Web and product analytics
Marketing attribution models explained
The short answer
An attribution model decides how credit for a conversion is shared across the touchpoints that came before it. Last-click gives it all to the final touch; first-click to the first; linear splits it evenly; time-decay favours recent touches; position-based favours first and last; data-driven uses your own conversion paths to assign weights. GA4 uses data-driven by default.
Key takeaways
- Models only share out credit; the same $1,000 sale can be credited to three different channels.
- Fixing tracking (UTMs, calls, lead outcomes) matters more than the choice of model.
- Only a holdout test shows what a channel adds on top of what would have happened anyway.
On this page 6 sections
One customer, six models
A customer finds you through a Google search ad, comes back from your newsletter a week later, then returns by typing your address and buys a $1,000 job. Here is how each model splits the credit:
| Model | Paid search | Direct | |
|---|---|---|---|
| Last click | $0 | $0 | $1,000 |
| Last non-direct click | $0 | $1,000 | $0 |
| First click | $1,000 | $0 | $0 |
| Linear | $333 | $333 | $333 |
| Time decay (illustrative) | $200 | $300 | $500 |
| Position based (40/20/40) | $400 | $200 | $400 |
What actually changes decisions
For most small businesses the choice of model matters less than whether the data reaches the model at all. Untagged links, phone calls with no source, and leads that never get marked as sales distort every model equally. Fix tracking first: UTMs on every campaign, call tracking if the phone matters, and lead outcomes in your CRM.
Picking a model
- Short purchase cycles (one or two visits): last non-direct click is usually fine.
- Long B2B cycles: look at first-touch and multi-touch reports side by side to see which channels open conversations and which close them.
- Enough conversion volume: GA4's data-driven model, accepting that it is a model and not a measurement.
Incrementality
Attribution shares out credit; it does not tell you what would have happened without a channel. Brand search ads, for example, often capture people who would have found you anyway. Turning a channel off in one region or for a few weeks and watching total leads is the honest test. Our significance calculator helps judge whether the difference is real.
Questions
- Which attribution model does GA4 use?
- Data-driven attribution by default for key events, with last click available as an alternative in the attribution settings.
Sources
Next step
Compare the tools for this job
Related guides
UTM parameters explained (with a free UTM builder)
What the five UTM parameters do, naming rules that keep reports clean, common mistakes, and a free in-browser UTM builder.
How to track phone calls from Google Ads
Three ways to track calls from Google Ads (call assets, website call conversions and a call tracking platform), what each one misses, and how to set them up.
A/B testing for small business: traffic and significance
How to run an A/B test on a small site, how much traffic you need, how to read statistical significance, and when not to test. With a free calculator.