Scaling research without scaling headcount
How I built a distributed research program at CHG Healthcare that moved four researchers out of the bottleneck. I scaled through training and playbooks first, then built a research skills library so any designer, PM, or engineer could run rigorous research at a high bar.
When I joined CHG, research was split across four initiatives: Sales Enablement, Client Digital Experience, Provider Digital Experience, and Credentialing. Each had one researcher supporting three to four product teams. The math didn't quite work, but the bigger issue was no shared process or taxonomy, so we were reinventing the wheel every time.
"Finished studies went into Confluence and got lost. There were no baselines, no taxonomy, no way to search across work. Work was siloed and had no through lines."
When product teams couldn't get research fast enough, they moved without it, relying on subject matter expertise and gut feel. Products shipped with friction that could have been caught early. Nobody had baseline measurements, so there was no way to even prove something had gotten worse. The researchers were doing real work; it just wasn't connecting to anything.
The first move wasn't a training program or a workshop. It was getting the basics in place. I created team structure, moved scattered Confluence pages into Airtable with real tagging and taxonomy, and built a bandwidth tracker so we could actually see who had capacity and when. Work that used to disappear became findable. With everyone's bandwidth clear, we had accountability.
"He organized the research team into a cohesive, high-performing unit and introduced an Airtable system that made research projects and insights easily findable and shareable across the organization."
Jonathon J., Principal Product Designer, formerly at CHG
As the team grew, we made the pitch for better software and we moved from Airtable to Dovetail, gaining better search and richer structure. We rebuilt the taxonomy properly: initiative, user type, method, theme, date. The goal was boring and practical. If a PM had a question on Monday, they should be able to search and find prior work before asking for a new study.
I was able to make the pitch to leadership for more headcount and we added a fifth researcher too. Internally, we started to formally bring PMs and designers into the fold, starting with those who had run guerrilla research in the past and wanted to grow their skills.
We didn't open it up to everyone at once. We started with designers and PMs who already had decent instincts — people who asked real questions instead of fishing for validation. They came in as observers first. We ran interviews on Zoom and piped them into a Teams observation room so people could watch without being in the session. Seeing how research actually works matters before you try to run it yourself.
I built a research playbook — study design, question framing, synthesis basics, common mistakes. People worked through it on their own time. Once grounded, they ran unmoderated tests in Maze. Designers and PMs could initiate evaluative studies without a researcher in the room.
As the program matured and AI became viable, I saw an opportunity to scale further. I built a research skills library with AI that acts like a mentor sitting next to you. Instead of a static playbook, the library responds in real-time. It asks whether research is even needed before you start. It checks what you already know. It pushes back on leading questions in your guide. A designer in the design tool can run a usability study with the same quality standards a trained researcher would enforce.
We also built out participant databases to take recruitment off the critical path. That was one of the main reasons studies stalled before — not methodology, just the time it took to find participants.
The guardrail was straightforward. Each researcher was the quality check for their initiative. They reviewed study designs before anything went live, coached on question framing, and were available when someone got stuck. Not a bureaucratic review — more like a quick gut check with someone who knew what good looked like.
People earned independence. Designers ran a few interviews with a researcher watching before we let them go solo. The coaching moments were predictable: leading questions, talking over silence, jumping to conclusions. Most people fixed those habits fast once they heard themselves doing it.
"At the end of the day, a good research conversation is a very guided conversation where you avoid leading questions. Once you internalize that, you're almost all the way there."
No studies produced bad data. The model worked.
The speed told the story. One designer-led study ran from kickoff to results in 2.5 weeks, with findings ready in 2 days, using the AI-assisted analysis workflow the team built and shared across the org. That pace was not possible when every study waited on a researcher to run it.
Designers and PMs started asking questions earlier, validating before building, actually curious about what users thought. The shift was visible: product decisions started being grounded in research, not assumptions.
For the research team, the time they got back went to foundational work. Longitudinal studies. Journey research. The big-picture questions that nobody had time to ask when everyone was buried in usability tests. Designers and PMs handled directional research. Researchers focused on strategic depth.
Research stopped being a bottleneck. It became something the product org could do together. Researchers set the standard. Everyone else had the tools to meet it.
Most programs like this start with workshops and talks. Nothing sticks without the foundation underneath. The repository, taxonomy, playbook, participant database — those had to exist first. Infrastructure compounds. Once it's there, behavior change happens naturally.
The second principle is trust. People make good decisions when they have tools, coaching, and clear standards. A research conversation means asking questions instead of leading. That's most of it. Give people the principles and let them practice, they get there faster than you'd expect.
The third: tooling multiplies capacity. Training and playbooks work to a point. Then you hit a ceiling. Intelligent tools that behave like a mentor can scale further. Designers can run studies at research quality. PMs don't need a researcher in the room. The research function doesn't shrink. It shifts from execution to strategy.