The Explore section provides insights into the performance of specific search queries and results. You can see how often people searched for particular queries and how these queries performed over time. Additionally, this page provides insights on which query/result pairs are performing the best, which in turn influences the machine learning ranking in Sajari. Pairs that perform better will be boosted over time, ensuring your users receive the most relevant search results possible.
To dive deeper into how a particular query performed, choose Query from the dropdown in the top left and enter the search query into the text field.
The graph below shows useful query statistics over time:
- Searches, the number of times people searched for the query term. Select this to see the number of searches over time.
- CTR (click-through rate), the percentage of times a result was selected for this search query. Select this to see CTR data over time.
Note: Make sure you have click tracking or conversion tracking set up to access those statistics.
Search result ranking
This report removes individual bias and shows results ranked based on past user behavior. Aside from the result ID, which is typically a URL or a numeric ID, you can see how often the result appeared for the specified search query.
Using click and PosNeg tracking, the table also shows the performance of each result along with a confidence score. At a glance, you can see which results have excellent performance and are most likely to be selected by users. The confidence score provides additional certainty and is calculated based on the number of interactions and signal strength.
If the confidence score is high, you can be certain that excellent performing results are great matches for the query. If the confidence is low, the performance can be seen as a directional signal, but more data is required to be certain of the performance.
To dive deeper into how a particular result performed, choose "Result ID" from the dropdown in the top left and enter the ID of the record (typically a URL or numeric ID).
The graph shows the number of views for the specific result over time. This allows you to understand the impact of promotions, the introduction of new products or content, or understand how changes to the search algorithm affect individual results.
- Views, how often the specified result was viewed.
In the query result below you can see how often the specified result appeared for various queries. Just like for queries above, a performance and confidence score shows how likely it is that this query/result pair performs well.
Consider the following example in which a user wants to understand click-behavior for the search query "water restrictions".
The graph shows how often people searched for the term water restrictions, and the percentage of time they selected a result.
The result table shows that the first URL for the search term "water restrictions" is most likely the best result, indicated by the "good" performance. At a 73% confidence level, the change of the performance rating being correct is also fairly high.
The second result received far fewer interactions than the first result and hence has a low confidence rating of 37%. With it's "indifferent" performance and low confidence rating, it's not yet clear how good of a match this result is and more interaction data is required.
All other results in the list received a poor performance at a low confidence rating. However, the mix of poor performance and low confidence does indicate that these results are likely to be not good matches for the query and Sajari's machine learning algorithm will rank these URLs lower as a result in future searches.