A new bulk creation feature was in prototype, and engineering wanted evidence it worked before committing. One user group wouldn't tell the whole story: internal experts, paying customers, and external marketers each bring different context.

I ran usability testing across three groups, internal team members moderated, current customers and external growth marketers unmoderated, writing the screeners, recruiting participants, moderating the think-aloud sessions, and building the unmoderated tests myself. One cross-group view separated universal findings from group-specific ones, with severity scored by product familiarity. The findings reshaped the plan: we cut a complex in-platform editing approach that proved hard to use and rarely needed, and shipped the simpler, higher-value path first.

Before designing anything, we needed to know how enterprise customers worked at scale, where the tooling fell short, and what they'd built to compensate. Foundational discovery: no prototype, just open questions about real workflows.

I ran discovery sessions with enterprise customers across retail, grocery, hospitality, media, and sports, then synthesized them into cross-interview theme documents. The strongest signal recurred in every vertical: customers had independently built elaborate parallel spreadsheet systems because the in-platform experience didn't match how they worked. The recurring themes, naming and discoverability at scale, export and audit gaps, a hard code dependency for non-technical users, each became a structured deliverable with quotes, implications, and next steps the team could design against.

One study at a time couldn't keep pace with the roadmap, and ad hoc recruitment was slow and inconsistent. We needed a repeatable way to source the right participants and run parallel studies across regions without findings bleeding together.

I founded a customer advisory community as a standing recruitment pool: a hand-selected founding cohort of enterprise customers across travel, food, fintech, retail, and tech; a year-long cadence of an in-person kickoff, monthly virtual sessions, and quarterly events and newsletters; and a dedicated channel where members answered each other's product questions within the first week. Its scoring model expanded the program across Europe, North America, and Asia-Pacific. With the pool in place, concurrent studies ran across product areas at the roadmap's pace, with screener quality and clear synthesis boundaries keeping every insight attributable.

Research pays off only when the study is built to answer the real question. The wrong method, or framing so loose that any result confirms the plan, wastes participants and yields findings nobody can act on.

I matched method to question: moderated think-aloud sessions to understand reasoning and probe hesitation, unmoderated prototype tests for independent task completion at volume. Study goals were written as falsifiable hypotheses with explicit success metrics, scope lines made clear what each study could and couldn't produce, and discussion guides kept sessions consistent, so no participant was wasted and no finding arrived unactionable.

Demand arrived through three disconnected channels: a sales request log, a research repository, and live meeting notes. Each has its own bias. The log favors the loudest account, the repository favors whoever was studied, the meetings favor whoever was in the room. Prioritizing from one channel meant building for the loudest voice.

I synthesized each channel on its own terms: 340 logged requests from an 18-month window coded into weighted themes, a multi-year research repository mined, months of meetings across 20+ enterprise accounts distilled. Cross-reading the three produced one demand picture where confidence comes from agreement: themes confirmed by all three channels became high-confidence priorities, two-channel themes carried their caveats, and single-channel signals were held for validation. The roadmap now prioritizes from evidence no single channel could provide.