The comparison people actually want
The standard advice is: talk to twenty customers first, then build a landing page. It is good advice given badly, because it hides the reason. Interviews come first not because they are more rigorous — they are less rigorous — but because a landing page written before you can say the problem in the buyer's own words will get the wrong traffic and produce a confident, useless number.
The two methods are not competing instruments measuring the same quantity. An interview tells you what the problem is called, how it is currently survived, and what the person believes it costs them. A landing page tells you how many strangers will act on a claim about that problem when a price is attached. Neither answers the other's question, and running only one is how founders end up with a validated idea nobody buys.
What follows is the honest comparison, including the part where interviews mislead you and the part where landing pages do.
Side by side, on the things that decide it
Hour estimates assume one founder, no research budget, and no existing audience.
| Dimension | Customer interviews | Priced landing page |
|---|---|---|
| What it can prove | That the problem exists, and what it is called | That a stranger will act on it at a price |
| What it cannot prove | That anyone will pay. People are generous with hypothetical money. | Why they acted, or what they thought they were buying |
| Hours to a first answer | 15–25, mostly spent on scheduling | 6–10 to build, then weeks of waiting |
| Main failure mode | Leading questions and polite answers | Wrong traffic, confidently counted |
| Sample it produces | Five to twenty people you found | Whoever a search engine sends |
| Cost of being wrong | A week, and a false sense of certainty | A week, and a number you may over-trust |
Hour figures are our own estimates from running both on the sixteen ideas listed on this domain, not published benchmarks.
Why interviews mislead
The classic failure is well documented and still nearly universal: you ask whether someone would use a thing, they say yes because saying no to an enthusiastic person is unpleasant, and you write down a data point that means nothing. Rob Fitzpatrick's book on this is short and correct, and the fix — ask about the past, never about the future — is easy to state and hard to hold to after the fourth call.
The subtler failure is selection. You interview people you can reach. People you can reach are, by construction, people in your network, which for a founder means people who are unusually tolerant of new tools and unusually willing to be helpful. That population is not the market. It is the friendliest possible sample of it, and its enthusiasm is close to worthless as a predictor.
There is also a ceiling problem. Twenty interviews is a lot of work and still a very small n. If the true rate at which people care is one in twenty, you will very likely see zero and conclude the idea is dead, or see one and conclude it is alive. Both conclusions are noise.
Why landing pages mislead
A landing page's number is only as good as the traffic behind it, and traffic is the part founders control worst. Post the page in a community full of other founders and you will measure how much other founders like the idea of your idea. That number can be high and mean nothing, because those people have the same problem you do — a taste for new tools — rather than the problem you built for.
The second failure is the copy. A page is a claim, and a well-written claim about a mediocre product outperforms a badly written claim about a good one. If you are testing two ideas and one page is better written, you have measured your writing.
And there is a floor below which the method simply cannot see. With forty visitors, the difference between one signup and three is not a difference. A landing page needs enough traffic to distinguish a rate from a rounding error, and most pre-launch pages never get it, which is why the interview stage is not optional.
The order that works, and why
Three stages. The middle one is the one people skip, and skipping it is what makes the third stage unreadable.
- 1
Five to eight interviews, about the past only
Not "would you use this". Ask when they last had the problem, what they did, how long it took, and what it cost them. Stop when the fourth person in a row uses the same phrase to describe it.
That phrase is the deliverable. It becomes the headline, the target keyword, and the first line of the page. Interviews that produce no repeated phrase have told you the problem is not sharp enough to write about yet.
- 2
Find where those people already look
A search term, a subreddit, a trade forum, a Slack group. If you cannot name the place, the landing page has nowhere to get traffic from and the third stage will produce a number built on friends.
This is the stage that is genuinely skipped most often, and its absence is invisible until the page has been live for a month with nine visitors.
- 3
Publish the page, with the price on it, and wait
Written in the words from stage one, aimed at the place from stage two. Then leave it alone for sixty days, because checking it daily converts a slow experiment into a source of anxiety and changes nothing.
Decide the pass mark before you publish. Ours is fifty signups with fifteen on a committed tier, or five urgent-tier signups in any fortnight.
Roughly forty hours across three weeks, most of it waiting. The alternative is a year of building and the same information at the end.