The AI Mom Test Coach skill for Claude

The problem the skill solves is more familiar than the solution. You've had ten conversations, everyone was enthusiastic, and nobody buys. It isn't bad luck, and usually not a pricing problem either. It comes down to the questions you asked, and nothing else in the chain. Rob Fitzpatrick wrote a book about it in 2013, The Mom Test, and his diagnosis is uncomfortably simple: people are nice. Ask someone whether they think your idea is good, and you'll get a polite yes. Friends say yes to be nice, not to buy.

The name comes from the test itself. A good question is one that even your mother can't answer with a polite lie, because you're not asking for her opinion but about her life. That sums up the whole method: the moment a question is about your idea rather than the other person's life, the answer is worthless, however enthusiastic it sounds.

What the skill does with that is translate it into something you can use this afternoon. You supply your questionnaire, your target audience and your learning question, and for each question you get a label, the rule it breaks, a rewritten version, three follow-up options per topic and a compliment check. It's written for founders validating a product, but the file itself names a broader group: sales discovery, market research, intake conversations and customer success calls. Wherever the other person might want to be nice, the same problem applies.

There's one thing you should know beforehand, and the skill says so itself: it judges questions, it doesn't simulate customers. It won't tell you what your target audience would answer, and it doesn't cite market figures you haven't supplied yourself. Better questions produce better data, not a better market. This skill belongs to our AI and automation service and sits alongside the rest of the skill library NL.

Compliments. What a great idea, I'd definitely look into that, this really has potential. They feel like validation and they aren't. All they say is that the other person likes you or wants to keep the conversation pleasant. This is so important in the skill that there's a separate, compulsory step for it: the compliment check, which you must never skip.

Fluff. This is the largest category and the hardest to spot, because fluff sounds like an answer. The file distinguishes three flavours: generalities (I usually do that), hypotheses (I'd definitely), and future promises (next month I'm going to). All three are about a life that hasn't happened yet. A conversation full of fluff fills a notebook and delivers zero facts.

Ideas from the other person. Could you also build in a calendar link, then I'd use it. A remark like that feels like free product management, but the file explicitly says it's not an instruction. Treat it as a signal and go looking for the problem underneath that idea. Building to the letter of customer requests is exactly how roadmaps go off the rails. If you want to order those requests later by what they're actually worth, you do that with the Kano Model Analyst skill or the MoSCoW Prioritisation Coach skill, not with the conversation itself.

The second part of the book is about commitment and advancement, and that's the part most often skipped in practice. A conversation only counts as a result once the other person hands over something of value: time, reputation or money. An enthusiastic reaction without a next step is not validation. That makes the third follow-up option in the skill, the commitment question, the most important of the three. Without that question, every conversation ends in politeness.

In diff form, with lines from the skill's example: what goes out and what comes back in its place.

A theory chapter follows, and that's unusual for a skill of this size. It explains where the name comes from, sets out the three rules, splits bad data into compliments, fluff and ideas from the other person, and closes with commitment and advancement: a conversation only counts as a result once the other person hands over time, reputation or money. That theory isn't there for show. Because the rules are in the file, Claude can name which rule each question breaks instead of just passing judgement.

The core is a fixed process of six steps. Step 1 is reading and recognising: the skill reads your list literally and feeds back in three lines what it infers from it, goal, target audience and learning question, so you can correct it straight away. Step 2 is asking follow-up questions when context is missing, with an explicit maximum of two questions. If the learning question is missing, it can't judge anything, because a question isn't good or bad in itself, only in relation to what you want to learn.

Step 3 is the diagnosis per question, with the three labels from FIG.02. Step 4 is rewriting every amber and red question, with four requirements for a good rewrite: it asks about the last time it happened and names a moment when doing so, it asks what the other person did and not what they thought, it doesn't name your solution, and it leaves room for the answer you'd rather not hear. The original question stays alongside it, deliberately, so you can see the difference and learn it yourself.

Step 5 is the three follow-up options per topic: facts, cost and commitment, set out in FIG.03. Step 6 is the compliment check, and it's compulsory. It scans your list for questions that invite compliments and describes how to respond if a compliment still crops up during the conversation: don't say thanks and move on, redirect to behaviour instead. There's a literal example sentence for this, along with the note that this step must never be skipped, not even if the rest of the questionnaire is green from top to bottom.

