| name | write-discovery-plan |
| description | Set up a Teresa Torres continuous-discovery cadence — weekly customer interviews with the product trio, automated recruiting, and calendar discipline. Produces a discovery plan that runs continuously rather than as a one-off research sprint. Use when standing up discovery for a product slice or fixing a stalled interview cadence. |
| argument-hint | [product slice and its desired outcome] |
| user-invocable | true |
| allowed-tools | Read, Write, Edit, Bash, Glob, Grep |
Set up a continuous-discovery cadence for $ARGUMENTS.
Most products fail because teams build the wrong thing, and the cure is talking to customers continuously
rather than in big-bang research sprints that go stale before you ship. This skill produces a discovery
plan based on Teresa Torres's continuous discovery:
weekly touch points, run by the team building the product, in pursuit of a desired outcome. The plan must
be runnable — a protected calendar slot, a self-sustaining recruiting mechanism, and the trio committed.
Follow every step. The output is a plan the team can start running next week.
Step 1: Confirm the desired outcome
Discovery serves a specific outcome — the metric you're trying to move. Agree it with the product leader
before anything else. "Increase weekly active users" or "reduce week-one churn", not "learn about
customers". If the slice has a roadmap (/product-manager:write-roadmap), the desired outcome is the
roadmap root. This step is complete when one desired outcome is written and agreed.
Step 2: Form the discovery trio
Name the product trio: the PM, a designer, and
an engineer. All three attend every interview. The rationale is non-negotiable: each role hears different
things — the engineer hears constraints and workarounds, the designer hears interaction friction, the PM
hears business context. If only the PM attends, the PM becomes a filter on what reaches the team, and
filters distort. This step is complete when three named people have committed.
Step 3: Protect a recurring calendar slot
Set one recurring weekly slot for interviews — Torres's example is "Tuesday at 11am" — and protect it.
The target is at least one customer conversation per week. Bi-weekly is the floor for teams that genuinely
can't hit weekly; anything less frequent collapses back into periodic research, which is a weaker practice.
This step is complete when a recurring slot exists on the trio's calendars.
Step 4: Automate recruiting
The most common reason a weekly cadence collapses is recruiting friction — hunting for participants each
week becomes a bottleneck that kills the cadence within a few weeks. Automate it from the start:
| Situation | Recruiting mechanism |
|---|
| Active product with logged-in users | In-product prompt to a % of users: "Do you have 20 minutes to talk about your experience in exchange for [incentive]?" Agreers enter a scheduling pool; a calendar integration books the next slot |
| Pre-launch / small B2B / enterprise | Recurring pool via customer success, LinkedIn, industry communities, or a specialist panel — build a pool, don't recruit per-interview |
The goal: wake up Monday with an interview already on the calendar, having done nothing to arrange it that
week. If you reference a paid panel or recruiting service, flag the cost explicitly — prefer the in-product
or customer-success route first. This step is complete when the recruiting mechanism is specified and the
first participants are in the pool.
Step 5: Define the interview type and target segment
State whether the cadence runs generative discovery interviews
(/product-manager:write-interview-guide), switch interviews
(/product-manager:switch-interview), or alternates. Name the target segment narrowly enough that themes
will saturate — 20-30 interviews across a reasonably homogeneous segment typically reaches 90-95%
saturation. This step is complete when interview type and segment are stated.
Step 6: Set the synthesis and OST update rhythm
Discovery without synthesis is just conversation. Set the rhythm:
- After every interview: log the participant, type, and key signal to
docs/product/discovery-log.md
- Every 3-4 interviews: revisit the opportunity space and update the OST
(
/product-manager:synthesise-interviews, /product-manager:write-opportunity-solution-tree)
- Monthly minimum: full OST review
This step is complete when the synthesis cadence is written into the plan.
Rules
- The trio attends every interview. Researchers reporting back to a PM is not continuous discovery — it reintroduces the filter the trio exists to remove.
- Protect the slot. A discovery cadence that gets bumped for "more urgent" work isn't a cadence. The recurring slot is sacred.
- Automate recruiting before the first interview, not after the cadence stalls. Manual recruiting is the single biggest cause of collapse.
- Weekly beats a sprint. A one-time discovery sprint produces understanding that is months old by ship date. Continuous discovery keeps pace with a changing world.
- Flag paid recruiting services. In-product and customer-success recruiting are free; panels cost money. Name the cost and prefer the free route first.
- Discovery serves an outcome, not curiosity. If interviews aren't connected to a desired outcome and changing decisions, you're running discovery theatre.
Output Format
Write the plan to docs/product/discovery-plan-[slice].md:
# Discovery plan: [slice]
| Field | Value |
|-------|-------|
| Desired outcome | [metric, baseline → target] |
| Trio | [PM / designer / engineer] |
| Recurring slot | [day / time] |
| Cadence target | [weekly / bi-weekly] |
| Recruiting mechanism | [in-product / CS / panel] — [cost note] |
| Interview type | [generative / switch / alternating] |
| Target segment | [narrow segment] |
| Synthesis rhythm | [per-interview log; OST update every 3-4; monthly review] |
## First two weeks
- [ ] Calendar slot created and protected
- [ ] Recruiting mechanism live, first participants in pool
- [ ] First interview scheduled
- [ ] Discovery log initialised
Related Skills
/product-manager:write-interview-guide — the guide for each generative interview./product-manager:switch-interview — the guide when the cadence runs switch interviews./product-manager:synthesise-interviews — turn the interview window into OST updates./product-manager:write-opportunity-solution-tree — the artifact discovery keeps current.
