Three line items, only one of which gets counted
Ask a founder what building an unwanted product cost them and they will give you a number in weeks. Twelve weeks, or a summer, or most of a year. It is an honest answer to a question that has been asked wrongly, because the weeks are the cheapest item on the bill and the only one that stops accruing when you stop working.
The second item is judgement. Once you have spent twelve weeks, your reading of the evidence changes in a direction you cannot feel from the inside. Four signups becomes early days. Two users who stopped logging in becomes an onboarding problem. This is not weakness; it is the ordinary operation of sunk cost, and it operates most strongly on the people most committed to their work.
The third item is the tail. A product with four users is not finished when you stop building it. It has a domain, a dependency tree, a support inbox and four people who will reasonably expect it to keep working. That tail runs for years, costs an afternoon a month, and occupies the specific mental space that a new idea needs.
What building first actually costs you
Your own numbers. This one is uncomfortable on purpose — most people have never put the maintenance tail in the same total as the build.
Be honest. Then add a third, which is what everyone leaves out.
Evenings and weekends, for most people, not forty.
Your billable rate, or what you turn down to do this.
Domains, dependency updates, and answering four people. Longer than you think.
- The build
- $10,800
- The tail nobody counts
- $4,320
- Total, before a single user
- $15,120
At the defaults — 12 weeks at 15 hours, an hour worth $60, then 24 months of 3 hours keeping it alive — the build is $10,800, the tail is $4,320, and the total before anyone has used it is $15,120.
Assumptions behind these numbers
- Your time is priced at what you would otherwise charge for it. If you would not otherwise be paid, set the rate to what you would need to be paid to give up the evenings.
- The tail assumes you keep the product running rather than shutting it down. Shutting it down is cheaper and almost nobody does it.
- Nothing here counts the opportunity cost of the idea you did not build, which is real and impossible to price honestly.
- Nothing here counts the effect on your judgement, which is the item this page argues is largest and which has no dollar figure.
Against eight hours
Fifteen thousand dollars of your own time against roughly eight hours to publish a priced page. The comparison is so lopsided that stating it feels like a trick, and there is a catch worth naming: the page does not save you fifteen thousand dollars. It saves you fifteen thousand dollars in the cases where the answer is no, and in the cases where the answer is yes it costs you eight hours and six weeks of waiting.
So the real question is your base rate. If four in five of your ideas are ones you would abandon after launch — which is roughly the historical rate for most people who ship regularly, including me — the expected value is overwhelming. If you are the rare person whose ideas mostly work, validation is a tax on your throughput and you should ignore it.
Six weeks of waiting is a genuine cost and the one people underweight in the other direction. It is six weeks during which you are not building anything, and for a lot of people that is six weeks of losing interest in the idea entirely. That, too, is a result — but it is a result about you rather than about the market, and it is worth knowing which one you are reading.
How it usually goes when you build first
Compressed from three of my own, and recognisable enough that most people who ship will find their own version in it.
- Week 0
The idea is obviously good
It is obviously good because you have not yet said it out loud to anyone who would have to pay for it.
- Week 4
The interesting part is done
The hard technical problem is solved, which feels like most of the work and is about a fifth of it.
- Week 12
It works, and there is a landing page
The landing page is written last, in a hurry, in your vocabulary rather than the buyer's. This is the single largest reason launches get no traffic.
- Week 14
Nine signups, two of them people you know
Here the second line item starts running. Nine becomes early, and the plan becomes marketing rather than a rethink.
- Month 6
Four users, none of them paying full price
The product is fine. The problem it solves was one you had rather than one they had, which is the most common way this fails and the hardest to see from inside.
- Month 30
Still paying for the domain
And still updating dependencies, because two of the four users are real people who were nice about it.
Source: Three of my own projects between 2021 and 2025, compressed. Not a study.
The reasonable objections
- Building is how I think. I cannot evaluate an idea in the abstract.
- That is true of a lot of people and it is not a defect. The compromise that works is to build the smallest thing that produces the page's substance — the calculator, the comparison, the template — rather than the product. You get to write code, the artefact is publishable, and it takes days rather than a quarter.
- Nobody signs up for things that do not exist. I would be measuring nothing.
- Signup rates are lower, sometimes a lot lower, and that is the design rather than a flaw. What matters is the ratio between ideas, not the absolute number: six pages published in one fortnight are comparable to each other even if all six convert badly. If all six convert at zero you have learned something about your traffic, which is also worth six weeks.
- My idea needs the product to be understood.
- Sometimes genuinely true, and it is the strongest objection here. The test is whether you can describe what it replaces. If you can name the spreadsheet, the manual process, or the tool it sits next to, a page can carry it. If you cannot, you may be building something new enough that no page will help, and you should be honest that you are taking a much larger bet.
Fifteen ideas that have not been built
Each one is a bet that has not been placed yet, with the price and the failure criteria published in advance.