Search Analytics Overview - SearchStax
SearchStax® Analytics provide the insights you need to make your search experience more accurate, more efficient, and more profitable. Analytics features are available to our premium clients (Gold support and above). SearchStax Analytics features let you track the history of searches, clicks, revenue, and searches per session, as shown in the summary graph above. At a lower level, the Analytics suite summarizes popular queries and critical metrics such as click-through rate, average click position, mean reciprocal rank, and frequency of no-results searches.
SearchStax Analytics are not just a set of pretty graphs. They provides insight and control where you need them most.
Contents of this page:
- Analytics Architecture
- Analytics Apps
- Analytics Reports
- Typical Test Metrics
This page is an overview of the Analytics features, which are broken out in more detail on subsequent pages.
The SearchStax Analytics features are based on search-page reporting of user queries and click-through events. Your application's search page already handles queries and responses. SearchStax Analytics requires some additional code to be added behind the scenes.
- The library issues five different types of event messages through HTTP.
- The event messages are collated by a SearchStax Analytics App.
- The Analytics App automatically serves several pages of graphs and tables describing how users relate to your search results.
An "event" is a user action on your search page, such as a mouse-click on a link or button. Event messages are sent to a SearchStax Analytics App for collation and display.
SearchStax supports the following event messages:
See Recording Events to learn how to capture user events on your web page.
An Analytics App is a dedicated listener that
accumulates HTML event messages
from your search pages. It then analyzes the events and displays a variety
of analytic tables and graphs (see below).
As a SearchStax admin, you can create as many Analytics Apps as you need. Each App is uniquely identified by a key number. Incoming event messages are addressed to the key. Therefore, an Analytics App can track events from any web application using any search engine, whether it is a SearchStax deployment or not.
See Analytics Apps for more information.
There are multiple pages of analytics displays, accessible through the Analytics App's Dashboard: See Analytics Dashboard, Usage Metrics, Relevance Metrics, Response Time Metrics, and Insights for more information on these displays.
An introduction to typical metrics used for testing appears below.
Typical Test Metrics
SearchStax comparison tests rely on common metrics of search behavior, such as these relevancy metrics provided by SearchStax search analytics:
These metrics are then compared under different search strategies to determine whether a configuration change has helped or hindered search success.
The simplest measure of search success is the click-through rate. This is the proportion of queries that resulted in the user clicking on one of the result items. Higher proportions are better.
In this illustration, the control received only one click-through event out of four queries. The treatment resulted in three click-through events out of four queries.
Average Click Position
Average click position reports how far an item was from the top of the results list when
the search user selected it. (Low numbers are better.) This illustration shows
how accumulated click-position metrics can reveal an improvement in the
ranking of desirable result items. In this case, the treatment moved the desired result item
closer to the top of the result list.
Mean Reciprocal Rank
The mean reciprocal rank is another measure of click position. The scoring normalizes the score to the range 0 to 1. Higher values are better.
How often does a user run a query that produces no matches? Some such behavior is unavoidable because users misspell, mistype, or simply abuse the input. Beyond that point, however, systems can be tuned to broaden unproductive searches.
Does the treatment system produce fewer no-result searches than the control system?
In this illustration, the treatment has cut the number of unresponsive searches in half. This would be very desirable.
These metrics are all part of SearchStax Search Analytics. You can adjust the behavior of the search and observe changes in the metrics.