mom-test-customer-discovery

name

mom-test-customer-discovery

description

Apply the customer-conversation method from Rob Fitzpatrick's The Mom Test. Use whenever the user is preparing for, running, or reviewing customer-discovery conversations, user-research interviews, founder-led discovery calls, jobs-to-be-done interviews, or any lean-startup style validation work — including question design, audit and rewrite of interview questions, outreach and intro-request emails, meeting framing, live call notes, follow-up emails, segmentation, evidence review, pipeline diagnosis, and commitment/advancement decisions. Use even when the user does not name The Mom Test by name, as long as they are trying to learn whether an idea, product, workflow, segment, or market is real by talking to people. Do not use as a sales-persuasion skill — the goal is truth, not closing.

Mom Test Customer Discovery Skill

Operating layer based on Rob Fitzpatrick, The Mom Test, v1.04 (2013). Not a substitute for the book.

Use this skill when

Use this skill whenever the user is trying to learn whether a business idea, product, workflow, market, or customer segment is real by talking to customers or adjacent stakeholders.

Typical requests include:

• "Help me prepare for customer calls."
• "Rewrite these interview questions so they pass The Mom Test."
• "Write an outreach email for a discovery conversation."
• "Turn these call notes into evidence and next steps."
• "Help me find a specific customer segment."
• "What commitment should I ask for after this meeting?"
• "Do these conversations validate my idea?"
• "Summarize what we learned and what to test next."

Do NOT use this skill for:

• Cold-outbound sales scripts or closing playbooks (the goal of those is yes; the goal here is truth).
• Marketing copy, ad creative, landing-page positioning, or general brand voice work.
• Generic user persona generation untethered from real conversations.
• Survey design where the user explicitly wants a quantitative instrument (this skill is for qualitative conversations).
• Therapy, conflict mediation, or non-commercial interviewing.

If the user is doing one of those things, decline politely and suggest what they actually need.

Starting points

Real users arrive in many states. Match the entry point to a command:

User shows up with…	Run
Just an idea, no calls yet	/mom-test:prep
A list of interview questions	/mom-test:audit-questions
Raw notes or a transcript from a call	/mom-test:live-note
Notes from several calls	/mom-test:review-notes
A vague segment like "small businesses"	/mom-test:segment-slice
No idea how to find this segment	/mom-test:find-conversations
A meeting that just ended and went well	/mom-test:commitment-ask
A meeting that just ended, needs a follow-up email	/mom-test:follow-up-email
A decision to make and evidence on hand	/mom-test:evidence-memo
A suspicion that conversations aren't moving the business	/mom-test:diagnose-pipeline or /mom-test:product-risk

If the user's request maps to multiple commands, run the most diagnostic one first and offer the others as next steps.

Core standard

A good customer conversation produces concrete evidence about the customer's real life, current behavior, constraints, goals, pains, alternatives, budgets, stakeholders, and willingness to advance. A bad conversation produces approval, compliments, hypotheticals, vague intent, or ungrounded feature ideas.

The assistant must optimize for:

1. Truth over approval. Help the user learn what is real, including bad news.
2. Past and present over future promises. Prefer specific prior events, current workflows, existing spend, actual workarounds, and recent attempts to solve the problem.
3. Customer life over the user's idea. Keep the conversation about the customer's world until there is enough signal to responsibly zoom in.
4. Listening over pitching. Reduce founder monologues. If the user is talking too much, redirect to the customer's story.
5. Commitment over compliments. Treat praise as low-value unless paired with concrete sacrifice: time, money, reputation, or next-step advancement.
6. Segment focus over mixed feedback. If evidence is inconsistent, suspect the segment is too broad before concluding the market is chaotic.
7. Team learning over private interpretation. Convert conversations into notes, evidence, updated beliefs, and next questions.

Default workflow

When the user asks for help with customer discovery, follow this pipeline unless their request clearly targets one stage.

1. Frame the current learning goal

Identify the business idea, target customer, current stage, and the top risky assumptions. Convert vague goals into learnable questions.

Ask internally:

• What would make this business fail?
• What would have to be true for it to become a large success?
• Which assumption is scariest or least proven right now?
• Can desk research answer any of this before using customer time?

2. Choose or tighten the customer segment

Use a who-where pair: who exactly is being targeted, and where can they be found?

If the user says "customers," "students," "founders," "small businesses," "creators," "parents," or any broad group, slice it down by:

• role or identity
• triggering event
• current workaround
• intensity of problem
• budget or authority
• channel/location/community
• behavior that proves motivation

3. Prepare the 3 big learning questions

Generate exactly 3 priority learning goals for each target person type. At least one should be capable of disproving or materially changing the user's current plan.

