An engineer configuring a new audio codec had no way to hear what a bad setting actually sounded like until the chip was already unusable. Getting it wrong wasn’t rare. Customers were burning through 500 to 800 devices per product cycle just finding a workable configuration, at roughly $30 a unit. I was brought in to lead the redesign, and the customers who actually felt that cost were the one group I had no access to.

The Constraint That Became the Method

Direct customer access was off the table, blocked by competitive confidentiality that kept the actual buyers of these codecs at arm’s length from any vendor’s research team. Instead of treating that as a dead end, I turned to the one population using the same tools and hitting the same failure points: our own engineers. They knew things nobody had thought to ask them yet, and getting them talking took less convincing than I expected once they realized the outcome would make their own jobs easier, not just add another meeting to their week. 50 immersive workshops and interviews across multiple regions gave me a large enough sample to build role-based personas grounded in real workflow friction instead of assumed behavior. A spectrogram-based prototype, tested against those personas and validated in lab simulations, replaced blind trial-and-error tuning with a visual reference an engineer could actually read. Cutting setup time and configuration errors mattered, but the harder problem was building evidence credible enough to redirect a team that had never had user research to lean on before.

The workshop and persona repository are the artifacts I own outright. The setup-time and configuration-error reductions are outcomes the product team reported following the redesign. I can’t fully isolate my team’s specific lift from concurrent hardware changes to the codec itself, since both shipped in the same cycle. What changed for certain is that the engineering org went from having no user research at all to treating internal proxy users as a legitimate, repeatable research population, a shift that outlasted this one project.

“…as an engineer, I usually think linear through processes, but your work changed that. I have more respect for design and its ability to find the best process for the user.”

DSP Engineer
(@ semiconductor company)