You can use attribution to build a report on traffic sources in different ways. When making a report, you can use information about three sources for each user: the first source, the last source, and the last significant source. For the “First click-through” and “Last significant” models, Yandex.Metrica takes information from the user's session history.
Let's look at an example:
A user clicked through to the site from a content ad, viewed a few pages on the site, and left. The user returned to the site later, but this time from search results. Some time later, the user entered the site's address in the browser and completed an order on the site (this is the moment of conversion). As a result, the user had three separate sessions on the site.
To pinpoint the referral source, we must use an attribution model:
When using this model, Yandex.Metrica detects the source for each session at the current time, without using the session history. In the example, the user had three sessions, each with its own source:
This model can be used for technical analysis of a site. For example, you can use internal referral analysis to detect pages that don't have the counter code.
This model is used for a site with deferred conversion — when the user takes a long time to make a decision about a purchase (or other goal) and may go back to the site from different traffic sources several times before taking action. You can also use this model if you need to know which of the sources attracts the most new users to the site.
This model uses the user history. In particular, it uses the traffic source of the very first session. All of the user's subsequent sessions are attributed to the first source. In the example, the user's first session was from an ad:
This model lets you calculate the conversion rate more precisely. All sources can be divided into significant and secondary (insignificant).
Then sessions from secondary sources (click-throughs from saved pages, entering the site address, or internal traffic) are attributed to a more significant previous source, allowing you to better measure its effectiveness.
In the example, the sources of the first and second sessions are significant (ad and search). In addition, the source of the first session can't be changed. The source of the third session is insignificant (direct traffic). So the user's sessions are attributed to the second significant source — search:
This attribution model also gives you reliable results for sites with fast conversion that occurs during a single session.