Search Improvements

The Challenge

The existing search system wasn't designed for people in crisis. Users encountering housing terminology for the first time, dealing with stress about losing their homes, and primarily using mobile phones during emergencies couldn't find the advice they desperately needed.

Working closely with the lead developer, I learned about the technical constraints and possibilities. This helped me prioritise improvements that could deliver immediate value while building towards bigger system changes.

The key problems: people couldn't spell crisis-specific housing terms they'd never seen before, didn't know the right search terms for complex legal issues, and kept landing on pages written for lawyers instead of people seeking help.

hen

Research and Discovery

The project began with existing research indicating that zero-result searches were contributing to increased helpline calls. I took ownership of deeper investigation, focusing on understanding the specific challenges faced by users in housing crisis situations.

Recognising the trauma-informed approach necessary for this vulnerable user base, I designed a research methodology using proxy users combined with comprehensive analytics analysis and screen recording review through Hotjar. This approach enabled us to understand user behaviour without subjecting vulnerable users to additional stress during housing emergencies.

Working closely with the lead developer, I analysed misspelled search reports to identify patterns specific to housing terminology. Collaborating with a web content editor, I systematically reviewed these files to uncover common misspellings of housing-specific terms that weren't being caught by the backend search system, as well as errors in specific letter correspondence numbers related to housing and eviction processes.

Key Research Findings

Problem Definition and Strategy

I worked with the lead developer to establish a two-phase implementation strategy. The developer guided me through backend possibilities, identifying improvements that could be implemented without significant design support for phase one, versus changes requiring more intensive design work for phase two.

I prioritised technical improvements for phase one: enhanced spelling suggestions incorporating housing-specific synonyms, improved search result ordering, and backend algorithm refinements. For phase two, I planned the more complex UX improvements including autocomplete functionality, result categorisation tabs, and mobile optimisation.

This phased approach proved essential when I discovered that implementation would need to demonstrate success on Shelter Scotland before gaining support from the wider Shelter organisation, which was initially resistant to change.

Testing Strategy and Validation

Technical Implementation

Working with developers, I advocated for improvements including enhanced spelling suggestions with housing-specific synonyms, autocomplete functionality, result categorisation showing content source, mobile optimisation with tabbed results above the fold, and separation of legal content into dedicated search.

I established success criteria that valued task completion improvement and reduced user frustration, recognising that effective search serves users in crisis situations who need immediate, accurate information.

Launch and Coordination

I secured senior stakeholder support for implementation on Shelter Scotland, positioning the project as a pilot that could demonstrate value before wider organisational adoption. This strategic approach proved effective when the wider Shelter site was initially resistant to implementing our recommendations.

The Scotland implementation included all proposed improvements, requiring coordination between design, development, and content teams to ensure seamless integration with existing site architecture.

Results and Impact

The implemented search system delivered measurable improvements across all key metrics: average search depth decreased by 0.22, search refinements decreased by over 3%, and bounce rates decreased by over 1%. Most significantly, 90% of users in post-implementation testing completed tasks more quickly than before.

Organisational influence: After reporting on our analytics success, the wider Shelter site adopted several of our design recommendations including search suggestions, autocomplete, and result categorisation. While they didn't implement the mobile tab system or legal content separation, the demonstrated success influenced their broader UX strategy.