Search Relevance – SearchStax
SearchStax® includes search analytics reports that track the relevance and click-through events of your deployment.
To view the displays on this page, you must:
- Create an analytics app.
- Insert analytics scripts in your search-results page.
- Let people use your search application so you can collect statistics.
- Navigate from the Analytics node on the left-side navigation column to the list of Analytics Apps.
- Click the name of the appropriate App to see its Dashboard.
- Select the Relevance node in the menu bar.
- Select Average Click Position, MRR, or Similar Searches from the menu.
Contents of this page:
Average Click Position
When a user clicks an item in the search results, what was the position of the item in the list?
Average Click Position is a good gauge of the effectiveness of the search. Low numbers (below 5 or so) mean that the users found the desired items near the top of the search results. Higher numbers indicate that the search was successful, but the user had to scroll down through the search results to find the desired items. If the user cannot find the desired item at all, that will appear under No-Result Searches.
In general, the lower the click position the better. Over time you would like to see the Average Click Position trend down as you tune your search application.
The data table below the graph lets us dig for more information. It can be filtered by a keyword or by setting a numeric limit on Average Click Position, Clicks, or Search Volume. It can be sorted up or down on any one of the columns.
|Query||The primary search query. The display does not include the settings of facets and filters.|
|Average Click Position||The average position of the result item when this query resulted in a clickthrough event.|
|Clicks||Number of clickthrough events in the requested time period.|
|Search Volume||Number of searches for this query in the requested time period.|
Mean Reciprocal Rank (MRR) is a weighted average of click-through positions resulting in a value between zero and one. Higher values indicate that the “right answer” was often near the top of the search results.
MRR gives us feedback on the quality of our search experience. It helps us tune the system. For instance, did the recent changes in the synonym list improve the search experience? If so, then the MRR value should have risen. An upward trend in the MMR value over time indicates that the search is becoming more focused on the needs of the users.
The Similar Searches display gathers similar user queries in order to identify patterns in user behavior when discovering a particular result. Look for Similar Searches under the Relevance menu on the left side of the Managed Solr dashboard.
This display is a powerful tool to improve the search experience for your users. It shows user queries that led to specific results. This lets the team optimize search relevance algorithms and extend the set of synonyms.
|Clicked Item||Various queries led users to click-through the same result item. This is how we know that the queries are related.|
|Queries (Top 50)||The most popular queries that led to this result item.|
|Unique Queries||Number of textually-distinct queries that led to a click-through on this item.|
Do not hesitate to contact the SearchStax Support Desk.