Growth Customer Discovery

Growth: Customer Discovery

Owns truthful learning and audience-building. Hub: growth-foundations (worldview, shared vocabulary, router). Does not redefine hub terms.

Stance. Talk about their life, not your idea. Opinions, compliments, and hypotheticals are bad data and must be excluded — not softened. Find your audience before/while building; "build it and they will come" is a predicted failure (hub Guardrail), not a strategy.

Frameworks index

| Framework | Source | Depth | |---|---|---| | Question hygiene · good vs. bad data · commitment & advancement · deflecting compliments | Fitzpatrick, The Mom Test | [references/the-mom-test.md](references/the-mom-test.md) | | Audience-first · Audience–Product Fit · Discovery → Exploration → Transformation · build in public | Kahl, The Embedded Entrepreneur | [references/embedded-entrepreneur.md](references/embedded-entrepreneur.md) | | Question-critique-and-rewrite procedure | this skill | [references/question-rewrite.md](references/question-rewrite.md) |

Question hygiene (the core filter)

Good data is specifics about the past and actual behavior. Bad data is generics, opinions, and hypotheticals about the future. Apply on every question:

| Bad (exclude/rewrite) | Good (seek) | |---|---| | "Would you use…?" / "Do you think…?" | "Walk me through the last time you…" | | "Do you like this idea?" | "What have you tried to solve it? What did it cost you?" | | "Would you pay for X?" | "What do you pay for today to deal with this?" | | Any compliment received | A commitment of time/reputation/money |

Default move: convert every hypothetical/opinion question into a past-behavior question before the conversation.

Commitment & advancement (did you actually learn anything?)

A conversation only counts as validation if the person spends a currency and advances:

• Currencies: time (a real next meeting/work), reputation (an intro, a

public endorsement), money (pre-order, deposit, paid pilot).

• Advancement: they move to a concrete next step. Compliments and "keep me

posted" are zero-commitment → no signal.

Decision cue: after each conversation, name the currency spent and the next step secured. None → you got bad data, regardless of enthusiasm.

Audience-first loop (Embedded Entrepreneur)

Audience Discovery  →  Audience Exploration  →  Audience Transformation
(find where they      (embed; learn their      (build the product the
 already gather)        language & problems)     audience pulls you toward)

Embed in an existing community of the audience and earn standing before pitching anything. Building in public turns the audience into co-creators and a distribution channel. If the audience is unknown → start here before positioning or channels.

Decision cues

• "Validate my idea" → don't. Validate the problem via past behavior; the

idea is not the subject of the conversation.

• "They loved it!" → compliment = bad data. Ask what they did / will

commit. Deflect the compliment, dig for specifics.

• "Who do we even build for?" → Audience Discovery first; positioning and

channels depend on a known audience.

• Interview script provided → run the critique-and-rewrite procedure

([references/question-rewrite.md](references/question-rewrite.md)).

When this skill does NOT apply

• "We know the audience; how do we frame it" → growth-positioning.
• "We know who; which channel reaches them" → growth-traction-channels.
• "They're validated but don't stick" → growth-habit-loops.
• "How do we quantify/decide on the learning" → growth-experiments.

Anti-patterns (this skill's catalog)

1. Interviews-as-pitches (hub anti-pattern 6) — talking about your idea,

collecting compliments, founder talks more than the user.

1. Hypothetical/leading questions treated as evidence ("would you…?").
2. Compliments as validation — enthusiasm without a spent currency.
3. Skipping audience discovery — building, then hunting for who wants it

("build it and they will come").

1. Asking for opinions on the solution instead of facts about the problem.
2. No advancement — every conversation ends with "great, keep me posted."

Worked example (illustrative — not a product spec)

family-agent bad question: "Would you use an app that coordinates your family?" (hypothetical → bad data). Rewritten: "Tell me about the last time something fell through the cracks coordinating the family — what happened, what did you do, what did it cost you?" Commitment test: ask for a 30-min follow-up with their partner present; a yes (time currency + advancement) is signal, "love it, keep me posted" is not.

See also

• Hub: growth-foundations
• growth-positioning (frame the audience once it's known)
• growth-traction-channels (the embedded community is often the first channel)
• growth-experiments (turn discovered insight into a falsifiable test)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: williancesar
• Source: williancesar/saasoned
• License: MIT

Install and usage instructions live in the source repository linked above.