GA4 Attribution Settings: Which Model to Pick and Why It Matters
GA4 defaults to data-driven attribution. For some sites that is the right call. For others, the default overcredits brand search and undercredits the channels that start customer journeys.
Updated 6 min read
GA4 defaults to data-driven attribution. Most people never change it. For some sites, that is the right call. For others, the default is actively misleading — overcrediting brand search and undercrediting the channels that actually start customer journeys.
The attribution model changes which channels get credit for conversions. It does not change the total conversion count. The same 100 conversions happen regardless. The model changes whether Google Ads gets credit for 60 of them or 35 of them, which changes whether you increase or decrease your Google Ads budget next month.
What the models actually do
Data-driven attribution (DDA) — the default. DDA uses machine learning to distribute conversion credit across touchpoints based on observed patterns in your data. It trains on your conversion paths: which sequences lead to conversions versus drop-offs. The advantage is that it adapts to your specific business. The disadvantage is that it requires enough data to train. Google recommends 300 or more conversions in the last 30 days for reliable modeling. Below that threshold, DDA falls back to a simpler model internally, and the attribution becomes unpredictable.
Last-click attribution. 100% of credit goes to the last touchpoint before conversion. This is the simplest model and the one that most overcredits brand search and direct. People ready to buy often type your URL directly. Last-click gives that touchpoint all the credit while ignoring the ad that introduced them.
First-click attribution. 100% of credit goes to the first touchpoint. Useful when the question is "which channels introduce new customers?" rather than "which channels close deals?"
Linear attribution. Credit is split equally across all touchpoints. Simple and fair-sounding, but rarely accurate — some touchpoints matter more than others.
Position-based attribution. 40% credit to first touchpoint, 40% to last, 20% split among the middle. A reasonable heuristic that recognizes the introduction and the close matter most.
Time-decay attribution. More credit to touchpoints closer to the conversion, following a 7-day half-life by default. Makes intuitive sense for short sales cycles but undervalues the introducing touchpoint in long B2B cycles.
How to choose
Two factors determine the best model for your property: conversion volume and sales cycle length.
| Monthly conversions | Recommended model | Why |
|---|---|---|
| 500+ | Data-driven | Enough data for reliable ML modeling |
| 100–500 | Data-driven or Position-based | DDA works but verify against a simpler model quarterly |
| Under 100 | Last-click or Position-based | Not enough data for DDA to learn patterns reliably |
| Sales cycle | Recommended model | Why |
|---|---|---|
| Same session | Last-click | One touchpoint, one conversion — nothing to model |
| 1–7 days | Data-driven or Time-decay | Short enough for time-decay to be meaningful |
| 7–30 days | Data-driven or Position-based | Multi-touch journeys where intro and close both matter |
| 30+ days | Data-driven | Only DDA can model complex multi-session journeys |
What changes when you switch
Changing the attribution model changes historical data in GA4 reports. It is not just forward-looking. GA4 retroactively reattributes conversions under the new model. Your historical conversion counts per channel change immediately, your ROAS numbers shift, and your channel rankings may change.
Communicate this to stakeholders before switching: "We are changing how conversions are attributed. Historical numbers will shift. The total conversions do not change; only which channels get credit."
The cross-source reality
GA4's attribution model only applies to GA4 data. It does not change what Google Ads reports, what Meta reports, or what your CRM shows. Each platform uses its own attribution: Google Ads uses last-click for search and data-driven for other campaign types within its own ecosystem, Meta uses 7-day click plus 1-day view, and your CRM typically records first-touch or last-touch UTM parameters on the lead record.
Changing GA4 to data-driven while Google Ads uses last-click creates a discrepancy. GA4 might credit Google Ads with 40 conversions (shared credit) while Google Ads claims 65 (full credit). Neither is wrong — they answer different questions. For more on navigating these differences, see our guide to marketing attribution models and the GA4 vs Google Ads discrepancy explainer.
Comparing data across sources that use different attribution models requires care. The trend agreement between platforms (both show Google Ads as the top channel) is the useful signal, not the absolute count match.
The attribution lookback window
GA4 lets you set the lookback window: how far back GA4 looks for touchpoints to credit. Options are 30 days (the default), 60 days, and 90 days. A shorter window credits recent touchpoints more; a longer window captures more of the journey but may credit touchpoints that are not meaningfully related to the conversion.
Match the lookback to your sales cycle. If your average purchase cycle is 14 days, a 30-day window captures the full journey. If it is 60 days, you need the 90-day window to see the introducing touchpoint.
Practical recommendations by business type
E-commerce (short cycle, high volume): Data-driven with a 30-day lookback. High conversion volume feeds DDA reliably. Short cycles mean 30 days captures the full journey.
SaaS (medium cycle, medium volume): Data-driven or Position-based with a 60 to 90 day lookback. Trials often start with content and close with a direct visit or email. Position-based gives credit to both ends.
B2B services (long cycle, low volume): Position-based with a 90-day lookback. Low volume makes DDA unreliable. Long cycles need the full window.
Agencies reporting across clients: Use whichever model the client chose. Do not standardize all clients on one model. Each client's conversion volume and sales cycle is different. Document the model choice and reasoning per client.
Frequently Asked Questions
Does changing the attribution model affect total conversion counts?
No. The total number of conversions stays the same regardless of the model. What changes is which channels receive credit for those conversions. A switch from last-click to data-driven redistributes the same conversions across more touchpoints.
How many conversions does data-driven attribution need to work reliably?
Google recommends 300 or more conversions in the last 30 days. Below that threshold, DDA falls back to a simpler internal model, and the attribution outputs become unpredictable and hard to verify.
Why do GA4 and Google Ads show different conversion numbers for the same campaign?
Each platform uses its own attribution model. GA4 applies its chosen model (often data-driven) across all channels, while Google Ads uses its own attribution scoped to Google's ad network. The same event can receive different credit in each system. This is expected behavior, not an error.
Does changing the attribution model in GA4 affect historical data?
Yes. GA4 retroactively reattributes conversions under the new model. When you switch, historical conversion counts per channel change immediately. The total conversions remain the same, but the channel distribution shifts.
Which attribution model is best for lead generation with low traffic?
Last-click or position-based. With fewer than 100 monthly conversions, data-driven attribution does not have enough data to learn reliable patterns. Position-based is a good middle ground that credits both the introducing and closing touchpoints.