🤖 Crush customer interviews using the Mom Test
Everyone talks about the importance of customer interviews, but few seem to do it — and those who do tend to do it poorly. Let's fix that, with the help of the Mom Test and our AI co-founder.
Hey friend 👋
Interviewing customers is critical to ensuring you’re solving a problem worth solving.
And yet, doing it badly costs time and money, and kills companies.
But there’s a simple test you can use to ensure you remove bias and uncover the real opportunities in their problems. It’s called the Mom Test, and your AI co-founder is here to help.
Let’s go 👇
There’s a weird paradox in the startup world:
Everyone talks about the importance of customer discovery, but few seem to do it.
It seems every week I meet a founder who has some app that they think is pretty good and they want help finding someone who can market it for them.
Unfortunately, that’s not how startups work.
And when I do meet founders who conducted customer interviews, they tout stats with gusto: “78% of people we surveyed want an app like this!”
It’s always unfortunate to meet a cautionary tale... but they’ve made two critical mistakes:
They’ve tried to turn qualitative research into quantitative research.
They asked people a question rife with bias.
Let’s break it down:
It’s not a numbers game.
Customer discovery is qualitative research — we explore markets to gain an understanding of the underlying reasoning, opinions, and motivations of the people involved. This usually begins with understanding their pain, because pain is opportunity, but quickly moves into what drives buying behaviour.
In customer discovery, our goal is to gather non-numerical data (such as through interviews) from small, non-random samples of a population to provide depth of understanding. And we can’t analyse that data using statistics. Qualitative analysis is interpretive and thematic — we’re after patterns or themes.
Contrast that with quantitative research, where the goal is to test hypotheses and examine relationships between variables using measurable data. We study pre-determined variables with large, random samples in search of outcomes with statistical significance.
But in customer discovery, we don’t yet know enough about our customers’ world to pre-determine those variables!
The key takeaway?
Aim to have conversations with customers — not to run surveys of them.
Now let’s talk bias:
Rule #1: stop fooling yourself!
Few things are as relevant to early stage startups as this advice on science from physicist Richard Feynman:
The first principle is that you must not fool yourself, and you are the easiest person to fool.
While there are innumerable ways to introduce bias into our customer interviews, the easiest way to get bad data is to ask questions that open up what I call a Needs Gap.
Let’s say you’re at a restaurant with some friends.
The food has been thoroughly mediocre, and it’s been a bit noisy to boot. You didn’t have bad experience, but you aren’t rushing to come back. It’s meh.
At the end of the meal, the waiter comes up to you and asks, “how was everything?”
What do you say?
“It was good.”
In other words, you lied... and right to his face. But it’s even worse than that!
Because we know the waiter doesn’t care that much. This isn’t the French Laundry. This probably isn’t his career — it’s just his job, and he works on tips.
And yet... you lied.
Let’s add the twist: when you talk to customers about your startup, they know you care.
How likely are they to tell you truth?
So when you say 78% of respondents said they’d buy your app, I say:
All this means that the thing you fundamentally want to learn (will they buy) is the one thing you can’t ask!
So what to do?
Mother knows best (except when it comes to your startup).
Believe it or not, not being able to ask direct questions isn’t a handicap — it’s a superpower!
If you ask questions that pass the Mom Test, you’ll gather an extraordinary amount of valuable information — more valuable than just learning if they’d buy.
It’s like this:
You go and ask your mother, “hey, mom, what do you think of my startup idea?” How will she respond?
Oh, it’s amazing, honey! I can see it! This is the next big thing! You’re going to be as big as Elon Musk! I’m so proud of you!
Because just the like the waiter to the patron, you created a Needs Gap with your mom: you created a gap between what you need to hear, and what they think you need to hear. Your mother thinks you need encouragement. Patrons think the waiter needs the nicety.
And everyone wants to avoid the unnecessary conflict.
You can remove the bias by ensuring that your question creates no Needs Gap — that there’s no difference between what you need to hear and what they think you need to hear.
