: a referrer, a campaign tag, or neither.

Acquisition source

Where a visit came from: a search engine, a social site, a newsletter, an ad, another website, or Direct.

Conversion

The outcome being credited: a signup, a trial, a payment.

Attribution model

The rule that decides which touchpoints get credit for a conversion, and how much each gets.

Attribution window

How far before the conversion a touchpoint can still count. Some tools let you set it; in others it's limited by how long the tool recognizes a visitor.

Direct traffic

Visits with no referrer and no campaign tag: typed addresses and bookmarks, but also links from apps, documents and email clients that don't pass a referrer.

A model can only choose among the touchpoints the tool recorded and connected to the same person. An ad someone saw but never clicked, a podcast mention, a recommendation from a friend, or a visit on another device usually isn't in the data at all. Attribution reconstructs the recorded part of a journey, not the whole journey.

First-touch attribution, also called first-click or first-interaction attribution, gives 100% of the credit to the earliest touchpoint recognized for the customer. Everything after it gets nothing.

Where it helps

First-touch answers "which channels introduce us to the people who later pay?" That suits discovery: search content, communities, launches and partnerships often start journeys without being there at the end. It's also stable: the first touch doesn't move while the customer keeps visiting.

Where it falls short

• It ignores everything after the first visit: the comparison page, the reminder email, the demo that convinced them.
• It depends on identity. "First recognized" isn't "first ever": if the visitor first came on their phone or cleared cookies, the first recorded visit is a later one.
• Long sales cycles make it harder. The longer between discovery and purchase, the more likely the identifier expired or the device changed in between.

Last-touch attribution, also called last-click or last-interaction attribution, gives 100% of the credit to the final touchpoint recognized before the conversion. In the same example, the newsletter receives the $99, if it was the last recorded visit before the payment.

Where it helps

Last-touch answers "what brings people back at the moment they buy?" It's close in time to the purchase, so it depends less on connecting visits weeks apart, and it's useful for comparing conversion-stage channels such as email to existing leads, branded search or retargeting.

Where it falls short

• It ignores how the customer found you. A channel that only introduces people looks worthless.
• It favors channels that reach people who have already decided, which can flatter branded search, email and Direct.
• Its result depends on the product's rules, above all on how it treats direct visits.

Tools don't agree on Direct. Some count a direct visit as a valid last touch. Others skip it and credit the most recent visit with a known source. Check which rule your tool uses before you compare last-touch numbers between tools.

Direct means the analytics tool saw no referrer and no campaign tag. It's a label for missing information, not a channel anyone chose. Google Analytics 4, for example, defines its Direct channel as source "(direct)" with medium "(none)" or "(not set)".

Returning visitors often arrive this way. They type the address, use a bookmark or browser autocomplete, or click a link in a chat app, a PDF or an email client that strips the referrer. Someone who found you on Reddit on Monday may well come back as Direct on Friday.

If a tool treats Direct as a touchpoint like any other, and that return visit is the last one, last-touch gives Direct the credit. That's why some tools use last non-direct touch: direct visits are skipped and the most recent visit with a known source gets the credit, unless every visit was direct. GA4 states that its attribution models exclude direct visits from credit unless the whole path is direct. The aim is to stop a visit with no source from overwriting a known one; the cost is that a genuine direct return, from a customer who already knew you, never shows as such.

Take a fictional SaaS month: 100 visitors from Reddit, 50 from Google and 20 from a newsletter, some of whom come back several times. Six customers pay $49 each through a supported payment provider, and each payment is matched to the customer's visits. A seventh $49 payment can't be matched to any visitor.

Under first-touch, Reddit looks like the strongest channel: $147 from 100 visitors. Under last-touch it brings $49, and Direct shows $98 even though nobody in this example discovered the product directly. Google is $98 either way, because it started two journeys and ended two others. The unmatched payment is unattributed under both.

Changing the attribution model redistributes credit. It does not create additional revenue.

SuprAnalytics connects payments to the visits before them, so every source, campaign, page and country report can show revenue. This is how it decides which visit gets the credit:

• Two models: first touch and last touch. Last touch is the default for every new website.
• A setting per website, in Settings → Revenue → Attribution model. It applies to the website's revenue reports: the chart, the payments list, and revenue by channel, referrer, campaign, page or country.
• First touch credits the visitor's earliest recorded visit.
• Last touch credits the visit the payment belongs to when the checkout passes its session id, otherwise the latest visit that started before the payment. Direct visits count.
• Visits are sessions. A session ends after 30 minutes without activity, and its source is decided when it starts. A utm_source or ref tag on the landing page takes priority over the referrer.
• History follows the setting. Each matched payment stores both its first and its last visit, and the model is applied when a report is read, so switching re-credits past payments at once, in either direction. Nothing is reprocessed, and nothing missing is recovered: a payment that was never matched to a visitor stays without a source under both models.
• Lifetime value is always first touch. The Lifetime value page groups customers by the visit that first brought them, whatever the setting.
• Recognition has limits. The visitor id lives in a first-party cookie for a year after the last visit. In cookieless mode a visitor is recognized for a day. A phone and a laptop are two visitors; devices aren't merged.

