Validate a business idea in four steps, cheapest first

Validation is a cost ladder, and almost everyone climbs it backwards. The correct order is search volume, then forum reading, then five conversations, then a fake door. Total cost: roughly twelve working hours and $200. The order most people actually use is idea, build for three months, then go looking for customers. That version costs about fifty times more and answers the same questions later, when the answers are expensive to act on.

The ladder, in order

Each rung answers one question, and each one is more expensive than the last:

1. Search volume. 20 minutes, $0. Does anyone go looking for this?
2. Forum reading. 2 hours, $0. What are they doing about it today?
3. Five conversations. 4 hours of talking, 10 days of elapsed time. Will they say a number out loud?
4. Fake door. One weekend and $200. Will a stranger who owes you nothing take an action?

The order is not arbitrary and it is not just about cost. Each step produces the input for the next. Search gives you the vocabulary. Forums give you the workaround people already pay for. Conversations give you a price anchor and the three objections you did not anticipate. The fake door tests whether the pitch you assembled from all three survives contact with someone who has never heard of you.

One rule holds the whole thing together: no step can confirm an idea, only kill it. Passing all four means you have earned the right to build. It does not mean you will succeed.

Step 1: search volume, 20 minutes

Open Google Keyword Planner, which is free once you create an Ads account, or the free Ahrefs keyword generator. Check three families of terms: the product category, the problem stated in plain language, and the names of the incumbents.

Then do the arithmetic instead of squinting at the number. Suppose you want to build invoicing software for freelance videographers and the term "invoice software for videographers" comes back at 250 searches a month. Assume you rank third and take 10% of clicks: 25 visits. Your landing page converts at 3% to trial: 0.75 trials. Trial converts at 25%: roughly 0.19 customers a month. At $20 a month you are adding $3.75 in MRR per month. To reach $5,000 MRR you need 250 customers, which at that rate takes longer than your life.

That arithmetic does not say the idea is bad. It says search is not your acquisition channel, and you now have to name the channel that is: which subreddit, which newsletter, which conference floor, which agency owner with a list. If you cannot name one before you write code, stop. Nobody has ever back-filled a channel after launch by trying harder.

The inverse failure is just as common. "Invoicing software" at 40,000 searches a month is not an opportunity, it is Intuit and FreshBooks bidding $12 a click for the next four years. High volume with entrenched incumbents is a channel you cannot afford, which is the same finding as no volume at all.

Step 2: forum reading, two hours

Run site:reddit.com searches on the problem phrasing. Use the Hacker News Algolia search for the same terms. Pull up the incumbent on G2 or Capterra and filter to two and three star reviews, which is where people describe what the product fails to do rather than venting. Search Facebook groups and any public Discord for the trade.

Read thirty threads and tally what you find. You are looking for exactly one thing: a workaround with a cost attached. A spreadsheet with fourteen tabs. A Zapier chain that breaks every month. A VA at $8 an hour doing manual reconciliation. A Notion checklist someone rebuilt three times. A workaround is proof that a person already pays in time or money. A wish is proof of nothing. "I wish there was an app that did this" is the cheapest sentence on the internet.

Set the bar before you start: five posts, from five different people, in the last eighteen months, describing the same workaround in language you did not put in their mouth. Fewer than that and you have a hunch, not a pattern.

Copy their exact phrasing into a document as you go. Those sentences become your ad copy and your headline in step 4, and they will outperform anything you write yourself.

This is the slowest step to do by hand, which is why it is the one people fake. It is also the reason we built Unthinkable: it gathers demand signals from across the web and groups the recurring ones into business ideas, so your two hours go on judging patterns rather than running searches.

Step 3: five conversations, ten days

Five, not fifty. You are not sampling a population, you are disqualifying an idea. Five is enough because if the first five people describe five different problems, you do not have a market, you have five anecdotes.

Find them in the threads from step 2. Message the posters directly, referencing the specific thing they wrote. Reply rates on short, specific, non-salesy messages tend to run somewhere around a quarter, so send twenty to get five calls. Track it, because your real reply rate is itself a signal about how reachable this audience is.

