Search Interaction Design
Redesigning the job search experience for a growing audience of mobile-first users - making it faster to find relevant jobs and easier to narrow down results
Role
Lead Designer
Scope
End-to-End: Research to Delivery
Company
HeyJobs
Overview
tl;dr
HeyJobs’s search engine had outpaced its interface, leaving mobile users struggling to find relevant roles. Following two rounds of research (19 participants), we discovered that result quality, not filter complexity, was the primary pain point. I designed a mobile-first search experience, featuring a full-screen modal, scrollable quick filters, and live result counts, that successfully shifted user behaviour from manual keyword typing to intuitive filter browsing.
Key Impact
context
Why search & filtering?
HeyJobs serves essential workers, a segment that is overwhelmingly mobile-first and job-hunts on the go. The existing desktop-centric search experience failed to meet the needs of this audience, who rarely use traditional filters as a primary navigation tool.
PROBLEM
Experience built on assumptions
Search is the core of the HeyJobs experience.
A major backend upgrade had significantly improved the quality of our results - but users couldn’t tell. They were navigating an interface designed for a different era, one that assumed they knew exactly what to type and penalised anyone who didn't.
Two types of users were struggling in different ways:
Since the majority of our users are on mobile, these UI flaws were magnified:
Screen Hogging: The search form dominated the screen even when users were just browsing
Irrelevant Logic: The radius filter appeared regardless of whether a location had been entered
Invisible States: Filters were hard to discover, and active states were invisible, making it impossible to see what was currently narrowing the results
Process
Research & Design
Following our Lean UX process, we ran two rounds of ideation and validation - each building directly on what we learned from the last.
Round 1 Discover & Respond
13 participants · Remote usability testing · Mixed age range, 24–60
We observed participants searching on competitor sites and our site in a natural setting, then had them test a prototype with full-screen search, quick filters, and an extended filter panel.
The single most important finding from this rounds was simple:
If results didn't match intent, users left - without ever touching a filter. This reframed our entire focus: filter improvements only matter if search quality is already working.
Search vs. Filter: Filter usage was only 5%, as users preferred scanning results. Full-screen search, however, was adopted immediately and intuitively.
Clarity Issues: Quick filter active states were unclear for the majority, and advanced filters suffered from low discoverability.
User Priorities: Location and job type were the most utilised categories, while recent searches were specifically requested to reduce friction.
Armed with these findings, we began exploring new interaction patterns. The goal was to establish structural patterns, how components would behave individually and together.
key questions
Round 2 Refine & Validate
6 participants · remote usability testing · Mixed age range, 24-45
We focused this round on the refined designs - specifically the scrollable quick-filter strip and full-screen search with autosuggest - combined with an new feature 'Quick Apply' to observe how the two flows interacted.
Overall the design was perceived very well:
Consistent Intuition
Full-screen search remained the primary, intuitive entry point for all participants, validating the core navigation model.Effective Visual Affordance
The "visual cut-off" on the filter strip successfully prompted horizontal scrolling, making the full range of quick filters discoverable.Improved Clarity
Active filter recognition increased significantly, and advanced filters were easily located and used without friction.
User Requirements
The two rounds converged on a clear priority order - the foundation for every decision we made in the final design.
With the user requirements prioritised and stakeholder buy-in secured, I refined the final interaction patterns and paired with the Front-End team to define the component logic and implementation details, ensuring the design was ready for production.
IMPACT
Validation & Results
Validated through an A/B test with 33,425 users (95% significance), then rolled out to 100% traffic two months later.
Filter usage: The core shift
Behavioural Change
Primary Metric
Job card clicks per session, bounce rate, and conversion from search to application showed no significant change. But that's the point: filter-driven feeds performed comparably to query-driven ones - which validated the whole premise. Broad Searchers using filters were getting to relevant jobs just as effectively as Deep Searchers typing queries.