It’s called the Mom Test:
Ask a question in such a way that even your own mother has no choice but to tell you the truth.
In other words, close the Needs Gap — and remove the bias.
You can’t ask your mother if she likes your idea, or if she’d buy... but you can ask her to tell you about a time she had a particular problem.
You can’t ask her how much she’d pay for your app... but you can ask her how much she’s paying for the competition, or how much time or money she’s losing because she hasn’t yet solved the problem.
Here’s how it works:
It’s qualitative. Questions that pass the Mom Test are designed for qualitative analysis, and not quantitative analysis. They are open ended, rather than yes/no or multiple choice, and are designed to elicit stories from customers — rather than short answers. Perhaps most importantly, the questions ask about the present or the past, but never about the future, because statements about potential future behaviour are unreliable.
It’s unbiased. Questions that pass the Mom Test are asked in neutral ways, and do not lead customers in any particular direction. Further, they are phrased in ways that encourage honest responses, and never give customers the impression that there are any right or wrong answers.
It’s customer-centric. Questions that pass the Mom Test are centered on the customer’s life, and not on your startup idea, your product, or any of your assumptions. They do not seek validation of any idea, but aim to discover genuine insights. They are asked with the humility necessary to accept that we don’t know in advance where we’ll find those insights. And finally, the questions engage customers emotionally — we see the whites of their eyes when they respond.
We can distill all of this to a few core principles:
The 9 constraints of the Mom Test
To pass the Mom Test, a question must be:
Centered on the customer’s life, and not on your startup idea.
Seeking genuine insights, rather than validation of any idea.
Engaging emotionally with customers.
Open ended, rather than yes/no or multiple choice.
Eliciting stories from customers, rather than short answers.
About the present or the past, but never about the future.
Phrased in ways that encourage honest responses.
Neutral, and do not lead customers in any particular direction
Eliciting responses that are specific and tangible, rather than vague or broad.
This is a lot to keep in mind, and it takes practice.
Fortunately, we know someone with a lot of practice...
Bring on the AI co-founder!
And let’s ask it to critique our discovery questions.
This prompt takes your list of customer interview questions and provides a PASS/FAIL grade for each of them, along with an explanation of why it got that grade. For a question that doesn’t pass the mom test, it will also provide 3 to 5 suggested alternatives for you to consider.
Unlike every other prompt we’ve done so far in Founding with AI, we’re are NOT giving our AI co-founder any knowledge or insight into what we’re working on — no elevator pitch, no startup core, nothing. We don’t want to introduce bias into the questions, and the AI doesn’t need to understand the intent behind a question in order to validate it.
However, in a future edition, we’ll talk about using AI to come up with customer discovery questions for what you want to learn. Drop a comment if you want this higher in my queue!
As usual, this prompt should work in both free and paid ChatGPT, as well as in Claude and Gemini. My screenshots are from ChatGPT 4o.
Here it is:
# ROLE
You are an experienced startup founder and advisor with a track record of success. You are expert in customer discovery, validation, and market research.
I am a startup founder with an idea for a product, and I need to validate through customer interviews that the problem is worth solving and that the solution resonates.
# TASK
I am going to give you a list of questions that I want to ask customers, and you’re going to ensure they pass the Mom Test.
For each question, provide a PASS/FAIL determination and a one-sentence rationale for that determination. If the question fails, provide 3 to 5 alternative questions that might get me what I want to learn.
# OUTPUT
Return a bulleted list of my questions. Preceding each question, provide “PASS” or “FAIL”, followed by a colon. In a single sub-bullet, provide the one-sentence rationale for that determination. When a question gets fails, provide a second sub-bullet with “Potential alternatives”, and list each suggestion as a sub-bullet to that sub-bullet.
For example, given this question list from me:
- Tell me about a time you felt frustrated managing your to-do list.
- How much would you pay for this app?
You would produce this output:
- PASS: Tell me about a time you felt frustrated managing your to-do list.
