What is the difference between multi-touch and last-click attribution?
Last-click attribution gives all the credit for a conversion to the final touchpoint, usually the ad or link a customer clicked right before buying. Multi-touch attribution (MTA) spreads that credit across every tracked touchpoint in the customer's journey.
Last-click is simple and easy to read. Multi-touch is closer to how people actually shop across channels, but it needs more tracking and more modeling choices.
How does multi-touch attribution work?
Multi-touch attribution first stitches together the touchpoints a person had before converting, then applies a rule or model to split the credit.
- Collect touchpoints. Ad views, clicks, emails, and visits are tied to a user or device.
- Build the path. Touchpoints are ordered into a journey that ends in a conversion.
- Split the credit. A rule (linear, time decay, position-based) or a data-driven model assigns each touchpoint a share.
- Roll it up. Credit is summed by channel and campaign for reporting and optimization.
What is an example of last-click vs. multi-touch attribution?
The same journey, two answers
A customer watches an online video ad, clicks a social ad two days later, and buys after clicking a branded search ad.
Last-click gives search 100% of the credit. A linear multi-touch model gives video, social, and search a third each. A team using last-click might cut video, even though it started the journey.
Multi-touch vs. last-click attribution compared
| Last-click attribution | Multi-touch attribution | |
|---|---|---|
| How credit is assigned | All to the final touchpoint | Shared across tracked touchpoints |
| Setup | Simple, built into most platforms | Needs cross-channel tracking and a model |
| Tends to overvalue | Search, retargeting, and branded terms | Channels with many trackable touches |
| Tends to undervalue | Awareness and upper-funnel channels | Untracked and offline influence |
| Sensitivity to signal loss | High | High, since paths need stitching |
| Shows causation | No | No |
Which attribution model should you use?
Multi-touch gives a fairer picture of the journey than last-click, but both have the same blind spots: they only see what can be tracked, and they assign credit rather than measure cause.
- Use attribution for daily optimization. It's fast and granular, but treat its channel credit as directional.
- Use mix modeling for budget allocation. It covers every channel, including offline, from aggregate data.
- Use incrementality tests to check. Experiments confirm whether a channel actually moves results.
How does Alli do this?
Alli by PMG doesn't ask teams to pick one attribution model and trust it alone. Many teams keep their existing attribution for day-to-day optimization, and add Alli Scenario Planner for budget decisions.
It pairs media response modeling with brand equity signals from Alli Brand Equity, so the brand conditions that attribution misses are part of the forecast, and every number uses the KPI definitions in Alli Semantic Layer.
Explore Alli Scenario PlannerMulti-touch vs. last-click attribution FAQ
Why is last-click attribution misleading?
It gives all the credit to the final touchpoint, so channels that start or build demand, like video and social, can look like they do nothing, and branded search can look more effective than it is.
Is multi-touch attribution still worth it?
It can be, for understanding journeys and optimizing digital channels. But it depends on user-level tracking, which privacy changes keep reducing, so most teams pair it with mix modeling and experiments.
What is data-driven attribution?
Data-driven attribution is a type of multi-touch attribution that uses a model, trained on observed conversion paths, to decide how much credit each touchpoint gets instead of a fixed rule.
Does multi-touch attribution measure incrementality?
No. It assigns credit among the touchpoints it observed. Measuring incrementality needs a comparison with what would have happened without the marketing, such as a holdout or geo-lift test.
Last updated ‹ Back to the glossary