Four letterpress type blocks standing in a loose circle, each carrying a mirrored capital letter

In 2006, mymuesli was the name we almost didn't pick. We had no clue it would later become a category-defining, loved brand.

What held us back at first was that it simply felt too banal. For weeks, it sat on a longer list, next to names that sounded cleverer – like ‘Pureganics’. It grew only slowly on us, as it captured the core idea behind what our startup was about perfectly: mixing together your favourite muesli/granola. The whole story and promise in a few characters.

Since then, I've co-founded more companies, named a few of them, and helped other founders name theirs. Admittedly, I've thought about brand names a little longer than what’s healthy ;)

Most naming advice I come across still assumes the world of the 2010s. Secure the .com, register the trademark everywhere, then launch. Two things have changed since: 1) almost nobody types a domain anymore, and 2) most of the good words are taken!

So, here is how I'd go about naming a company today: the process, the rules and how I'd bend them how.

A name has a job to fulfill, replacing it is costly

In my last article, I described a brand as a stored association between a cue and an experience. The name is a first cue – something everything else gets attached to.

Startups live on fast iteration, digital or physical. A name is one of the few things you can’t iterate on cheaply. So, you want to get it right from the start.

Its job is not to explain your company. Its job is to be easy to remember, to pass on and to retrieve: by the person who heard it once over dinner, and by machines when somebody asks for a recommendation. Additionally, a good name is one that people can identify with – a simple test is asking if you would feel good wearing this on your chest for the next decade.

I sort names into three categories. They come with different costs attached.

Descriptive names – mymuesli, Salesforce, Booking.com. Such names tell the story and give away the promise. You need almost no money to educate people on what it means. The downside is that descriptive names limit you to the initial category. The narrower you are the stronger your brand might be, but only for a particular solution. Expanding wider or pivoting is often impossible.

Abstract names – Zalando, Kodak. A word that means nothing initially. A white canvas. Easy to protect, and expensive to fill with meaning. Zalando could afford years of TV campaigns. Most seed-stage companies can't.

Hybrid names – familiar sounds forming a new word, sometimes with associated meaning borrowed from linguistic relatives. Lexus for instance sounds like a mixture of ‘luxury’ – and ‘sex’. Cheaper to fill with meaning than a fully abstract word. Also the hardest to find and get right.

Explore all three, but choose on purpose. The category is really a marketing-budget decision.

To name your baby, go wi(l)de first

Before you even start, write the brief. What is the core of your business? What changes for the customer when your solution does its magic? What should people feel when they hear the name? What is the change of emotions (before vs after) that people buy from you? Which misunderstandings must be prevented? One line each.

Then make two lists: the territory you want to be associated with, and the territory you don't. The second list turned out to be very useful in my experience. It kills names before anyone falls in love with them.

Then, generate names. A lot of them. Two hundred candidates is a normal number, not an excessive one – for Project Eaden, we had more than 200 before we filtered anything. Mind maps, word association, roots from other languages, forced combinations of two words that have nothing to do with each other. Quantity first. Judgement comes later.

LLMs are pretty useful at this stage. They give you two hundred names in an afternoon and can walk you into territories you would have never entered alone. With search tools, they can quickly spot what's already taken – only a first sweep, though. In my experience, they'll often invent an etymology with complete confidence. Use them as the generator and the first filter, not as the final judge.

Next, filter rigorously

Filter from 200 down to about 10-15. I filter on story/meaning (does it carry something true?), distinctiveness (does it stand out, or sound like the five companies next to you?), potential collisions and sound. The last one gets skipped too often I think.

Sound. Names carry feeling before meaning. In one well-known study, people judged an ice cream called Frosh creamier and richer than the same ice cream called Frish.1 One vowel. Back vowels like oo and oh tend to feel big and substantial, front vowels like ee and i small, quick and light. Replications are mixed, so treat it as a tiebreaker rather than a strategy. Still worth considering.

Sounds can also make things feel small/light or big/serious, e.g. Cutipie vs Argon.

With sound comes the mouth, or rather how it moves when we say it. Saying mymuesli moves your jaw as if you chew, swallow and finish on a smile – what I called oro-kinetic symbolism in my last piece. Research points the same way: words whose consonants travel from the front of the mouth to the back, like swallowing, tend to be liked better than words travelling outward.2 Why, exactly, is still being argued about. So say every shortlisted name out loud ten times, and notice which ones you enjoy saying. (Starbucks nearly became Pequod, after the ship in Moby-Dick, until somebody asked who'd want a cup of Pee-kwod.)

On familiar vs new. Pick a name too familiar and you don’t stand out. Pick a name too strange and you're unapproachable. Names that work often pair one element people know with one they don't. Project Eaden did exactly that: a word everyone knows, next to a new one – eat and Eden in one. With Eaden carrying the vision for food to contribute to sustainable abundance.

Collisions. Do a rough check early, not at the end. Let an LLM with search do the first sweep, then Google each name together with and without your category. Search TMview and the WIPO Global Brand Database, both of which are free. Look for companies in neighbouring spaces with something that looks, sounds or means the same or similar. Having to replace a name weeks after the whole team fell for it is a major downer.

Probe it with the real world

Shortlists look different once they leave the room they were made and argued over.

The phone test. Say the name to someone on a call. Can they spell it back without help?

