Skip to content

Suggestions

Suggestions in Ecommerce Search are phrases suggested to the user while the user is typing in the search box on a webshop. The purpose of these suggestions is to guide the user to choose a phrase that gives the most relevant results.

Suggestions are generated from stored suggestions candidates with the text from the suggestion candidate. The suggestion candidates are read by Ecommerce Search and converted into suggestions in regular intervals of once every day.

Suggestions candidates

Suggestions candidates are used to generate suggestions. Examples of suggestion candidates might be parts of the product description, proper names, or popular search phrases.

Sources

Suggestion candidates can come from a number of sources. Some could be:

  • Product names
  • Brands
  • Top 1000 searches from Google Analytics

Management

The suggestion candidates are created automatically, so there is no UI for creating them.

Suggestion candidates can be generated from product data by using Suggestion Generator Rules.

The following functionality is available in the UI and in the API:

  • Search returns suggestion candidates matching the phrase, segmentId, and datasource in the request. Paging is also available.
  • Get returns the suggestion candidate that has the supplied id.
  • Create creates a suggestion candidate.
  • Update updates the suggestion candidate with the supplied id.
  • Delete deletes the suggestion candidate with the supplied id.

Taboos

Suggestions Taboos are used to avoid creating suggestions containing any taboo phrases. It can be because of regulatory requirements. An example could be painkillers.

Searching for suggestions will return results even if the word is not complete (partial matches) and if it is spelled incorrectly (fuzzy search). So a search for “bic” or “bick” can return “bicycle” as a suggestion.

Suggestion relevance

Suggestions have a simple boosting model. When manually creating a suggestion candidate, the boost is zero. If a suggestion candidate is generated by a suggestion genrator rule, it will have a score equal to how many times that candidate was generated.

Example

There are three products that result in the suggestion "chainsaw". There is one product that result in the suggestion "charger 24v". If a user searches for "cha", the chainsaw will be first with a boost of 3 and then the charger with a boost of 1. Manually created suggestions will be last with a boost of 0.