Module 6: Cadences & Health

Install the four recurring discovery cadences and baseline your Discovery Health Metrics.

Time: Half day Coding required: No. Prerequisite: Modules 0–5. You now know the full loop; this module turns it into a sustained habit. You will produce: The recurring discovery rituals scheduled in your own calendar/tooling, plus a baseline read of your Discovery Health Metrics with flag thresholds set.

Why this module

A course you finish and a practice you keep are different things. Continuous discovery lives or dies on a weekly habit, and the failure mode is silent: synthesis sessions quietly stop, the ledger goes stale, the OST drifts, and months later it shows up as a bad product decision. This module installs the rhythms that prevent that — and the diagnostics that catch slippage before it costs you.

The key reframe: the habit is the goal, not the metric count. A PM who runs one solid synthesis session a week for a year learns more than one who does a 40-interview sprint quarterly. You’re not optimizing numbers; you’re making sure you keep showing up.

Learning objectives

By the end of this module you will be able to:

Run the four discovery cadences (weekly synthesis, bi-weekly OST pruning, monthly outcome check, continuous passive-feedback triage) and say what each is for.
Treat passive feedback (tickets, reviews, NPS) as a continuous stream with a sustainable synthesis rhythm — distinct from interview synthesis.
Read your Discovery Health Metrics across cadence, coverage, and evidence quality, and recognize a flag.
Set the flag thresholds that tell you the practice is slipping before product decisions suffer.
Decide who beyond you participates in discovery (cross-functional involvement).

Concept reading

The four cadences

Cadence	Frequency	What happens
Weekly synthesis	Every week, no matter what	At least one synthesis session — interview, support review, NPS, reviews, or sales-call notes. One ledger entry. The OST gets touched. This is the heartbeat; if it holds, the rest follows.
Passive-feedback triage	Continuous / weekly	A 10-minute weekly pass over in-app feedback, tickets, and reviews — bulk-clustered, not deep-dived. Thin signals only become useful in aggregate.
OST pruning	Bi-weekly	Archive dead solutions (with reasons), merge duplicate opportunities, refresh confidence, demote branches with no recent signal. Keeps the tree honest.
Outcome check	Monthly	Step back to the root: is the desired outcome still the right one? Is it moving? Should the cycle extend or change?

Interview vs. passive synthesis are different crafts. Interviews are high-richness, low-volume, scheduled — synthesize deeply, one transcript at a time. Passive feedback is low-richness, high-volume, continuous — synthesize by bulk clustering across a batch, looking for convergence. Don’t apply the interview method to 200 tickets, or the ticket method to a 45-minute transcript.

Discovery Health Metrics — diagnostics, not a report card

These answer “is my habit healthy?” — not “am I a good PM?” Three categories worth watching, each with a flag threshold:

Category	Metric	Flag when
Cadence (are you showing up?)	Weekly synthesis session; ledger entries; OST updates	Synthesis missed 2+ consecutive weeks; ledger gap >2 weeks; OST untouched 2+ weeks
Coverage (learning broadly?)	Segment diversity; source-type diversity; unmapped-signal rate	Single segment or single source type for 30+ days; >30% of new signals still unmapped after 2 weeks
Evidence quality (earned the right to act?)	Sources per prioritized opportunity; verbatim coverage; stale experiments	Any Prioritized opportunity with a single source; paraphrase-only clusters; an experiment Running 4+ weeks with no result

A flag isn’t a failing grade — it’s an early warning to fix the practice before it shows up downstream.

Who else participates

Discovery isn’t a solo sport. Decide which engineers, designers, or support folks join interviews or synthesis — even occasionally. Shared exposure to customer signal is what stops the team from quietly reverting to opinion-driven roadmapping the moment you’re out of the room.

Go deeper: Discovery Health Metrics (all four metric categories + targets), Continuous Feedback Streams (passive-feedback streams, drip synthesis, and building a warm interview pool), and the Cadences section of the Agentic PM Playbook.

