Global Search

This one started with a Hotjar anomaly. We noticed unusual behaviour in session recordings — users opening multiple tabs, strange navigation patterns, what looked like frustration clicks. The data suggested something was wrong but didn't tell us exactly what. My first assumption was that operators were juggling multiple worker records simultaneously and using tabs as a workaround for something the product wasn't giving them natively.

I confirmed it on a site visit. I conduct in-person research sessions with clients throughout each quarter — informal, observational, just sitting alongside operators as they work. Watching their real workflow made the problem immediately obvious. Payroll bureaus are phone-heavy environments. Operators field calls mid-task constantly, and every call potentially requires accessing a different record. The workaround was exactly what Hotjar had suggested: either a long click-train back through the navigation, or a right-click to open a duplicate tab and search again from scratch. Neither was fast, and neither was invisible to the caller waiting on the line.

The solution was straightforward once the problem was clear — a global search that lets the operator find any record in the product instantly, from wherever they are, without disrupting their current context. I designed the interaction to work two ways: hotkey-first for power users who live in the keyboard, and click-accessible for everyone else. Both paths were important. The power users — typically senior operators running multiple clients — were the ones feeling the pain most acutely, but the broader team needed something they could adopt without changing how they worked.

What made this case study more interesting than a typical utility feature was what happened after launch. I ran follow-up interviews to understand what else was hard to find, and the answer was consistent: invoices, payslips, agencies. We expanded the search index iteratively based on that feedback. For larger customers — some with up to 60,000 employee records — the impact compounded significantly. A search that started as a navigation shortcut became the primary way those users interacted with the product.

What I'd do differently: I'd have instrumented search query patterns from day one. Understanding what terms people were searching for most — and what returned no results — would have accelerated our decisions about which record types to index next and flagged gaps in our data model earlier.