# Attribution models and lookback

> Five ways to split credit for a sale between the touches that led to it, and how far back Statsy looks for them.

Section: Revenue attribution  
Canonical page: https://statsy.co/docs/revenue/attribution-models  
Last updated: 2026-09-15

Almost nobody buys on the first visit. Somebody reads a post, forgets about you, searches your name a week later, and buys from an email a fortnight after that. An attribution model is the rule for deciding which of those three touches earned the sale.

Statsy applies the model when you look, not when the payment arrives, so switching models re-reads history instead of starting a new record. Both controls sit in the header of the **Revenue** page.

## The five models

| Model | How credit is split | Good for |
| --- | --- | --- |
| Last touch | All of it to the last touch before the payment. | The default. Answers "what closed the sale?" |
| First touch | All of it to the first touch in the window. | Answers "what found this customer?" — the top of the funnel. |
| Linear | An equal share to every touch. | A fair, unopinionated split when you have no theory. |
| Time decay | More to touches nearer the payment: weight halves every 7 days. | Long sales cycles where recent activity matters more but early work still counts. |
| Position based | 40% to the first touch, 40% to the last, the remaining 20% shared by everything in between. | The compromise most teams settle on: finding and closing both get credit. |

Every model normalises to 100% of the payment, so totals agree whichever you pick. With one touch it gets everything; with two, position based splits it evenly rather than giving 80% away and losing the rest.

## The lookback window

The lookback decides how far back before a payment a touch may sit and still receive credit. The selector offers 30, 60, 90, 180 and 365 days, with **90 the default**; through the API any whole number from 1 to 365 works.

Match it to how long people actually take to buy. A short window on a slow enterprise sale throws away the touch that started it; a 365-day window on an impulse purchase credits a blog post somebody read last spring.

## Why direct is quieter than you expect

Direct traffic — no referrer, no campaign — earns credit only when a payment has no other source anywhere in the window. If the buyer ever arrived from a search, a post or an email, that is what gets the credit, and the direct return visits before the purchase do not dilute it. This is deliberate: "they typed the URL" is not a marketing channel you can spend more on.

## What the report shows

Revenue by channel gives you, per channel: visitors, customers, conversions, revenue, refunds, net revenue and ARPU (net revenue per visitor).

> **Note.** The attribution report reads payments directly, so dashboard filters and segments do not apply to it, and unmatched payments are not in it. The revenue figures on the overview page do include them, which is why the two can differ.

The same report is available through `GET /api/v1/sites/<site>/attribution?model=&lookback=`, from the CLI with `statsy attribution`, and from an AI assistant over MCP.

Previous: [Record payments by hand](https://statsy.co/docs/revenue/manual-payments)  
Next: [Payments not showing up](https://statsy.co/docs/revenue/troubleshooting)