First-touch vs last-touch attribution

Most buyers touch more than one channel before they pay. An attribution model is the rule that decides which of those touches gets credit for the sale. The two simplest models, first-touch and last-touch, often tell opposite stories about the same customers, so it is worth knowing what each one rewards.

One customer, three touches

Consider a single buyer:

1. Day 1: they click a link in a Reddit thread and read your landing page.
2. Day 4: they search for your product name on Google and look at pricing.
3. Day 9: they click a link in your newsletter and buy a $49 plan.

First-touch gives all $49 to Reddit, because Reddit introduced them. Last-touch gives all $49 to the newsletter, because it was the final click before the payment. Both are “correct”; they answer different questions.

What first-touch is good at

First-touch measures discovery. It tells you which channels put new people in front of you, which is exactly what you want to know when you are deciding where to find an audience. Its blind spot is everything that happens after the introduction: a channel that reliably closes deals, such as email or a retargeting campaign, gets no credit at all.

What last-touch is good at

Last-touch measures what tipped the decision. It is simple, it matches what most people intuitively expect, and it reacts quickly when you change a campaign. Its blind spot is the opposite one: it over-credits channels that sit at the end of the journey. Branded search is the classic example, since people often search your name right before buying, even though something else made them aware of you.

Last non-direct touch

A common refinement is to ignore visits that carry no source. If the buyer in our example had typed your address straight into the browser on day 10 and bought then, pure last-touch would credit “direct”, which tells you nothing. Last non-direct touch skips that visit and credits the newsletter instead. Typing your address is the result of earlier marketing, not a channel of its own.

Multi-touch models

Multi-touch models split the credit across several touches:

• Linear: every touch gets an equal share.
• Time-decay: touches closer to the sale get a bigger share.
• Position-based (U-shaped): the first and last touch get most of the credit, the middle ones split the rest.
• Data-driven: an algorithm estimates each touch’s contribution from many journeys.

They sound more accurate, and at large volumes they can be. For a small business they mostly add false precision: the weights are a choice, the data behind them is thin, and a fractional “0.33 of a sale from Reddit” is harder to act on than a clear answer plus a look at the journey.

The lookback window matters as much as the model

Every model needs a limit on how far back it looks. With a 7-day window, our day-1 Reddit visit would not count at all; with 30 days it would. Short windows favor fast, impulsive channels. Long windows favor slow, research-heavy ones, but they also credit touches that may have had nothing to do with the decision. Match the window to how long your customers typically take to buy, and keep it the same when you compare periods.

Which model should you use?

• If you are trying to find new audiences, look at where customers first came from.
• If you are trying to improve campaigns that convert, use last non-direct touch.
• Either way, read individual journeys for your best customers. Ten real paths teach you more than a weighting formula.

How theBizness.ai credits a sale

theBizness.ai uses last non-direct touch with a 30-day window. Each sale is credited to the source the visitor most recently arrived with, through a referrer or UTM tag, in the 30 days before the purchase. A visit without a source never overrides a real one, and a buyer returning from a checkout or sign-in page keeps the channel that brought them in.

Subscription renewals have no visit of their own, so they are credited to the source of the customer’s earliest attributed purchase. That keeps recurring revenue with the channel that won the customer. To see the discovery side, open a customer’s path to purchase, which lists their visits from the first source to the sale. The exact rules, including the fallbacks when a visit has no source, are in the revenue attribution docs.