Redesigning search for public sector sales platform

Stotles helps companies find and win UK public sector contracts. The core feature was a notice search that relied on a keyword filter, with users maintaining long lists of exact-match terms (some over 160). The system missed relevant notices when buyers used different terminology, it took customer success teams over a month to onboard new accounts, and search results surfaced persistent noise even when keywords were well-tuned.

My role

When engineering created a proof of concept for hybrid search combining keyword and semantic matching, my job was to determine whether it could meaningfully improve the experience. I scoped a 20-user co-development programme, structured research questions around the value to users and the business case for semantic search replacing keyword search. I designed an MVP that replaced the keyword input with a topic-based search. Initial testing with our co-dev group surfaced a trust problem as well as technical adjustments to the search mechanism. Users had built confidence in the existing search through years of effort, and a natural language based system, however technically better, struggled to earn that confidence without clear explainability. These insights directly shaped the development of the next phase of testing, a freemium new user sign up flow designed to gain quantitative insights from self-serving users.