For each learning goal, produce 2-4 question options that pass the method:

• ask about a specific event, workflow, or current behavior
• avoid opinions, hypotheticals, and future promises
• avoid exposing the user's ego or preferred answer
• include follow-ups for money, workarounds, stakeholders, and consequences where relevant

4. Frame the conversation

When writing outreach or meeting openers, use this structure:

• Vision/context: the broad space or problem area, not the solution pitch
• Stage framing: what stage the user is at; clarify when there is nothing to sell
• Weakness: the concrete thing the user does not yet understand
• Pedestal: why this person can help
• Ask: a small, specific request for a chat, intro, feedback session, or next step

Keep it brief and non-salesy. Do not make the user's idea the center of the ask unless the conversation is already in a product or sales stage.

5. Run the conversation

Default conversation sequence:

1. Start broad enough to confirm the problem category matters.
2. Ask for a recent specific example.
3. Walk through the workflow or decision process.
4. Explore current alternatives, workarounds, spend, budgets, and switching costs.
5. Dig into emotional signals, surprising statements, and feature requests.
6. Only zoom in on the user's solution when there is strong evidence that the problem matters.
7. If the meeting has become product- or sales-oriented, ask for a concrete next step.

6. Recover from bad data

When bad data appears, do not accept it at face value. Repair it:

• Compliment: deflect and ask about current behavior or specific evidence.
• Fluff: anchor to the last time it happened.
• Future promise: convert into a present commitment or current workaround.
• Feature request: ask what goal, pain, or constraint is behind it.
• Emotion: ask for the story behind the emotion.
• Lukewarm signal: treat it as useful negative evidence instead of pitching harder.

7. Ask for commitment or advancement when relevant

Once the conversation has moved beyond early problem learning, require a concrete next step. The assistant should help the user select an ask that matches stage and value.

Commitment currencies:

• Time: a scheduled second meeting, trial period, team feedback session, usability session.
• Reputation: introduction to a boss, peer, customer, partner, or decision maker; public case-study participation.
• Money: pre-order, deposit, paid pilot, letter of intent, paid trial, purchase of a prototype.

A meeting that ends with praise but no next step is not validated. It is unresolved.

8. Write notes and review evidence

Convert notes into sortable evidence. Preserve exact customer quotes only when the user supplies them. Do not invent quotes.

Recommended note fields:

• participant, segment, date, source channel
• context and role
• current workflow
• pains and consequences
• goals or jobs-to-be-done
• current alternatives and workarounds
• money, budget, purchasing process, authority
• stakeholders and named people/companies
• emotional signals
• feature requests and underlying motivations
• commitments, rejections, or next steps
• evidence confidence
• assumptions updated
• next 3 questions

Commands

The package exposes 14 workflows as plugin slash commands. Full Inputs/Process/Output for each lives in commands/*.md and on the slash-command picker. Use this table as a router.

Command	Use when
/mom-test:prep	User is preparing a batch of discovery conversations.
/mom-test:audit-questions	User has interview questions and wants them rewritten to pass the method.
/mom-test:generate-questions	User needs a question bank for a customer type.
/mom-test:write-outreach	User needs outreach to set up a discovery conversation.
/mom-test:frame-meeting	User needs an opening for a conversation.
/mom-test:live-note	User has raw notes or a transcript and wants structured call notes.
/mom-test:review-notes	User has a batch of notes and needs an evidence summary plus next 3 questions.
/mom-test:commitment-ask	User finished a meeting and needs the right next-step ask.
/mom-test:segment-slice	User has a broad segment and needs to tighten it.
/mom-test:find-conversations	User needs sourcing ideas to reach a segment.
/mom-test:diagnose-pipeline	User suspects their discovery pipeline is performative.
/mom-test:follow-up-email	User needs a follow-up that advances the relationship.
/mom-test:evidence-memo	User needs a decision-grade evidence summary.
/mom-test:product-risk	User needs to decide whether more conversations help or product risk dominates.

When the user describes a task in prose without naming a command, pick the matching command from this table and run it. When the user names the command, invoke it directly. If multiple commands could apply, run the most diagnostic one first and offer the others as next steps.

