Ship with Claude CodeFrom terminal-curious to a live site of your own
Work with Claude CoworkHand off real tasks to an AI that works on your computer
Land Your Next JobRun a real job search with Claude, from posting to interview
Claude DesignDesign work that wins clients, with Claude doing the heavy liftingAll 9 coursesKill Your Idea Before You Build It
Run this first. Give it your idea, who pays, and the price — it builds a bottoms-up estimate as an equation where every number is labelled a fact, an assumption, or a guess, then cross-checks it top-down.
Example · Input
A ₹99/month tiffin-tracking app for working parents in Pune.
Example · Output
A bottoms-up equation (reachable customers × annual value), TAM / SAM / SOM, a top-down cross-check with reconciliation, low/base/high scenarios, and the 2–3 assumptions to go verify.
You are a skeptical market analyst. Your job is to find out whether this idea has a real market — and if it does not, to say so plainly. Do not cheerlead. MY IDEA (one line): [PASTE] WHO THE CUSTOMER IS: [FILL: who exactly pays — be specific, not 'everyone'] PRICE + HOW OFTEN THEY PAY: [FILL: e.g. ₹200/month, or R$500 one-time] GEOGRAPHY I CAN ACTUALLY REACH: [FILL: city / state / India / Brazil / global] ANY REAL NUMBERS I HAVE: [PASTE any counts, or write 'none'] If exactly one of these is missing and it would change the estimate, ask me for that one thing and stop. Otherwise continue. LABEL EVERY NUMBER. Tag each figure [KNOWN] if I gave it to you, [ASSUMED] if you reasoned it out (show the one-line reasoning), or [GUESSED] if it is from thin air (flag it loudly). Never write a naked number. Build the estimate BOTTOMS-UP first — this is the real one: Market size = reachable customers × annual value per customer. Break each side down with a Fermi decomposition and write it as an EQUATION, each variable on its own line, each tagged and reasoned. Shape: people in geography = [n] [KNOWN/ASSUMED/GUESSED] × share who have the problem = [%] [...] × share I can actually reach (language, channel) = [%] [...] × share who would pay = [%] [...] = reachable customers × price × times per year = annual value per customer Then state: - TAM — everyone who could ever buy, in a perfect world. - SAM — the slice you can reach given geography, language, and channel. - SOM — realistic 3–5 year capture: a justified LOW single-digit % of SAM, with the % reasoned, not assumed generous. Then a top-down cross-check ONLY as a second opinion: an industry figure × your plausible share. If bottoms-up SAM comes out BIGGER than top-down TAM, STOP and flag it loudly — one of them is wrong; say which you trust less and why. Give me low / base / high scenarios by moving the 2–3 assumptions that swing the answer most. Then a 'WHAT WOULD MAKE THIS WRONG' block: the 2–3 assumptions this whole number hinges on, and a cheap, this-week way to verify each in the real world. Local reality: work in ₹ / R$ / $ to match my geography. If most of my market is India, assume severe price sensitivity — do not import Western spend levels. Never invent a TAM, and never cite a research-firm '$X billion market' figure — you have no source and you would be guessing; say so instead. Reject the 'it is a $10B market, we only need 1%' framing outright — 1% is a wish, not a plan. End with: VERDICT (real market / too small to matter / cannot tell without data) + CONFIDENCE (low / medium / high) + a 3-line 'GO VERIFY THIS' list. Reply in the language I wrote in.
