AI statistics

AI statistics

AI statistics give us an insight into how the AI ​​performs and the parameters that influence the recommendations displayed. We can see various metrics, including the number of generated recommendations and the effectiveness of recommendations across different pages.

image of ai statistics  recommendations performance tab in rek.ai dashboard

AI recommendations

This metric indicates how many instances of recommendations have been generated. For instance, if a news article displays 3 recommendations at the bottom of the page, it counts as one recommendation. The figures presented are divided into:

  • Links
  • Autocomplete
  • Questions and answers

Performance

Avg. response time:

This metric, represented in milliseconds, signifies the duration of time taken for the recommendation to be processed on the server prior to being transmitted to the user.

Variation score:

This metric provides insight into the diversity of recommended pages that visitors have engaged with. A higher value indicates a wider array of recommendations, showcasing their appeal to a diverse audience. In this context, it's evident that 271 unique pages have been recommended, highlighting the breadth of content offered to users.

Engagement

This metric represents the proportion, expressed as a percentage, of visitors who engaged with the displayed recommendations by clicking on them. One engamenet is defined as when a user has a recommendation in the viewport for 1 second and then interacts with them.

The percentage is an average of all pages on the website that present recommendations. This is to equalize the differences between how the recommendations are presented on different pages.

The numbers are broken down by Desktop and Mobile for the different types of recommendations made.

  • Recommendations of links and news
  • Autocomplete engagement

Pages with high engagement in recommended links

This list identifies the pages where recommendations yield the most favorable outcomes. These pages exemplify instances where users have derived the utmost benefit from the recommendations provided. In the list provided, we can see that "Page-1" is the top-performing page with an 83.3% success rate.