To compare both views without changing the setting, the MCP tools accept a model argument, first or last. The revenue attribution docs cover the setup, and subscription revenue explains how renewals are credited.

Stripe knows the customer and the amount. It doesn't know which marketing source sent them. That part comes from your analytics:

1. 1A visitor arrives through a tracked source. The script records the visit, its source, and gives the visitor an id.
2. 2They browse, perhaps leave and return, perhaps several times.
3. 3They pay through Stripe Checkout or a Payment Link.
4. 4Stripe sends the payment to SuprAnalytics through a webhook. Polar, Paddle, Lemon Squeezy and Dodo Payments work the same way; other providers can use the Payment API.
5. 5SuprAnalytics matches the payment to a visitor: the visitor id passed to checkout, else the same Stripe customer id on an earlier matched payment, else the same email.
6. 6The website's attribution model decides which of that visitor's visits receives the revenue.

An unmatched payment can't be assigned to a source under any model. It still counts in total revenue, without a source, so the sum of the source rows can be lower than the total. Setup is in the Stripe guide; what you get is on the Stripe revenue attribution page.

First-touch may be useful when you are

• Evaluating discovery channels
• Understanding top-of-funnel acquisition
• Measuring which channels reach new people at all

Last-touch may be useful when you are

• Evaluating the final recognized interaction before purchase
• Understanding conversion-stage channels
• Comparing traffic sources close to the purchase

Many founders look at both. Where the two agree, the signal is easier to trust. Where they disagree, the gap is informative: a channel that is strong under first-touch and weak under last-touch is probably starting journeys that other channels finish. Tag your own links with UTM parameters so the visits you can label aren't lost to Direct, and read individual visitor journeys to see the steps no single-touch model shows. None of this guarantees better decisions; it gives you two consistent views to reason with.

• One touchpoint receives all the credit; every other one receives none.
• Earlier or later interactions are ignored, depending on the model.
• Identity matching can fail: a visitor who isn't recognized starts a new journey.
• Cross-device journeys are incomplete when devices aren't linked.
• Cookieless tracking can shorten how long a visitor is recognized.
• Direct traffic can hide the earlier source that actually introduced the customer.
• Attribution doesn't measure incrementality: whether the sale would have happened without the channel.

Incrementality needs an experiment, such as holding back part of an audience and comparing the two groups. Ad platforms offer lift studies for this because attribution rules can't answer it.

Multi-touch models, briefly

The wider industry uses models that split credit across several touchpoints. These are general concepts; SuprAnalytics doesn't offer them.

Linear

Every recorded touchpoint gets an equal share.

Time-decay

Touchpoints closer to the conversion get a larger share.

Position-based

The first and last touchpoints get the largest shares (often 40% each) and the middle ones share the rest. Also called U-shaped.

Data-driven

Shares are estimated statistically from converting and non-converting paths. It needs a lot of data and is a model, not a measurement.

Tools change here too: Google retired first click, linear, time-decay and position-based models from Google Ads and GA4 in 2023, leaving data-driven and last-click variants.

Treating attributed revenue as incremental

"$5,000 attributed to Reddit" doesn't mean you'd lose $5,000 without Reddit.

Ignoring Direct

A large Direct share under last-touch often means returning customers whose earlier source is hidden. Look at first-touch before cutting a channel.

Comparing models with different definitions

Different tools, windows and Direct rules produce different numbers for the same data. Compare like with like.

Assuming cross-device recognition

A visitor on a phone and on a laptop is usually two visitors.

Treating first-touch as first exposure

It's the first recorded visit, not the first time someone heard of you.

Treating last-touch as the final influence

A review, a friend's message or a sales call may have come after the last tracked visit.

Forgetting unattributed payments

Revenue by source adds up to matched revenue, which can be less than total revenue.

Confusing goals with revenue

A signup or trial goal shows intent; only payment data shows revenue. Keep the two reports apart.

Related: revenue attribution · SuprAnalytics vs Google Analytics · pricing