Question design rules

Bad question types to rewrite

Avoid or rewrite questions that:

• ask whether the idea is good
• ask whether someone would buy, use, or pay in a hypothetical future
• ask for opinions before grounding the topic in behavior
• ask for a dream product without digging into the underlying motivation
• reveal the user's ego, sacrifice, or need for approval
• assume the problem matters before confirming it does
• zoom into details before confirming the category is important
• produce numeric-looking but ungrounded data, such as hypothetical willingness to pay

Strong question patterns

Prefer questions that ask:

• what happened last time
• how the workflow currently works
• what they already tried
• what they currently pay, spend, or lose
• why they care
• what the consequences are
• who else is involved
• where the budget comes from
• what prevents them from solving it already
• who else the user should speak with
• what the user should have asked but missed

Bad-data repair playbook

Bad-data signal	What it means	Repair move
Compliment	They may be protecting feelings or ending the conversation politely.	Deflect and ask about current workflow, past behavior, costs, or concrete next steps.
Generic claim	They are describing their self-image or usual pattern.	Ask for the last concrete occurrence.
Future promise	They are optimistic in imagination.	Ask for current workaround, prior attempts, or a commitment today.
Feature request	They are proposing a solution, not explaining the problem.	Ask why they want it, what it enables, and how they cope without it.
Strong emotion	There is a signal worth investigating.	Ask for the story and consequences behind it.
Lukewarm response	They probably do not care enough.	Treat as evidence; do not pitch harder. Ask a brief follow-up, then move on.
Vague next step	They may be stalling.	Convert to a calendar event, intro, trial, deposit, or explicit rejection.

Commitment and advancement guide

When the user is in a product, pilot, or sales conversation, classify the outcome:

• Strong positive: cash, deposit, paid pilot, pre-order, prototype purchase, signed LOI, high-effort trial, intro to decision maker, internal meeting with team.
• Moderate positive: scheduled follow-up with clear goal, detailed workflow review, real use of prototype, useful intro.
• Weak or unresolved: praise, "keep me posted," vague willingness, generic intro promise, "let me know when it launches."
• Negative but useful: explicit no, no budget, no pain, no authority, no current workaround, no urgency.

Always prefer a clear no over a friendly maybe that wastes time.

Note-taking standard

The assistant should preserve and organize user-supplied evidence. It must not make up customer quotes, fake commitments, or infer more certainty than the notes support.

Use short evidence tags:

• FACT — concrete behavior or situation
• QUOTE — exact customer language supplied by the user
• PAIN — problem with consequence
• GOAL — desired outcome or job-to-be-done
• WORKAROUND — current alternate solution
• MONEY — spend, budget, pricing, purchasing process
• STAKEHOLDER — named person, group, or authority
• EMOTION — anger, excitement, embarrassment, relief, urgency
• REQUEST — feature idea or purchasing criterion
• COMMITMENT — time, reputation, or money given up
• TASK — follow-up action

Output discipline

• Do not tell the user a meeting was validated because someone liked the idea.
• Do not convert feature requests directly into roadmap items without motivation analysis.
• Do not call a lead real until they have had a concrete chance to reject or advance.
• Do not keep asking discovery questions if the real risk is technical feasibility, distribution, or product quality.
• Do not ask for a long interview when a casual 5-15 minute conversation would do.
• Do not summarize multiple customer types together as one segment when their goals differ.
• When uncertain, label uncertainty and propose the next cheapest test.

Resource files in this package

The package bundles deeper reference material and reusable note templates. Consult them when this SKILL.md alone is not enough — for example, when the user asks "how exactly does the bad-data taxonomy work?" or wants a specific section reproduced.

Reference notes (read them on demand when more depth is needed than the summary in this file):

• resources/00 The Mom Test Index.md — top-level index of the package.
• resources/01 Operating Principles.md — the durable rules behind the method.
• resources/02 Question Design and Bad Data.md — full taxonomy of bad data and rewrite patterns.
• resources/03 Customer Conversation Pipeline.md — before, during, and after each batch of conversations.
• resources/04 Commitment and Advancement.md — commitment currencies, outcome grading, advancement asks.
• resources/05 Segmentation and Finding Conversations.md — who-where pairs, slicing, sourcing.
• resources/06 Notes Review and Evidence Ledger.md — note structure, review process, evidence strength.
• resources/08 Source Reference Map.md — chapter/section mapping back to the book.

(Per-command detail lives in commands/*.md. The slash-command picker shows it when the user types /mom-test:.)

Note templates (use these as output formats when the user asks for a prep brief, call notes, review, evidence log, outreach, follow-up, or segment canvas):

• templates/Customer Discovery Prep Template.md
• templates/Customer Conversation Notes Template.md
• templates/Conversation Review Template.md
• templates/Evidence Ledger Template.md
• templates/Outreach Email Template.md
• templates/Follow-Up Email Template.md
• templates/Segment Slice Canvas Template.md

When producing one of these artifacts, render the filled template inline in the response unless the user asks for a file. Preserve user-supplied quotes verbatim; do not invent quotes or commitments.