The main attribution models
| Model | Logic | When to use it |
|---|---|---|
| Last click | 100% → the final channel | Direct response, short cycle |
| First click | 100% → the first channel | Assessing reach campaigns |
| Linear | Split evenly across channels | No clear priority |
| Time decay | More weight → recent touches | Long decision cycle |
| Position-based | 40% → first + 40% → last | Valuing both awareness and the close |
| Data-driven | Machine learning on your data | 400+ conversions a month |
How attribution changes budget allocation
Picture a path: SEO → email → retargeting → brand search → purchase.
Last click: Brand search = 100% (SEO and Email = 0%)
First click: SEO = 100% (Retargeting and Email = 0%)
Linear: SEO=25%, Email=25%, Retargeting=25%, Brand=25%
Data-driven: SEO=35%, Email=20%, Retargeting=30%, Brand=15%
The choice of model determines where the budget goes. Under last click, SEO and email look ineffective, even though they are what starts the funnel.
Attributing personalization
Personalization mostly works in the mid-funnel and lower funnel:
– Recommendations on the site (onsite personalization)
– Personalized emails sent through the retailer’s own ESP
– Personalized retargeting in ad platforms
Under last click its contribution is systematically understated. For a fair read on personalization, use a holdout group — a control segment that receives no personalization — and compare its total revenue with users who did receive a personalized experience.