Search That Learns Case Study

The Business Problem

Every search that finds nothing is a missed chance to connect two members. People use slang, abbreviations, brand names and misspellings, and no word list can predict them all in advance. Without a record of searches, head office had no way to know which ones were failing, which trades were in demand, or whether members were clicking on the results they were shown.

The Solution

wildesoft.net made the search record what happens. Every search is logged, apart from empty or punctuation-only ones, along with the members people click on. Head office has a new search terms page showing what was searched for, how often, and which searches found nothing.

Each night, the new search service looks at the searches that found nothing and uses AI to suggest words that mean the same as a trade members already list, such as a plural, a misspelling or a common nickname. Only confident matches are added automatically. Anything less certain is kept as a suggestion for head office to add by hand, and searches for people's names, company names or towns are left alone. When a new business category is added, it's given a search group automatically, so its members show up in the right searches straight away. Small fixes came from the same record too. For example, "it" at the start of a search now means IT, so "it support" finds IT companies.

The Result

The member search improves every week from real searches, and head office can see what members need.

See all case studies for Member Search