Playbook 5 · Setting up the work

How to do discovery research

Discovery is how a hypothesis stops being a hunch. The Discovery Workbench keeps the evidence linked to the work, not buried in a Notion archive.

Why it matters

Most research dies the day after the read-out. The findings sit in someone's deck and the same theme gets re-discovered six months later by a new PM. Discovery in Tenhaw stays alive because every interview, insight and theme links back to the outcome it informs, and themes accumulate evidence over time instead of being rediscovered each cycle.

Step by step

  1. Pick a hypothesis to test. Tied to a specific outcome or epic. "We think week-one users churn because the onboarding doesn't show their data fast enough."
  2. Recruit 5–8 interviewees. Five is enough to spot a pattern; eight is enough to be confident.
  3. Run the interview. Open questions, no leading. Record (with consent) or take notes.
  4. Ingest the transcript into Discovery Workbench. Paste text or upload audio (Whisper transcribes it).
  5. Let the AI synthesise. It extracts verbatim insights, dedupes them against existing themes, and bumps severity on themes that pick up new evidence.
  6. Review themes weekly. Promote the high-opportunity-score ones into stories on the relevant epic.

What good looks like

  • Every interview linked to an outcome or epic
  • Consent flag set; retention TTL where needed
  • Themes accumulate quotes from multiple interviews
  • High-opportunity themes promoted to draft stories within 2 weeks
  • Findings cited in the relevant outcome's Coach briefing

How Tenhaw runs this

Paste a transcript or upload audio (Whisper transcribes inside the E2B sandbox). The AI extracts verbatim insights and dedupes them into your existing themes. Themes accumulate evidence over time instead of being rediscovered every quarter, and each insight links back to the outcome it informs.

  • Tenhaw Discovery Workbench
  • Tenhaw Whisper integration (E2B sandbox)
  • Tenhaw theme deduper
  • Tenhaw outcome links

Try this on the Tenhaw platform →

In Tenhaw → AI Coach → Discovery Workbench → Add interview. Each ingestion is a long-context LLM call and is rate-limited (12 per 5 minutes per user), go for quality, not volume.

Related playbooks

4How to break an epic into stories 3How to write a value-focused epic 2How to put together a quarterly roadmap 1How to set an outcome

Run this in your team

The Tenhaw product enforces every rule in this playbook so it doesn't sit on a wiki gathering dust. Try Tenhaw or book a working session with the founder.