Besides the six steps, the file contains four further fixed parts. An input checklist of five points: the questionnaire, the target audience as specific as possible (with the note that Entrepreneurs is not a target audience), the learning question, the stage you're at, and whether you also want to sell something in this conversation, because then different rules apply. An output format with fixed headings, including a diagnosis table, a was-and-becomes table, the follow-up options per topic, the compliment check and a closing block that explicitly states what you know after this conversation and what you don't. A shortened example, the photographer example from FIG.01. And finally two chapters that guard the boundaries: eight things the skill never does, and six honest limits of the framework itself.

If you want to write a file like this yourself, or adapt this one to your own trade, that's the whole point: it's released under an MIT licence. Writing a SKILL.md NL explains how to build a file like this, from frontmatter to refusals block.

The method comes from Rob Fitzpatrick. The Mom Test, from 2013, is not academic work but practical knowledge from an entrepreneur, and the file says so itself. It's good tooling, not proof, and that difference is worth remembering. What makes the book unique is that it isn't about conversation technique but about the question itself: change the question, and the problem of polite answers largely disappears on its own.

The process around it comes from Steve Blank. The Four Steps to the Epiphany, from 2005, introduced customer development: customer research as a repeating cycle alongside product development, not a phase you tick off once. The Mom Test supplies the conversation form, Blank supplies that conversation's place in the process. Anyone wanting to map the bigger picture of his model ends up at the Lean Canvas Strategist skill or the Business Model Canvas.

Separating outcome from solution comes from Anthony Ulwick. What Customers Want, from 2005, is about the difference between what a customer wants to achieve and the solution they propose for it. That's exactly why the skill treats an idea from the other person as a signal and never as an instruction. The same line of thinking runs through the Jobs to Be Done Analyst skill and the Value Proposition Canvas NL.

The conversation technique comes from Steve Portigal. Interviewing Users, from 2013, is about the craft: letting silences fall, following up without leading, and accepting that it feels uncomfortable. The skill can improve your questions, but during the conversation you're the one who has to keep quiet. Rule three, and immediately the only rule no file can ever carry out for you.

The evidence comes from Jakob Nielsen and Daniel Kahneman. Nielsen wrote in 2001 that you shouldn't listen to what users say but watch what they do. Kahneman explained in Thinking, Fast and Slow, from 2011, why self-reporting about future behaviour is unreliable: people predict their own choices with a system that has different priorities from the system that eventually makes those choices. Together they answer the question of why you never ask whether someone would buy something. Anyone wanting to learn to spot those thinking errors more broadly can put the First Principles Thinker skill alongside this one: it takes assumptions apart the same way the Mom Test does with questions.

Even without installing the skill, you can use these principles to improve your next conversation: ask about the last time instead of the next time, don't name your solution, and end with a question that asks something of the other person.

After that, using it is simple: paste your questionnaire and add your target audience and learning question. The more specific those two are, the sharper the diagnosis, because a question can only be judged in relation to what you want to learn. If something's missing, the skill asks at most two questions and then carries on. If you get stuck with the installation itself, the full route is in installing Claude skills NL. Want to learn to set up this kind of work with your team, that's the subject of our AI training NL.

There are also situations where you're better off leaving it aside, and the file names them itself. For quantitative research with statistical claims, this is the wrong tool: for that you need a survey with a sampling design, and that follows entirely different rules. The same applies to conversations without a learning question. A demo or a negotiation is not a learning conversation, and a question asked there serves a different purpose.

Then there are the honest limits of the framework itself, and they're all in there. The Mom Test improves the quality of conversations, not their representativeness. Five good conversations with the wrong target audience remain five wrong conversations. Anyone who hasn't yet sharpened their target audience won't solve that with better questions.

Past behaviour predicts better than an opinion, but not perfectly. People who paid last year don't automatically buy again. A stronger signal than a hypothetical yes, certainly. Not a guarantee.

The framework is practical knowledge, not an academically validated method. Fitzpatrick is an entrepreneur who wrote down what he saw working. That makes it good tooling and not proof, and the file says so itself.

Sometimes you're talking to the wrong person. In procurement at large organisations, the person you're speaking to is often not the one who decides. Then you're measuring the behaviour of someone without the budget, however good your questions are. For that kind of process, the questioning technique from the SPIN Sales Conversation skill NL is a better addition, since it's built for complex sales with multiple decision-makers.

And one last remark from the file that's often misquoted: a good conversation usually lasts twenty to forty-five minutes, and shorter is fine too. Length says nothing about quality. The questions do.

Better questions make your conversations useful, but someone still has to be sitting across from you. If it's hard to get conversations at all, the real work lies before that: being found by the people looking for you. The free SEO scan shows in a few seconds where your site stands. And if you want to talk further about what AI can do for your business, from individual skills to full automation, we simply do that in a conversation. With good questions, of course.