Table of Contents
Prefixbox Portal has a tool called ‘Offline measurement’ that can be used to relevance measurement. It helps customers experiment with finding the best configurations for setting up their search engines in such a way that the shopping experience and business results can be improved.
With the help of this tool crawlings can be created that can be executed on search boxes with different configurations to analyze the search results so the customers can find the most relevant ones.
The starting point is to collect the keywords that the visitors of the website are using during a search. The keywords used in a certain timeframe can be collected to create a random sample. Then the experiment can be started with different search settings to find out the best settings that would lead to more and more relevant results when using these keywords.
To analyze the crawlings executed with different search settings, experiments can be created where the metrics calculated at each experiment can be compared.
The present document has been created to help users understand how this tool can be used, what the main components of it are and how the related functions work.
The document has three main parts. The first part describes the prerequisites that are needed to experiment with offline measurement.
The second part describes the offline measurement components and how the related functions work.
In the third part of this document, we present some best practices that serve as examples on how the search settings will influence the outcome of the crawlings and the experiments.
The ‘Offline measurement’ tool can be reached at the ‘Tools’ Menu:
The three main components are: ‘Keyword sets’, ‘Crawlings’, ‘Measurements’.
To use the ‘Offline measurement’ tool, first there are some prerequisites that need to be considered. These will be described in the following chapter.
Of course, already existing subscriptions , websites, suggestion lists and search boxes can be used as well. (E.g. if there is an A/B test going on, the search box created and used here can be compared to a LIVE search box)