The Mom Test for Customer Conversations
Plan and audit customer discovery using Rob Fitzpatrick's Mom Test to uncover past behavior, improve interview questions, and assess commitments.
Based on Rob Fitzpatrick's public method.
Not affiliated with or endorsed by Rob Fitzpatrick. Each step cites the public work it comes from: 8 sources.From flattering forecasts to past behavior
Replaces dashboard opinions with questions about the latest report, its preparation, and its recipient.
- Do you like the idea of a campaign reporting dashboard?
- Would you use a dashboard with more charts?
- Would you pay for a better reporting tool?
- Walk me through last Friday's report: what did you do?
- How did you prepare and send that report?
- Who received the report, and how did they read it?
Illustration from the skill's worked example, not a recorded run.
What it does
When your agent should reach for this skill.
Use this skill to prepare customer conversations, rewrite leading questions, review supplied interview notes, or decide what an interested prospect's actions actually establish. It produces a customer discovery brief: a practical conversation guide, an evidence review, or both. The aim is to learn facts that change a decision and, when an offer exists, identify concrete progress toward a purchase. 12
Apply it before committing development effort, when feedback sounds uniformly encouraging, or when repeated meetings produce enthusiasm without advancement. It also helps investigate feature requests and understand why customers retain apparently awkward workflows. Do not turn a conversation into a vote on the founder's idea. Specific behavior is more informative than predicted adoption. 35
This package draws on public interviews and essays rather than book text. The output schema, evidence labels, and review gates are agent adaptations. They organize Fitzpatrick's advice without claiming that he prescribed this exact reporting system. Source citations identify the public method; customer findings must point to supplied notes or observations separately.
The method
11 steps, restated from Rob Fitzpatrick's public work. Numbers link to the sources.
1Choose the deliverable and decision
Open templates/output-template.md. Select planning, evidence review, or mixed mode. State the customer group, product stage, and decision the work must inform. If context is incomplete, make narrow assumptions explicit; do not invent interview evidence.
2Separate learning from selling
Read the distinction in references/method.md. Identify whether success means understanding existing behavior or securing progress on an actual offer. In mixed work, keep those outcomes separate and identify where the conversation changes mode. 2
3Set three learning objectives
Follow references/method.md. Choose the uncertainties that could change the plan, including a threatening assumption. Record what an unwelcome answer would imply. Prioritize current behavior, failed alternatives, and purchasing constraints over a long questionnaire. 45
4Choose relevant participants and context
Use references/method.md to identify people with direct experience and realistic access. Friendly contacts can help practice; record that relationship and broaden the group. Keep a short exchange useful without forcing every encounter into a formal interview. 46
5Rewrite the question set
Apply the three filters in references/method.md: the customer's circumstances, a concrete past episode, and room to listen. Replace approval requests, forecasts, and imagined prices. Consult examples/worked-examples.md for before-and-after illustrations. 3
6Reconstruct an actual episode
Follow references/method.md. Capture the trigger, sequence, tools, participants, costs, outcome, and reasons. Prepare conditional probes rather than a rigid script. Ask to see relevant work when practical; distinguish a reported event from something observed. 4
7Recover from praise and investigate requests
Use references/method.md. Return from compliments to a concrete event. For each feature request, uncover the goal and current workaround before considering implementation. Mark product pitches that may have influenced later answers. 5
8Test fit with existing habits
Consult references/method.md. Identify where the proposed product should begin and end, what tools it should cooperate with, and what behavior must change. A painful problem does not establish that the suggested solution fits the customer's world. 7
9Choose the next form of evidence
Use references/method.md. Decide whether the remaining uncertainty needs another conversation, an observable prototype task, or a purchase step. Experience-dependent value often needs an earlier prototype. Learning reduces particular risks; it does not create certainty. 46
10Assess commitments when an offer exists
Follow references/method.md. Specify a meaningful ask involving time, reputation, or money. Record requested, agreed, and completed actions separately. Explain what each action demonstrates and what remains unproven, especially purchasing authority. 24
11Synthesize, review, and iterate
Complete templates/output-template.md, linking conclusions to concrete evidence and preserving contradictions. Apply checklists/review-checklist.md. Repair failed checks and state the next decision-relevant action. Use reflection to explain what changed, rather than merely collecting highlights. 8
Worked example
How the method plays out on one case.
