Most founders name their company in an afternoon. Somewhere between the idea and the first commit, a name gets picked, a domain gets bought, and the decision is filed away as branding. Something to revisit later, when there’s budget for a proper agency and a proper opinion.
That framing is the mistake. Naming isn’t the first branding decision you make. It’s the first distribution decision you make, and it’s the one you’re least likely to reverse.
Why naming has become more important, not less
Speaking on Lenny’s Podcast, a16z general partner Anish Acharya made an argument about where distribution now comes from. An entire generation of founders was trained on network effects. Come for the tool, stay for the network. The problem is that every network that already exists learned that lesson too, and each one is now built specifically to stop anyone from growing a network on top of it. That route is closed.
So the network effect has reverted to something older and messier. Grassroots word of mouth. When a product is getting mentioned organically across X, YouTube, and Instagram, that is the strongest third-party distribution available to a founder today. And unlike the mobile era, there’s no app store and no growth hacking industry to plug into. You build your own channels off the back of people talking about you.
Here’s the part worth sitting with, which follows from his argument rather than being part of it.
Word of mouth is spoken. That’s what the phrase means. It happens in a podcast, a group chat, a conversation at a conference, a reply to someone asking for recommendations. Someone hears about you once, with no spelling and no context, and then goes looking for you an hour later or a week later.
Every step of that depends on your name. If distribution now runs on word of mouth, and word of mouth runs on a name being heard once and found later, then the name is load-bearing infrastructure. Not decoration.
The three tests
A name has to survive three separate journeys to work as distribution. Most names fail at least one.
It survives being said out loud. Someone hears it in conversation and can reproduce it. This is where creative misspellings die. Drop a vowel, swap a c for a k, add a doubled letter, and you’ve built a name that works perfectly on a slide and falls apart the moment it leaves someone’s mouth. The test is simple. Say it to somebody who has never heard of you, don’t spell it, and ask them to write it down.
It can be typed from memory. Different test, and harder. Hearing a name correctly is not the same as recalling it accurately hours later. Names that are one letter away from a common word tend to get autocorrected in someone’s head. Names built from two joined words tend to get remembered in the wrong order. Names that rely on a prefix like get- or try- tend to get remembered without it.
Searching it returns you. This is the test almost nobody runs before committing, and it’s the most expensive one to fail. Take the bare word, search it, and look at what owns the results. If a category term, two other startups and a VC firm with the same name are sitting above you, you’re not going to displace them. You’ll spend years buying your way to the top of a query that should have been free.
That third test has changed shape recently. It’s no longer only about search results. When someone asks an assistant about tools in your category, the model needs a distinct string to attach facts to. A name that collides with a generic industry term gives it nothing to hold onto. The company that owns its own word gets cited. The company sharing a word with a whole category gets folded into the category.
What happens if you fail all three
Plenty of successful companies have names that fail every test above. This is worth being honest about, because the counterexamples are obvious and anyone reading this can name three.
What those companies did was become the retrieval mechanism themselves. If the name can’t be found, the founder can be. You put a face on it, build a following, post consistently for a year or two, and people search the person rather than the product. Someone hears about your app, can’t remember what it’s called, but remembers who built it and finds you that way.
This genuinely works. It’s also slow, expensive, and fragile. It requires one specific person to keep showing up, and the distribution collapses if they stop. It’s a full-time job layered on top of the job of building the thing.
Naming well is the cheap version of the same outcome. You do the work once, at the start, and then every mention anyone makes of you for the next decade routes correctly without further effort.
How to actually run this
Three practical shifts.
Treat the name, the domain, and the handle as one decision. Founders who split them end up explaining for years which one is really them, and every explanation is friction on a channel that’s supposed to be frictionless. Check all three before you commit to any of them.
Test the bare word before you fall in love with it. Search it. Look at what’s already there. Do this on day one, not after the logo is designed. The name that survives this test is worth more than the name that sounds better in your head.
Prefer a coined word you can own over a real word you’ll share. A made-up but pronounceable word has no incumbent competition in search results, no category to be confused with, and no VC firm sitting on the same query. The cost is that it means nothing on first hearing. That cost is usually smaller than it feels, because names acquire meaning from the product rather than the reverse.
Don’t let availability pick your name
The practical objection to all of this is that every name passing the three tests is already taken. That’s mostly a .com problem.
Short .com domains are largely gone, and the ones that remain are listed at aftermarket prices that make no sense for a company yet to launch. So founders compromise, and the compromise almost always lands on the name rather than the extension. They add get or try or app, drop a vowel, tack on a number. Every one of those choices fails the second test on purpose. A padded domain is a name that can’t be typed from memory, and it was chosen not because it was better but because the clean version wasn’t available.
That’s the wrong thing to give up. The name is the part doing the distribution work. If you’ve found one that says what the company is and passes all three tests, the answer is to put it somewhere it fits rather than damage it to fit somewhere.
For a tech startup that usually means looking at .io, .ai or .tech, all of which read as native to the category and none of which carry the availability problem. Which one suits depends on what the company actually is. .ai signals the model layer specifically, which is useful if that’s the whole product and limiting if it isn’t. .tech is the broader one, and it does its clearest work when the technology is the substance of the idea rather than a layer on top of something else. bizname.tech tells anyone who hears it, once, what kind of company they’re dealing with, and it leaves the name itself intact.
The thing to take away
Naming is the cheapest distribution decision you will ever make and the most expensive one to change. Changing it later means abandoning every mention, every link, every person who half-remembers you, and starting the accumulation over.
Do the work in the afternoon you were going to spend on it anyway. Just run the three tests while you’re there.
