A pre-launch page has two jobs and they fail differently. It has to make somebody believe the problem is theirs, and then it has to make giving you an address feel small.
Most advice addresses the second — button colour, form length, above the fold. The first is where nearly all the loss is, and it is diagnosable.
Which half is broken
You can tell without analytics, which is useful because there is none on this site by design and you may not have any either.
Time on page is short and signups are near zero. They did not believe the problem was theirs. This is a copy problem, and specifically a first-sentence problem.
They read and did not sign up. They believed you and the ask felt too big. This is the form, the absence of a reason to act now, or an unanswered question about what happens next.
They signed up and did not verify. Neither of the above — see below.
The first case is the common one and it is the one people spend the least time on, because rewriting a sentence feels less like work than adding a testimonial section.
The first sentence is doing all of it
A visitor decides in about a second whether this is about them. That decision is made on one sentence, and the sentence most founders write describes their solution.
The one that works describes the reader's situation, in the words they would use. Which is why the sequence matters: you cannot write this sentence before you have talked to twelve people. The words you need are the ones you heard, not the ones you would have chosen. A page written before those conversations is a page describing an imagined person.
Concretely, the test: could somebody with the problem read your first sentence and think "that is what I was going to say"? If it reads as a product description, it is not there yet.
What actually loses people after they have read it
In rough order of size:
No sense of what happens next. "Join the waitlist" answers nothing. When? What arrives? How often? Are you going to sell my address? One sentence under the button fixes this and almost nobody writes it.
Too many fields. Every field past the address needs to justify itself. A price question is worth its cost. A "company size" dropdown you will never use is not.
Nothing that makes today different from next week. This is what referrals are actually for — a queue position is a reason to act now that costs you nothing to give, and it works precisely because it is visible and relative.
Visible signals of not-realness. Placeholder text, a broken link in the footer, an obviously unfinished section. People are checking whether you exist.
The third failure, and its number
Signing up and not verifying is a separate loss and it is invisible unless you look.
Every Osokoro plan including the free one does double opt-in, which means you get two numbers — signups and verified signups — and the ratio is one of the more informative figures available. Around 80% is unremarkable. Around 50% means something is wrong with where the traffic is coming from. Below that, you are largely counting things that are not people.
Watch it per source. A blended figure hides the one channel that is producing nothing.
The verification email itself loses real people, and the losses are mostly mechanical: send it immediately, say on the confirmation screen exactly who the email is from, make the subject line unmistakably yours rather than "Confirm your email", and let people re-request it without signing up again.
Worth knowing now rather than later: the shared sending domain works on every plan. Everything includes one custom sending domain and Studio allows several; mail stays on the shared domain until the DKIM records you publish have been verified. A from-address a reader does not recognise is one of the things that costs you verifications, so finish that verification before a large launch.
Tag your sources or none of this works
A conversion rate without a source is an average across audiences who behave nothing alike, and it will point you at the wrong fix every time.
The specific trap: one channel with volume and terrible conversion will drag the blended number down, you will conclude the page is broken, and you will rewrite a page that was working fine for the other three channels.
UTM tags on every link. Osokoro records source and shows it, and the useful view is conversion and verification rate side by side, per source. The two together tell you whether a channel is sending people who are wrong for you, or bots.
The order to fix things in
- Talk to twelve people. Write down their words.
- Rewrite the first sentence using those words. Nothing else matters until this is right.
- Add one sentence under the button saying what happens next.
- Remove every field you will not act on.
- Tag your sources and read conversion per source, not blended.
- Only then consider the referral mechanic, the design, and the rest.
Steps 1 and 2 account for most of the available improvement and are the cheapest things on the list. They are last on most people's because they do not feel like building.