When choosing recommendations, our algorithms look at a variety of factors to determine the best links for each user, including contextual similarities, post popularity and audience patterns. Since our recommendations are personalized for each user, we can also filter out links to posts he/she has already read. Our algorithms are all geared to help your readers find interesting content and boost engagement.
Strossle content recommendations come in many forms with different recommendations types.
Our recommendations can be of multiple "recommendation types". Think of them as boxes of candy, each box containing a unique flavor - you chose what boxes to pick candy from based on your unique preferences.
The following is a list of recommendation types and a description of the capabilities of the recommendation.
Optimized for displaying the latest and greatest articles and is perfect to use if you want a widget that are focused on displaying the newest articles.
Machine Learning (ML)
Optimized for providing a high CTR. Many parameters (e.g. CTR, context, publish date etc.) are taken into consideration when we recommend articles using our machine learning models.
Recommends the most "read" articles for a specific source. This recommendation is based on statistics on number of times our widget has been loaded on a specific page. A strength of this recommendation is that it is able to pick up on articles heavily promoted by the publisher.
This recommendation type promotes articles frequently shared on social channels.
Just as for internal recommendations (within a specific site) we can specify what recommendation to use for external recommendations (cross site recommendations through the Accelerator).
Currently we only use Machine Learning (ML) for external recommendations.
Our ad recommendations follows a different logic than all other recommendations. To select what ad to display, each ad is given a score. This score should reflect how relevant the ad is to the user and the site the user is visiting. The following parameters are involved in assigning a score to an ad:
- What (IAB) categories has been assigned to the site and do they match the categories of the ad?
- What kind of device is the user browsing the site from?
- What is the location of the user?
Once all ads has received a score determining their relevance in relation to the page user is browsing, ads with the same score will be ranked by eCPM. The highest scored ads (most relevant to the context), with the highest eCPM will win and be shown to the user