How we built Rolling Research at Thumbtack

The starting point

Thumbtack is a two-sided marketplace connecting homeowners with home service professionals. Like most product-driven companies at scale, it had research happening — but not on a cadence. Studies were reactive: a team needed an answer, they'd request research, and the process of scoping, recruiting, running, and synthesizing would take weeks. The prior year's work had been run through an agency pilot, which kept research at arm's length from the teams who needed it most.

The goal was to change both of those things: bring research in-house, and put it on a standing weekly rhythm rather than a project-by-project sprint. In nine months, we did both.

The operating system

Rolling Research at Thumbtack ran on a simple weekly loop:

Intake. Teams submitted research questions through a lightweight form — a clear question and a timeline, nothing more. Removing the kickoff overhead was intentional: the harder it is to request research, the fewer requests you get, and the more important-but-unasked questions pile up.

Friday triage. Every Friday, incoming requests were reviewed and prioritized. What needs a full moderated study? What can be answered with a survey? What's urgent enough to kick off Monday? Triage kept the queue honest and prevented the loudest stakeholder from automatically jumping the line.

Monday kickoff. Prioritized projects launched Mondays. Recruiting happened in parallel with protocol development. Sessions ran throughout the week.

Flash Findings. The output was always the same format: an executive summary, direct answers to the questions asked, and supporting evidence. No 60-slide decks. Documents short enough that people actually read them, structured consistently enough that insights could be compared and retrieved later.

Embedded presence. Work-in-progress insights landed in Slack as sessions ran. Stakeholders didn't wait for a final readout — they got signal in near-real-time and could redirect a study mid-week if something important surfaced.

What the numbers mean

21 studies across homeowners and pros in nine months represents a meaningful increase in the surface area of the company's contact with real users. 15 product teams served means Rolling Research reached every major pillar of the organization — this wasn't a tool used by one team; it became infrastructure. A 94% stakeholder CSAT score reflects that the work was genuinely useful, not just technically competent. And a 75% increase in research volume over the agency pilot represents what happens when friction comes out of the system: more teams ask more questions because asking is easy.

The AI angle

Flash Findings aren't just for humans. A consistent, dated, structured output format is exactly what a company's AI and RAG systems need to stay accurate. As Thumbtack's AI-assisted tools grew, the research corpus became the mechanism by which those tools stayed calibrated to what users actually experience — not what they experienced when someone last wrote a persona.

This is the part most companies haven't built yet: research outputs structured so that both the PM reading it on Tuesday and the retrieval system indexing it on Wednesday get what they need. Cadence and structure together are what turn a research practice into a knowledge pipeline.

Want this for your team?

The Thumbtack program was built from scratch — intake system, triage rituals, output templates, CSAT loops, Slack workflows, handoff training. It's reproducible. If you want to know what it would take to run this inside your organization, that conversation starts with understanding how stale your current user knowledge is and where the gaps are most likely to hurt you.

Start that conversation →

Numbers from the Thumbtack Rolling Research Program, 2025. Additional program detail available on request — ian@siblingsystems.limited