1: agency reporting product
Weak first draft
A founder proposes a campaign dashboard for small agencies. The interview guide asks:
- Would a dashboard save you time?
- Which charts would you want?
- Would you pay EUR 49 per month?
The invented agency owner responds enthusiastically and suggests several charts. The founder's draft conclusion is that demand and pricing are validated and the chart builder should be developed next.
Apply the method
Set three objectives: reconstruct the latest client report, understand the recipient's needs, and learn the last relevant software purchase process. Replace the opening with a request to walk through the most recent report. Probe the sequence, what took time, why that format was chosen, and what happened after delivery. This moves attention from the proposed product to past behavior. 34
In the invented follow-up, the owner describes spending 90 minutes exporting campaign data, copying it into a branded document, and emailing clients. Clients receive an attachment because they rarely open tool accounts. The owner wants charts to make that attachment readable, rather than to give clients a new dashboard.
Investigating the goal behind the chart request reveals a different design opportunity. Preserving delivery through the customer's existing tools also avoids imposing unnecessary workflow changes. 57
Improved result
Decision: Test document generation before building a client dashboard.
Evidence: Reported past behavior: the latest report required exports, manual document preparation, and email delivery. The 90-minute figure is the owner's account, not a measured result. The owner reports that client login adoption has been poor.
Weak signals excluded: Enthusiasm and imagined willingness to pay EUR 49 do not establish adoption or price acceptance. The feature request identifies a desired result to investigate, not a roadmap instruction.
Next experiment: Try a narrow prototype using a real report, measuring whether it creates an acceptable client attachment with less preparation work. Treat the prototype as a solution hypothesis rather than a proven consequence of the problem.
Commitment: The owner agrees to a 30-minute session next Tuesday using an existing export. Status: agreed. Attendance, actual testing effort, payment, and purchasing authority remain unresolved. Time is a stage-appropriate signal, with a narrower meaning than a purchase. 46
Rules that made the difference: Specific episodes replaced forecasts; the feature's purpose was investigated; the product boundary respected existing delivery; and the agreed test was kept separate from completed action. 357
1 more worked example in examples/worked-examples.md.
What's in the package
7 files, pinned at d5668a9. Open any file on GitHub.
SKILL.mdEntry point the agent loads: when to use the skill and the procedure.6,283 bytesLICENSEApache-2.0 license.10,756 bytesreferences/method.mdThe method in full, step by step, with citations.12,180 bytesreferences/sources.mdEvery public source the method is built from.3,505 bytesexamples/worked-examples.mdWorked examples showing the method applied end to end.5,875 bytestemplates/output-template.mdThe output format the agent fills in.5,935 byteschecklists/review-checklist.mdChecks the agent runs before handing back the result.4,745 bytes
Sources
The public work this skill is built from. Read Rob Fitzpatrick's originals.
- The Mom Test, official websitemomtestbook.com
- I'm Rob Fitzpatrick. I wrote The Mom Test book, went through YC (s07), learned how to do sales as a techie, and live off full passive income. AMA!indiehackers.com
- Learning Leader episode 451: Rob Fitzpatrick, How to Talk to Your Customers, How to Ask Great Questions, and How to Be Usefulpodscripts.co
- How to Talk to Customers and Validate Your Business Idea (When Everyone Is Lying to You), Everyone Hates Marketers interviewstfo.io
- EP 231: The Mom Test with Rob Fitzpatrick, The Innovation Showtheinnovationshow.io
- Rob Fitzpatrick: How an entrepreneur's journey yields compound interest, Get The Audience, episode 23gettheaudience.com
- Workflows and worldviews / Deciding where your product ends / Idea traps, Rob Fitzpatrickrobfitz.com
- If you aren't taking notes, stop reading / Slower process, faster progress / Healthy externalities, Rob Fitzpatrickrobfitz.com
The skill file also covers: Files in this skill. Read the full SKILL.md.
Edge wrote this skill from Rob Fitzpatrick's public writing and talks listed above. It is not affiliated with or endorsed by Rob Fitzpatrick. If something misrepresents the method, write to [email protected] and we will correct or remove it.