- This question passes because it is open-ended, asks the customer to tell a story, and is about past events.
- FAIL: How much would you pay for this app?
- This question does not pass the mom test because it is about future behaviour, and is subject to bias.
- Potential alternatives:
- How much do you currently pay for task management apps?
- Have you ever lost time, money, opportunities, etc., because of how you manage to-do list?
# The Mom Test
The purpose of the Mom Test is to remove as much bias as possible from our questions by ensuring questions meet the following criteria:
- They are designed for qualitative analysis, and not quantitative analysis.
- They are open ended, rather than yes/no or multiple choice.
- They typically elicit stories from customers, rather than short answers.
- They ask about the present or the past, but never about the future, because statements about potential future behaviour are unreliable.
- They are neutral, and do not lead customers in any particular direction
- They are phrased in ways that encourage honest responses.
- They elicit responses that are specific and tangible, rather than vague or broad.
- They are centered on the customer’s life, and not on your startup idea.
- They do not seek validation of any idea, but discovery of genuine insights.
- They engage customers emotionally.
# List of Questions =
{bulleted list of questions}
As a test, I gave it four questions I want to ask customers to validate our continuing example of Todorly, a task management app for financial professionals:
Would you find a task management app designed specifically for financial professionals useful?
Tell me about the last time you missed a deadline or task. What happened, and how did you handle it?
How much would you be willing to pay for an app that helps you manage your tasks better?
What tools or systems do you currently use to stay on top of your work? What do you like or dislike about them?
Questions 1 and 3 should fail, while questions 2 and 4 should pass.
Here’s the full output from my AI co-founder:
Let’s take a closer look at the questions that failed:
“Would you find a task management app designed specifically for financial professionals useful?”
It’s incredibly common for customer discovery “surveys” to include a question like this, which the founder then touts as evidence:
78% of people we surveyed want an app like this!
Sounds like amazing validation, except it’s not!
Not only did you not understand the assignment, and try to turn qualitative research into quantitative research, but you also asked a question that’s both rife with bias and about the future — so there’s no cost in signaling to you that they want it when they don’t.
When I ask founders what they’re trying to learn from a question like this, they’re often taken aback: “I’m trying to learn that! lol”
But we’re not there yet. That’s validation, not discovery, and it requires different tools.
In discovery, you want to know whether the problem is there, and whether they are using comparable solutions to try to solve it. If they’re using comparable solutions and are still feeling the core problem you’re trying to solve, that’s good evidence you might be on to something.
Now let’s look at the suggested alternatives from our AI co-founder:
Can you tell me about the last time you struggled to manage your tasks or deadlines? What were the consequences?
What are the specific challenges you face when managing tasks related to financial work?
How do you currently manage tasks related to financial work, and what issues, if any, have you encountered?
The first two two questions open the door for the customer to express that they have the problem (without biasing them toward the answer) and the third sees if they’re using a comparable solution (again, without biasing the result).
Boom.
Don’t stop at PASS/FAIL
The goal of this exercise is principally to get to a series of questions that all pass the Mom Test, but it’s also to help you think through the kinds of information you’re collecting.
If the suggested alternatives are completely unrelated to what you’re trying to discover, it’s time to take a step back and put further thought into what you’re trying to learn — and why.
And remember, like all our past conversations with our AI co-founder, it’s a conversation — an iterative back-and-forth. If you’re not getting the results you’re looking for, let your co-founder know. Ask questions, provide new questions, and keep at it until you have a set of questions that don’t open a Needs Gap and will get you the unbiased results you need to make decisions.
And then, go talk to customers.
Until next week,
—jdm
PS: If you enjoyed this issue of Founding with AI, I’d be super grateful if you took a few seconds to share it with a founder friend. Referrals are a huge help in growing a newsletter — and a huge help in helping more founders realise the awesome power of AI!
And, as always, I’d love to hear your thoughts. If you have ideas how to make this newsletter even better, have suggestions on future topics, or even just want to share what you thought of this issue, just hit reply or leave a comment.