Continuous Discovery System
Teresa Torres defined continuous discovery as weekly touchpoints with customers. Most teams read that and book a recurring Calendly slot they never fill. A real continuous discovery system has infrastructure: an always-on intake, an automated tagging layer, a weekly synthesis routine, and a direct line from insight to the opportunity backlog.
This skill builds that infrastructure. Two modes: Setup (build the system) and Weekly Run (execute the routine).
---
Context
The system has four layers:You don't need to build all four at once. Start with one layer and add the others.
---
Mode 1: Setup — Build the System
Step 1 — Audit existing feedback channels
Ask the user to list every place user feedback currently lives. Common sources:
CHANNEL | EXISTS? | VOLUME | CURRENTLY REVIEWED?
Support tickets | Y/N | [per week] | Y/N
NPS/CSAT comments | Y/N | [per week] | Y/N
In-app feedback widget | Y/N | [per week] | Y/N
User interviews | Y/N | [per week] | Y/N
Sales call notes | Y/N | [per week] | Y/N
CS escalations | Y/N | [per week] | Y/N
App store reviews | Y/N | [per week] | Y/N
Community / Slack | Y/N | [per week] | Y/N
Session recordings | Y/N | [per week] | Y/N
Churn surveys | Y/N | [per week] | Y/N
Identify the 1–2 channels with the highest signal-to-noise ratio. Start there.
Step 2 — Define what "good signal" looks like
For each channel selected, define the signal filter:
CHANNEL: [name]
Good signal = [specific criteria]
Noise = [what to ignore]
Volume target: [how many good signals per week]
Step 3 — Build the intake and triage process
Weekly triage workflow: export/collect, filter, tag (8–12 need tags + context tags), and spot patterns (3+ occurrences = flag for synthesis).
Step 4 — Build the weekly template
25-minute weekly review covering: raw intake, tagged items, patterns, opportunity statements, backlog actions, and open questions.
Step 5 — Connect to the opportunity backlog
Rules: opportunity statements appearing 3+ consecutive weeks get a backlog item. Format is need-based, not solution-based.
---