Multi-Touch Attribution Models: Linear, Time Decay and Position-Based
How the common multi-touch attribution models assign credit, how a linear model is calculated, and the limits every rule-based model shares.
Single-touch attribution gives all the credit for a conversion to one interaction, usually the last click. Multi-touch attribution (MTA) starts from a more realistic premise: several touchpoints usually lead to a sale, and each may deserve some of the credit.
Unlike marketing mix modeling, which works from aggregate data, MTA works at the level of individual journeys. That is what makes it detailed, and also what limits it. For the broader picture of how MTA works and where it stands today, see our guide to multi-touch attribution. This guide focuses on the models themselves.
The common multi-touch models
Linear. Every touchpoint in the journey gets equal credit. A journey with four touchpoints gives each one 25 percent. It is the simplest multi-touch model and an improvement on single-touch models because it recognizes every step. Its weakness is the same simplicity: not every touchpoint has the same influence.
Time decay. Touchpoints closer to the conversion get more credit, and earlier ones get less. The assumption is that the interactions nearest the decision mattered most.
Position-based (U-shaped). The first and last touchpoints get most of the credit, a common split being 40 percent each, and the remaining share is divided among the touchpoints in between. The assumption is that introducing the customer and closing the sale are the most important moments.
Data-driven (algorithmic). Instead of a fixed rule, a statistical model compares converting and non-converting journeys to estimate how much each touchpoint adds. It adapts to your data, but it is only as good as the journeys it can see.
How a linear model is calculated
Building a linear model from your own data shows how mechanical these rules are. You need event-level records with at least three fields: a user identifier (traditionally a cookie ID), the channel of each touchpoint, and whether the user converted.
- Keep converting journeys. Filter to the users who converted and keep all of their touchpoints.
- Count touchpoints per journey. For each converting user, count the number of interactions before the conversion.
- Split the credit. Give each touchpoint 1 divided by that count. A journey with 10 touchpoints gives each one a tenth.
- Add up by channel. Sum each channel's shares across all journeys to see its total attributed conversions and its share of the whole.
Time decay and position-based models follow the same steps with a different weighting rule in step 3.
Using model results, carefully
Attribution output is often read as a budget signal: if display collects a large share of credit, give it more; if a social channel collects little, cut it or change its strategy.
Treat that as a starting point, not a conclusion. The result depends heavily on which rule you chose. A channel that tends to appear early in journeys looks strong under position-based attribution and weak under time decay, though nothing about its real effect has changed. Comparing models side by side shows how much your conclusions depend on the assumption.
What every rule-based model shares
The differences between linear, time decay and position-based models are smaller than what they have in common:
- They only see tracked touchpoints. A TV ad, podcast or billboard that prompted a search never appears in the journey. Privacy changes, including Apple's App Tracking Transparency and browser tracking protections, have also left many digital journeys incomplete.
- The credit always adds up to 100 percent. Every converting journey hands out all of its credit, even when the sale would have happened without any ads.
- They describe, they don't measure cause. Attribution shows which touchpoints were present before a conversion. It doesn't show which ones made the conversion happen.
Major platforms have moved away from rule-based models as well. In 2023 Google removed first-click, linear, time decay and position-based models from Google Analytics 4 and Google Ads, leaving data-driven and last-click attribution.
Choosing a model
If you use multi-touch attribution, pick a model that matches how your customers buy, compare it against at least one alternative, and be cautious about any decision that changes depending on the rule.
To know what a channel actually contributes, pair attribution with incrementality measurement. Experiments and marketing mix modeling estimate how many sales would not have happened without a channel, including the channels attribution can't see. Our comparison of multi-touch attribution and marketing mix modeling explains how the two fit together.
Frequently asked questions
What is a linear attribution model?
A linear model gives every touchpoint in a converting journey an equal share of the credit. If a customer had four touchpoints before buying, each gets 25 percent. It is simple and fairer than last-click, but it assumes every interaction mattered equally.
What is the difference between time decay and position-based attribution?
Time decay gives more credit to touchpoints closer to the conversion. Position-based, or U-shaped, gives most of the credit to the first and last touchpoints and splits the rest among those in between.
Which multi-touch attribution model is best?
None of the rule-based models is correct in general, because each builds in an assumption. Comparing several models shows how sensitive your conclusions are to that assumption. To learn what a channel actually caused, you need incrementality measurement, such as experiments or marketing mix modeling.
Does Google Analytics still offer linear and time decay models?
No. In 2023 Google removed first-click, linear, time decay and position-based models from Google Analytics 4 and Google Ads, leaving data-driven and last-click attribution.