Hands-on exercise

🎯 This module is inherently about your practice. Unlike Modules 2–5, there’s no ShiftLoop substitute here — you’re scheduling your real rituals against your product and outcome. If you’re still practicing on the sample, schedule the rituals anyway as a dry run, then re-set them on your real product at the capstone.

Steps:

Schedule the four cadences in whatever you actually use (calendar, recurring task, or a scheduled agent). Weekly synthesis is non-negotiable; add bi-weekly OST pruning and a monthly outcome check. Make weekly passive-feedback triage a standing 10-minute slot.
Consider automating the heartbeat. A recurring agent that fires the weekly-synthesis prompt against your passive-feedback source makes the cadence hold even on a busy week. (This is also how the pilot satisfies its “ran twice unattended” gate.)

Recipe — how to wire it up: In a Claude Threads session inside Obsidian, invoke the schedule skill and tell it: “Set up a weekly recurring agent that runs every Monday and fires the pm-signal-synthesis skill against [describe your passive feedback source — e.g., ‘new Intercom tickets since last Monday’ or ‘App Store reviews from the past 7 days’].” The skill creates a CronCreate task in the vault. Once it’s running, your week-0 habit is: open the synthesis thread it created, review the output, write your ledger entry. The agent does the clustering; you hold the judgment.
Baseline your health metrics. Open the Discovery Health Baseline template and fill it in honestly — when did you last synthesize, how many segments/source types in the last 30 days, any Prioritized opportunity resting on a single source, any stale experiment. Keep the file; its value is in comparing it to a re-run 30–60 days later.
Set your flag thresholds from the table above, written down where you’ll see them — so a slip trips a flag instead of going unnoticed.
Name your participants. Write down who beyond you will be exposed to customer signal, and how often. (If you’re a solo founder with no team: a contractor you brief on synthesis outputs counts; so does an engaged customer you talk to regularly, a peer founder who reviews your OST, or a designer who joins one synthesis session a month. The goal is that someone besides you has unfiltered exposure to customer signal — even once a quarter is better than never.)

Deliverable: scheduled recurring rituals (show the calendar entries / tasks / agent), a one-page Discovery Health baseline with thresholds set, and a short list of discovery participants.

Success criteria

All four cadences exist as real recurring commitments — not intentions. Weekly synthesis is scheduled and protected.
You can explain why passive-feedback synthesis uses bulk clustering while interview synthesis goes deep per transcript.
You captured an honest Discovery Health baseline and wrote down the flag thresholds.
You named at least one non-PM who will be exposed to customer signal.
The weekly synthesis ritual is something you could keep for a year, not a sprint you’ll abandon.

Common failure modes

Symptom	What’s going wrong	Fix
“I’ll synthesize when I have material”	Cadence treated as event, not habit	Schedule the session regardless of volume. The standing slot is the practice; material follows the habit, not the reverse.
Deep-diving 200 tickets like transcripts	Wrong synthesis method for the stream	Bulk-cluster passive feedback; reserve deep per-signal synthesis for interviews.
Optimizing the metric counts	Mistaking the diagnostic for the goal	The metrics catch slippage; they aren’t a score. One consistent session a week beats a burst.
OST goes stale between sessions	No pruning cadence	Bi-weekly pruning: archive dead ideas with reasons, refresh confidence, demote signal-less branches.
Discovery stays a solo activity	No cross-functional exposure	Bring an engineer/designer/support person into a session. Shared signal is what holds the team to evidence.
A metric flagged and nothing happened	Thresholds not written / not watched	Put the thresholds where you’ll see them; a flag should trigger a fix, not a shrug.

You’ve run every part of the loop on the sample and installed the rhythms. The final step is the real one: run the whole cycle on your product, judged against a rubric.

→ Capstone: One Full Cycle on Your Real Product