Your friends say they'd use it. Should you believe them?

Because a yes from a friend is a social favor, not a buying decision. The question they are really answering is "do I support you?", and the answer to that is almost always yes.

Three things are broken in what friends tell you:

• The sample is skewed. These are people who already like you.
• The question is imaginary. "Would you use this?" measures imagination, not behavior.
• The cost is zero. A yes commits them to nothing, while a no makes the next dinner awkward.

Rob Fitzpatrick wrote a whole book, The Mom Test (2013), on this pattern. People who care about you tend to tell you your idea is good. It is politeness, not malice.

The tell is how specific they get. A friend who would really pay asks about price, timing and what it replaces. A friend being kind says "that's so cool" and changes the subject.

The stakes are real. In CB Insights' report The top 9 reasons startups fail (5 March 2026), poor product-market fit appeared in 43 percent of 385 failed investor-backed startups. That counts only failures where the causes could be identified. Product-market fit means enough people want what you built to pay for it or keep using it. Failures could have several causes at once.

That study covers investor-backed companies, not every side project.

Watch what people do, not what they say. Signs that someone may really pay: they ask about price or timing, or already spend money or hours on the problem. Stronger still, they agree to a preorder or deposit, or introduce you to someone else who has it. Praise for the concept is encouragement, and encouragement is not evidence.

Useful criticism tests the structure of the idea: the market, the timing, the competition, the money, and whether you are the right person to build it.

Founders at this stage often suspect the people around them are softening the read. For a way to rank what you hear, from praise up to real use, see how to tell a good idea from excitement.

It can be. Tools that build apps without writing code mean far more people can ship something in a weekend.

When building takes a weekend, skipping the question feels rational. In our view, the cost of building the wrong thing was never mainly the code. It is the lost months, the launch that lands on silence, and the slow admission afterward. Fast building does not fix that. It gets you to the disappointment sooner.

You have a few options, and each has a catch:

• Strangers in founder communities are more honest than friends, but you often get quick hot takes rather than a structured argument.
• An experienced investor can give a sharp read, but few founders at the idea stage have one on call.
• People who have the problem give the best evidence of all. Use interview questions that get honest answers, and find them where they already talk.
• A tool built for evaluation. Roast My Idea is set up to test an idea from several angles. It returns a verdict (Build it, Sharpen it or Kill it) with how confident it is, the biggest flaw it sees, and one test to run this week. It goes after the idea, never you.

What about a general chat assistant? It can give a useful critique, but it answers the question the way you frame it, and an excited founder rarely frames a hard question. A tool built for one explicit evaluation saves you designing the critique yourself.

No. An AI saying yes is not the market saying yes. Its praise reflects how you asked, not whether anyone will buy. The yes that counts comes from someone who has the problem and gives up time or money for your answer.

Here is an illustrative example, not a real founder or a real tool run. Say a solo developer wants to build a tool that judges whether a game will be fun from a written description.

A tough read might say Sharpen it, not Kill it. The flaw is structural: the tool judges the pitch, but fun lives in playing the game, so the buyer cannot trust the verdict. That makes it a product problem, not a marketing problem, so no amount of posting fixes it. A fair test: message ten indie developers who shipped a demo. Ask what they did the last time they doubted their game, and what it cost them.

If what you need right now is encouragement to keep going, a tough read is the wrong tool. Once you have paying customers, their behavior beats anyone's opinion, ours included.

There is also a real case for friends. Sometimes friends are your target customers, for example if you build for your own hobby group. Then their behavior counts, though their praise still does not. The missing fact is always the same: will people who have the problem act, not just agree?

Go back to three friends who said they would use it. Ask each for one small commitment instead of an opinion: a preorder, an introduction to someone with the problem, or an hour to try a rough version. Count how many say yes.