Attribution Models
By the end of this lesson, you should be able to: explain how the common attribution models split credit, say why they measure presence rather than influence, explain what data-driven models improve and what they can't fix, and describe how to measure what a channel actually adds.
What attribution is
Before buying, a customer often touches several marketing channels. Say someone sees an ad on Instagram, a week later searches for the product and clicks a search ad, then gets an email and buys.
Attribution decides how to split the credit for that sale between Instagram, search and email. The answer matters, because marketing budgets follow it. A channel that gets more credit usually gets more money.
The rule-based models
The common models each follow a fixed rule. Here's how each would split the credit for that one sale.
| Model | The rule | Search ad | ||
|---|---|---|---|---|
| Last touch | All credit to the final touch | 0% | 0% | 100% |
| First touch | All credit to the first touch | 100% | 0% | 0% |
| Linear | Equal credit to every touch | 33% | 33% | 33% |
| Time decay | More credit to touches closer to the sale | 14% | 29% | 57% |
| Position-based | 40% to first and last, the rest split in the middle | 40% | 20% | 40% |
The time decay row halves the credit for each step further from the sale. Real tools use different weights, but the idea is the same.
Each rule tells a different story. Last touch says email drove the sale. First touch says Instagram did. Teams can spend months arguing about which rule is fairest.