What is an attribution model?
An attribution model is the rule that splits the credit between touches. A touch is a visit from a source, such as a creator video, a search or a newsletter. First touch gives all credit to the first touch. Last touch gives all credit to the last one. Linear splits the credit equally.
Manalens runs six rule-based models: first touch, last touch, last non-direct, linear, position-based and time decay. Last non-direct is the default. It gives all credit to the last touch that is not Direct. The lookback is 30 days by default, and you can set it from 1 to 365 days.
Sources: Manalens attribution models documentation.
How does Steam attribute wishlists?
With UTM links to your store page. A link must have utm_source or utm_campaign. The report shows total visits with a UTM code and tracked visits, where the visitor was signed in to Steam. It counts a wishlist, a purchase or an activation only when it happens within 72 hours of the click. So players who were not signed in, or who convert later, are not counted.
Sources: Steamworks UTM Analytics documentation.
What does an example look like?
A player watches a creator video on Monday and visits your website. On Wednesday they find the game with a search. On Friday they come back Direct and buy it for $20. The table shows the credit for three models.
| Model | Creator video | Search | Direct |
|---|---|---|---|
| First touch | $20.00 | $0 | $0 |
| Last non-direct | $0 | $20.00 | $0 |
| Linear | $6.67 | $6.67 | $6.67 |
| Position-based (40/20/40) | $8.00 | $4.00 | $8.00 |
Sources: worked example; Manalens attribution models documentation.
What can attribution not tell me?
It cannot prove that a campaign caused a sale. The Manalens models measure observed clicks and visits. They do not use ad impressions, ad spend, probabilistic identity matching or causal uplift. A player who saw an ad and did not click leaves no touch. Use attribution to compare channels, and test the cause with a separate experiment.
Sources: Manalens attribution models documentation.
Frequently asked questions
Which attribution model should I use?
Start with last non-direct, the default in Manalens. It gives all credit to the last touch that is not Direct, so a return visit by bookmark does not hide the real source. Then compare first touch to see which channels bring players to the game for the first time.
Why does Steam show fewer wishlists than my campaign brought?
Steam counts a wishlist from a UTM link only when the player is signed in to Steam and adds the game within 72 hours of the click. Players who click in a social app browser, or who wishlist later in the Steam client, are not counted. So the UTM report shows a lower bound.
Is attribution the same as incrementality?
No. Attribution splits the credit for conversions that happened between the touches that were observed before them. Incrementality asks how many conversions would not have happened without the campaign. The Manalens models do not measure causal uplift, so you need a separate test for that question.