continuous-discovery

Continuous Discovery — post-launch discovery procedure

When to use

A live product with real users, and the question is "what should we build next?" or "will this idea work?". This is the post-launch complement to The Mom Test: Mom Test validates whether a problem/business is real before building; this skill structures the ongoing loop of choosing and de-risking what to build once you have users. For how to conduct the conversation itself (question technique, bad data, commitment), defer to the mom-test skill — it's stronger on that and this skill does not repeat it.

The procedure

1. Set one outcome

1. Pick a product outcome: a metric the team can move directly (e.g. "increase % of new users who reach the aha moment"), not a business outcome (revenue — lagging, cross-team) and not a traction metric (usage of one feature — too narrow). Prefer leading over lagging; iterate toward faster feedback (90-day retention → 5-day is a legitimate revision).
2. In a team: leader picks which outcome and shares the strategic why; the team commits how much movement ("+10% this quarter"). Neither side names solutions. Solo: still write the outcome down — it's the root of the tree and the referee of every later debate.
3. One outcome at a time, kept for multiple quarters. Pair it with a health metric so you don't optimize it destructively. New territory? Set a learning goal ("find the levers") before a numeric goal.

2. Map the opportunity space (the tree)

1. Draw an experience map of how customers do the job today. Scope it with a question calibrated to ambition: "How do people use our product to X" = optimization; "How do people X at all" = new markets. In a team, each person draws alone first, then merge (prevents groupthink). Visual — nodes (moments) and links.
2. Build the opportunity solution tree: outcome at the root; below it, opportunities = customer needs, pain points, desires in the customer's own words; below those, solutions; below those, assumption tests.
3. Construction rules — the tree's value is in following them strictly:
   • Child = subset of parent. Siblings = similar but distinct (you can address one without the others).
   • Top-level opportunities = distinct moments in time in the customer journey; branches must not overlap.
   • No solutions in disguise — test: "Is there more than one way to address this?" If not ("let me skip ads"), it's a solution; write the need behind it ("I don't like ads").
   • No company-perspective framings, no feelings-as-opportunities (capture the cause), no vertical single-child chains, no node that fits under two parents.
4. Feed the tree from interviews (step 3); it's never finished.