Guide 177 · Conversions · GA4 + Tag Manager
Key event attribution in GA4: data-driven credit, and the 90-day lookback window
Attribution is how GA4 divides credit for a key event between the touchpoints before it. The property default is data-driven attribution, and the key event lookback window decides how far back a touchpoint can sit and still earn credit. This guide reads the settings, compares two channels on the same key event, and works out what changes when the window is shortened.
After the steps, the generate_lead key event appears under the Key events tab in Admin, and the standard reports show it within a day. The worked example below gives the count and the rate to expect.
Download the reference: the generate_lead key-event reference (JSON) (776 bytes, MD5 4452e63b361df9a027b3610b31eb8c57). It holds the event name, the counting method, the worked-example figures recomputed for this guide, and the checkpoints that prove the key event is recording. The figures in the file are the same figures shown on this page.
What you need
- Editor access or above at property level in GA4. Marking an event as a key event is a property setting, and the star in the Events table does not appear for a Viewer.
- A linked view of the property's attribution settings: Admin, then Attribution settings. The reporting attribution model and the key event lookback window both live there.
- At least one key event marked, with traffic from more than one channel, so the credit split is visible. In this guide the event is generate_lead.
The trace
- Open Admin, then Attribution settings. Read the reporting attribution model. The default is data-driven attribution, which shares credit across touchpoints using the property's own data, rather than giving all of it to the last click.
- Read the key event lookback window. The default window is 90 days. A touchpoint older than the window cannot earn credit for a key event, however important it was.
- Open Advertising, then Attribution, then Model comparison. Pick generate_lead as the key event. The report shows the same key events credited under different models, so the effect of the model choice is visible on one page.
- Compare two channels under the default model. In the worked example the window shows 214 generate_lead key events: 96 credited to organic search and 58 to paid search, with 60 spread across the other channels. The shares are 96 / 214 = 44.86% for organic and 58 / 214 = 27.10% for paid.
- Check what a shorter window would do. A lead that arrives 120 days after the first touch sits outside a 90-day window, so that touch earns nothing under the default. Shortening the window to 30 days would move credit from early touches to later ones, and the channel shares above would shift with it. The window is a measurement choice, not a correction.
- Note the difference between reporting scope and attribution. Session-scoped dimensions such as Session source / medium follow the session, not the attribution model. Event-scoped dimensions such as Source and Medium follow the reporting attribution model, and only carry values for key events.
Dead ends
DEAD END — READING SESSION SOURCE AS ATTRIBUTION. Session source / medium credits the session the key event happened in. It is not the attribution model, and it will not match the Model comparison report. The two answer different questions.
DEAD END — CHANGING THE MODEL TO FIX A CHANNEL. Switching from data-driven to last click does not fix tracking; it re-labels the same key events. If paid search looks weak under data-driven attribution, check the tagging and the landing page before changing the model.
DEAD END — EXPECTING ATTRIBUTION TO CHANGE THE COUNT. The attribution model divides credit; it does not change the total. The 214 key events stay 214 under every model, and only the shares move.
End state
The generate_lead event keeps data-driven attribution with a 90-day lookback window. The window credits 214 key events, with organic search at 44.86% and paid search at 27.10% of them.
| Figure | Value |
|---|---|
| generate_lead key events in the window | 214 |
| Credited to organic search | 96 |
| Organic share | 44.86% |
| Credited to paid search | 58 |
| Paid share | 27.10% |
| Credited to all other channels | 60 |
| Other channels share | 28.04% |
| Key event lookback window | 90 days |
The shares are 96 / 214 = 44.86%, 58 / 214 = 27.10%, and 60 / 214 = 28.04%, all recomputed in Python 3 and rounded to two places. The three credited counts add to 214. The figures are a worked example, not a real property.
Sources
- Analytics Help — Conversions vs. key events in Google Analytics — what a key event is, the event to key event to conversion flow, and where Google Ads conversions appear
- Analytics Help — About key events — marking an event as a key event, the 30-event limit for a standard property, and 50 for Analytics 360
Checked 11 November 2026. The click path above is the one in Google's help pages, read 11 November 2026. Data-driven attribution as the default, the 90-day key event lookback window, and the split between session-scoped and event-scoped dimensions are as Google documents them. No step in this guide was run inside a live Analytics property, because Admin needs a Google sign-in and this site does not sign in as anyone. The figures on this page are a worked example recomputed in Python 3 with decimal arithmetic, not numbers from a real property. Google moves menu items without warning; if a path on this page no longer matches, the linked help page is the current map.