How do I know if my startup idea is any good?

The honest answer: you can’t know in advance. But you can replace the question you can’t answer (“is this idea good?”) with questions you can — and most founders never make that swap.

Stop asking the unanswerable question

Rob Fitzpatrick’s The Mom Test opens with a rule that sounds like a trick but is the whole method: you shouldn’t ask anyone whether your business is a good idea. Not because the answer is unknowable in principle, but because every human you ask — your mom, your friends, even your target customers — is incentivized to make the conversation pleasant. Compliments are free. The future is unfalsifiable. “I would totally use that” costs the speaker nothing and tells you nothing.

There’s a second, subtler problem, and it’s yours, not theirs. Daniel Kahneman calls it substitution in Thinking, Fast and Slow: faced with a hard question, your fast intuitive system quietly answers an easier one. “Is this idea good?” becomes “Do I enjoy thinking about this idea?” or “Did the last three people I pitched smile?” You will feel like you’re evaluating the business. You’re actually measuring your own enthusiasm and other people’s politeness.

Ask about the past, not the future

The Mom Test’s replacement questions are deliberately boring:

• “When did you last run into this problem? Walk me through it.” If they can’t recall a specific instance, the problem is hypothetical.
• “What did it cost you?” Time, money, embarrassment — quantify the pain.
• “What have you tried?” This is the killer. If they’ve never Googled for a solution, never duct-taped a spreadsheet together, never paid for a bad alternative — they will not buy yours. People solving real problems leave a trail.
• “Who else has this problem?” Real problems come with referrals attached.

Notice what these have in common: they’re about their life, in the past tense, with specifics. Lies live in the future tense; evidence lives in the past.

Count commitments, not compliments

Fitzpatrick’s currency test is the cleanest scoring system I know. After each conversation, ask: did the person give up anything that costs them something?

• Time: a scheduled follow-up with an agenda, an agreement to trial it for a week.
• Reputation: an introduction to their boss, their team, their budget owner.
• Money: a pre-order, a deposit, a signed letter of intent.

A pipeline full of “love it, keep me posted!” is a pipeline full of zeros. One person who introduces you to their boss outweighs twenty who called it cool.

Then run the outside view

Even with good interview data, Kahneman has one more correction. Your plan is built from the inside view — your specifics, your talent, your timeline. The outside view asks: of the last hundred startups roughly like this one, what happened? Base rates are brutal and informative: most ideas in crowded categories die of distribution, not product. You don’t consult the base rate to get discouraged; you consult it to find out which part of your plan needs to be unusual for you to beat it.

A practical closing ritual, stolen from Kahneman’s premortem: write the postmortem now. “It’s eighteen months from now and this idea failed — what happened?” If the failure story writes itself (“nobody actually had the problem,” “they had it but wouldn’t pay,” “we couldn’t reach them”) — that’s your interview script. Go collect evidence against the most plausible cause of death first.

The checklist

1. Never pitch during discovery. Talk about their life, not your idea.
2. Past tense, specifics, costs. “When did this last happen?”
3. No prior attempts to solve = no real problem.
4. Score conversations by commitments (time, reputation, money), not enthusiasm.
5. Check the base rate for your category — then identify what makes you the exception.
6. Premortem the idea and investigate the most likely cause of death first.

An idea that survives all six isn’t guaranteed to be good. But an idea that fails them is reliably bad — and finding that out in two weeks instead of two years is the closest thing to an edge a founder gets.