Skip to content

Commercial sorting

An important feature of Ecommerce Search is allowing for sorting product, based on the search request context and the commercial properties of the results. In a search result, commercial sorting is intended to be used as a tie-breaker when there are multiple search results with equal search relevance.

Commercial sorting is named as such because it makes it possible to sort results by their commercial relevance, and not only by the relevance to the search phrase.

Note

When a user searches or browses the order in which the results are presented, have been shown to increase the likelihood that they convert.

Commercial sorting takes advantage of this fact. It can be used to ensure search results the user is most likely to buy stand out at the top of the first page. Furthermore, since results that are presented early are more likely to convert, it makes sense to e.g. put products with a higher profit margin or some other commercial value, like brand awareness at the top.

Warning

Commercial sorting is designed to rearrange the most relevant results. It is possible to give so much weight to commercial sorting that relevance is compromised. One way to ensure this does not happen is to set a 'maximum score' in the general settings. Beware then the sum of all the weights must match the set max achievable score.

Commercial properties

The commercial properties are the collection used for calculating the commercial score of a result along with the context partitions and user affinities, which are resolved for each search request. A result contains commercial properties from both commercial parameters after normalization and commercial affinities.

Example

Commercial properties of a shirt might include:

  • Is in style at the moment
  • Is seasonal
  • Is second-hand
  • Is a designer brand

It might also include commercial properties that it shares with almost all products, like profit margin or having many clicks recently.

Which properties and how much each should weigh will depend on the context the user is in. Are they searching or browsing? Are they looking at shirts? Are they looking at shirts on a brand page or general shirts category? All of this can be set up with contexts.