How to launch a startup: the distribution week nobody plans for

Most founders plan the build and improvise the launch.

The build gets a scope, a rough date, and a list of things that have to work. Distribution gets a Tuesday. You finish the product, you feel the relief of finishing, and then you sit down to tell people about it and discover that telling people is its own project with its own week attached.

I run a launch directory, which means I am at the receiving end of that week for a few hundred products every month. I see which founders planned it and which ones are improvising, and the difference is visible from the submission form.

This is what the distribution half actually contains.

What a launch week actually contains

Write the list before you start, because the failure mode is not doing too little. It is doing the first thing you thought of, at length, and never getting to the rest.

Your own channels first. Whatever list, following, or group chat you already have. This is the highest converting thing you own and it costs nothing. Every product I have watched win a week brought its first twenty interested people with it rather than expecting to find them.

Communities where the question already gets asked. Not a broadcast, an answer. Threads asking for exactly the kind of thing you built already exist and already rank. Being genuinely useful in one of them outperforms announcing in five.

Launch directories. A listing, a link, and discovery traffic from people who browse new products on purpose. Steady rather than spiky. This is the row people over-invest in emotionally and under-invest in practically, which I will come back to.

One thing with a long tail. A written post, a video, a teardown. The thing that is still bringing people in three months from now when the launch-week attention is gone. Ordered by effort against return, your own channels win and directories are cheap. The mistake is treating the list as sequential and running out of week at step two.

The part that eats the days

Here is the specific thing that turns directory submissions from a cheap row into a lost weekend.

Every directory wants the same five things. Your product name. A tagline, usually capped somewhere around sixty characters. A description. Your URL. Your logo.

You will type those five things again, and again, into forms that are laid out differently every time, with different field labels, different limits, and different rules about what counts as a category. None of it is difficult. All of it is the same. By the twelfth one you are copying from a scratch file and pasting into boxes, and by the twentieth you are making mistakes in the tagline because you stopped reading it.

There are 188 directories in the catalog I maintain. Nobody should submit to all of them, but plenty of founders reasonably want to do thirty or forty, and thirty or forty of anything typed by hand is a day you did not plan for.

The engineering problem underneath it

This is the part that interested me more than the marketing problem, because the obvious solution does not work.

The naive approach is a rule per directory: this site puts the tagline in #product-tagline, that one calls it input[name=subtitle]. It works on the day you write it and rots immediately. Every redesign breaks a rule, and a catalog of 188 sites means you are maintaining 188 brittle mappings for a payoff that expires.

The approach that survives is to treat it as a scoring problem instead. Read every visible text input on the page. For each one, collect the evidence around it: its label, its placeholder, its name and id, the text sitting next to it. Score that evidence against what each field kind usually looks like, take the densest form on the page, and assign one input per kind where the score clears a threshold.

Negative evidence turns out to matter as much as positive. "Name" is a strong signal for the product name and a terrible one if the surrounding text says "first name" or "company name". Without penalties the thing confidently fills a newsletter signup.

The useful property is that it generalises. A directory nobody wrote handling for still works, because the heuristic is reading the page rather than recalling it. Add a site to the catalog and it is covered within the hour, with no release.

Two constraints worth knowing if you build something similar. Setting element.value directly does not register with React or Vue, because the framework tracks its own state and never sees the assignment. You have to go through the native prototype setter and dispatch the event yourself. And file inputs cannot be filled programmatically at all, by design, so logos and screenshots have to be handed over some other way.

What it does not do

Autofill is best-effort field detection, so review what a form contains before you submit it.

Forms embedded from another domain, the Tally and Typeform style embeds some directories use, are outside what a content script can reach without opening a hole you should not open. Per-field copy buttons are the fallback there.

File uploads, as above, are download buttons rather than autofill.

And it fills five fields. Directories that ask for a founder story or a long pitch still want you to write one, and you should, because that is usually the field a human reviewer actually reads.

Launch day is not the deadline

The habit worth breaking is treating launch day as a verdict.

That expectation is borrowed from boards that clear overnight, where whatever you posted yesterday is gone by morning. A permanent listing behaves differently. It stays in the archive, it stays in the sitemap, and it keeps sending a trickle of people who were not looking on the day you shipped.

So the useful work is the follow-up. Point at the listing again when you ship something. Answer the question in the thread that gets re-asked every month. Founders who keep mentioning the thing after the week ends are the ones who see it build, and a quiet first day is recoverable rather than final.

Where to start

Write the list before launch week rather than during it. Do your own channels first. Pick the communities where the question is already being asked rather than the ones with the largest member count.

Then do the directories in one sitting with something else doing the typing.

The catalog is free and filterable by domain rating, traffic and price, so you can pick the thirty worth your afternoon instead of working down an alphabetical list. The NL Launch Companion is the extension described above, and it is free on the Chrome Web Store.

Nick Launches is the directory I run. Listing costs nothing, the page stays up, and I publish the week's raw numbers there every Sunday, including the ones that do not flatter me.

Comments

Loading comments…