Use Rob Fitzpatrick's rule from The Mom Test: ask about the past, never about the future. "Would you pay for this?" is worthless because people are polite. Ask instead:

• Walk me through the last time you did this. What day was it?
• What did you use?
• What did it cost you, in hours or dollars?
• What did you try before that, and why did you stop?
• Who else touches this?

Signals that count: they share their screen and open the spreadsheet unprompted, they say a dollar figure, they introduce you to someone else, they follow up with you a week later without being chased. Signals that do not count: "great idea", "I would definitely use that", and any enthusiasm not attached to something they already did.

And if you cannot get five strangers on a call in ten days, that is the result. It means the audience is unreachable, and unreachable audiences produce customer acquisition costs that kill companies quietly over eighteen months.

Step 4: fake door, one weekend and $200

Build one page in Carrd or Framer. Headline in their words from step 2. Three bullets naming the workaround it replaces. A price. A button.

Put the price on the page. A page without a price measures curiosity, and curiosity is free. The harsher version, which I would run if the idea is B2B: the button goes to a Stripe payment link for a $29 refundable founding deposit, and the confirmation page says when it ships.

Spend $200. Split it between Reddit ads targeted at the exact subreddits from step 2 and Google search ads on the exact terms from step 1. At a $2 cost per click that is 100 clicks.

Be honest about what 100 clicks can resolve. It cleanly distinguishes a 3% signup rate from a 20% one. It cannot distinguish 15% from 20%, so do not pretend it can. Set a coarse threshold before you spend a cent: twelve or more emails from 100 clicks and I build, fewer than five and I stop, five to eleven and I spend another $200 testing a different headline against the same audience.

Write that threshold down with a date on it. After the money is spent you will find a reason that four emails were actually encouraging, because you will want to.

One condition on all of this: everyone who leaves an email gets a real reply within 48 hours, every deposit is refunded instantly on request, and you never take money for something you have decided not to build. A fake door is a test of a pitch, not a licence to lie.

What people actually skip, and the bill for it

Steps 2 and 3. Almost universally. Step 1 gets done because it is fast and feels rigorous, and then people jump straight to building, because building is the part that is fun.

Skipping step 2 costs you the real competitor. You benchmark against the $49 a month SaaS and discover at launch that you are actually competing with a free Google Sheets template that someone in the trade published in 2019 and four thousand people use. That is a pricing problem you cannot fix with features.

Skipping step 3 costs you an order of magnitude on price and a quarter on scope. Nothing tells you which part of the workflow is load bearing except hearing a person describe their Tuesday. Without it you build the three features that were obvious and miss the one they would have paid for.

Skipping step 4 costs you the conversion rate. You still learn it. You just learn it after 180 hours of work, when the fix means rewriting the product rather than rewriting a headline.

Run the numbers on the trade. An MVP over three months of evenings and weekends, at fifteen hours a week, is about 180 hours. At a modest $75 an hour contract rate that is $13,500 of your time. The full ladder is roughly twelve working hours plus $200, call it $1,100 at the same rate. Founders routinely risk the $13,500 to avoid spending the $1,100, because the $1,100 feels like procrastination and the $13,500 feels like progress.

There is a worse variant still: doing step 3 with friends, colleagues and your co-founder's network. They will be encouraging. You will come away with false confidence, which is strictly worse than no data, because it survives contact with the first bad month.

The one case where the order changes

Genuinely new categories have no search volume. Nobody searched for "Figma" before Figma existed. But people did search for "share Sketch file with client" and "Sketch file version control", which is the workaround's vocabulary rather than the product's.

So invert step 1: search the terms people use for the thing they do instead. If the workaround has volume, you have your channel and your ad copy. If neither the product nor the workaround shows any volume, and forums are silent too, then say it plainly rather than dressing it up as being early: there is no demand signal, and you are betting on manufacturing one. That is a legitimate bet. It is also a much more expensive game with a much lower hit rate, and you should know which game you are playing before you start.

Write the kill criteria first

Four thresholds, one per step, written down and dated before you begin: the minimum combined search volume, the five-posts-five-people bar, the number of calls that must produce a dollar figure, the signup count from 100 clicks. Ideas rarely die of bad data. They die because the founder renegotiates the threshold ten minutes after seeing the data, and then spends three months building something the numbers already refused.