Recommendations

7/20/2012

SiteSeeker can give recommended hits based on how other users have behaved after performing a search query. You can manage which user behaviour will trigger a recommendation.

SiteSeeker creates recommendations based on the user's query and her subsequent actions. In the simplest case, a user has clicked on a search result, but it could also be that the user has placed a product in the shopping basket or read an article during at least two minutes.

User needs for an e-commerce site

  • An analyst wants to be able to identify search queries with unusual high or low conversion rates, in order to be able to adjust the product line.
  • A webmaster wants to be able to follow the search functionality's conversion rate, in order to review changes and suggest improvements.
  • An e-commerce website visitor wants recommendations for products which many other visitors with similar interests (as based on the identical or almost identical search query) have bought, in order to make a better informed decision.
  • A visitor wants to receive recommendations on hits which many other visitors have clicked on/bought when they typed in the same search query in order to reach the desired page more quickly (these hits does not have to match the search query in question).
  • A visitor wants to be able to see which products are the most popular in a global perspective, in order to get advice about great buys from other visitors.

User needs for an intranet

  • A user wants to receive recommendations on documents which many co-workers clicked on/downloaded/viewed at least one minute after having performed a similar search query, in order to easily judge which documents are the most interesting.
  • An editor wants to see which documents that have an unusually high or low conversion rate in order to determine which documents to update and which to delete.
  • An intranet manager wants to be able to analyse the search functionality's conversion rate in order to evaluate changes and suggest improvements.
  • A user wants to sort the search results in order to display those hits that many other co-workers have found interesting first, thereby finding the relevant page more quickly.
  • A user wants to be able to view which documents are the most popular, from a global perspective in order to get advice about relevant information from other co-workers.

Solutions

Sorting by popularity

You sort the result of a query by how many clicks/conversions are registered for a page, by adding sort:popularity in the search query or by using the integration API. The result with the most conversions is displayed in the very top of the results list.

Search by recommendations

It is possible to search for recommendations manually in the template search page by adding siteseeker:recommendations after the search query. In a production scenario, a developer would add a second segment showing recommendations. The recommendations are presented in order of number of clicks/conversions the pages have for the search query in question.

Rank by recommendations

You can configure the ranking to be influenced by recommendations. See the tab Implementation details for instructions on how to proceed.