The native-speaker test. Ask native speakers of every language you'll sell in within three years to read each name cold and tell you what it reminds them of. Dictionaries don't catch slang, but people do. It's why the Mitsubishi Pajero is sold as the Montero in most Spanish-speaking markets.

The pitch test. Put the name into your one-line pitch. Does it feel natural or irritating?

The blind ranking. Every founder ranks their top five alone, before anyone discusses anything. The overlap is a great signal. Discussion before might contaminate it. It is easy to kill names with subjective associations and images projected into each other’s minds, too often the fastest and loudest voice wins.

The week test. Live with your top two or three names for seven days. Which one do you still like on day seven?

Then, own the decision. You must live with it. You'll say this name thousands of times before anyone else cares about it. Choosing it also says if co-founders actually see the same company.

Nobody types domains anymore

In 2015, Paul Graham wrote that if you have a US startup called X and don't own x.com, you ‘should probably change your name’. A marginal domain, he argued, signals a marginal company.3

Graham has had a lot of good insights for founders. On this one, the internet moved on without him. Yes, the dream is still a great name, the .com and a trademark in every market – if you can get all three, go for it. But that gets harder to find and more expensive to buy every year. And: people stopped typing domains.

So today I'd say: Don’t kill a great name over domain availability.

People search, tap, scan and ask AI. Some of the most searched terms on Google are youtube, amazon – and chatgpt.4 People google the name of the thing they want to open. And in a 2026 survey, three in ten shoppers had used ChatGPT to research their latest purchase, up from 2% two years earlier.5 Nobody in that journey types your URL.

Adding a few characters to an otherwise clean name for a domain has become acceptable. Facebook ran on thefacebook.com until it paid $200,000 for facebook.com in 2005. Tesla was teslamotors.com for more than a decade before it got tesla.com in 2016.6 They bought the .com once they could afford it. That's a perfectly fine order for today’s founders, too.
For Daily, a consumer brand offering ‘hydration with benefits’ I am a business angel at, we went for drinkdaily.com, same for eatfreda.com (Freda = fantastic frozen food). Owning the domain with just their brand name wouldn’t make much of a difference at all.

Where a .com domain still matters more is email. Email is the one place people still type an address, and there .com is muscle memory. Anything else is unusual enough to get mistyped. So if a .com is available in some form – even with a get-, try- or -hq added version – that's still the one I'd take, consistently for the website and emails. What isn't fine is a split identity: the product on one domain, the emails from another. Pick one string, route everything through it, and grab the same handle on the platforms you'll use.

The good names are taken

Two law professors at NYU went through 6.7 million US trademark applications. By 2016, more than 81% of the 1,000 most common English words were already registered as trademarks.7 Newcomers, they concluded, get pushed toward second-best names. Ten years and one AI boom later, I doubt it got any easier.

So when you ask ‘can we have this name?’, it is actually three different questions:

Question one: can we use it without getting sued? That's freedom to operate (FTO). Does anyone with older rights – in your markets, in your category or a neighbouring one – have something that looks, sounds or means the same or is quite close? This one is never optional as the risk of a nasty surprise is too high. A professional search costs much less than a registration and takes days to weeks, not months.

Question two: can we stop others from using it? That's registration. The official fees are the cheap part – an EU trademark starts at €850 for one class.8 The bill comes from searches, lawyers, extra classes, extra territories and the occasional opposition. In my experience, securing a name properly across the jurisdictions a European startup typically needs lands somewhere between €5k and €20k. It takes about nine months (the US being the bottleneck) before you can be sure it goes through.

After we registered mymuesli in Germany and the EU, someone else registered it in Switzerland and some Asian countries – neither is covered by an EU mark. Nor is the UK. But you don’t necessarily need to file all your core markets at once. The Paris Convention gives you a safety net: anything you file in other member countries within six months of your first filing keeps that first date.9 Had we filed in Switzerland within six months of our German filing. You should consult an IP lawyer instead of making a mistake here saving a couple of hundred Euros.

And keep in mind what registration doesn't do. It doesn't protect you from someone with older rights. That's question one again. On top of that, even when you have the right to stop others from using your name, this is quite an effort to enforce – especially for social media handles and foreign jurisdictions.

Question three: will people – and machines – think of you when it matters? No office grants this one. It's also the only one of the three that builds brand. More on this in my previous article.

Using a name you cannot own?

Most advice treats question two above as mandatory. I saw it that way too, until a couple of years ago. It’s rarely a great name you can protect versus a great one you can’t. More often it’s a great name you can’t protect versus a mediocre one you can. If you cannot register the name, and nobody else owns it, skipping registration is a legitimate option.

Supreme is a good example. The EU trademark office refused it as descriptive – and why the brand spent years fighting ‘legal fakes’ that had registered the name in countries it hadn’t. Yet its place in the heads of skaters and fashionistas was never in question. The brand lived there long before it was safe on paper.

If you can, own it. A name that's free in your classes, usually an abstract or hybrid one. This is the textbook route and the strongest legal position you can have, if you can get it. It just gets harder every year, for the reasons outlined above.

If you can’t own it, you might still use it. A name nobody can register for your category, including you, is not automatically a no-go. Save yourself the registration lottery and spend the money on being remembered instead. Booking.com is the famous case for this. The US trademark office refused it as generic, the company turned it into a household name anyway, and in this case the Supreme Court allowed the registration in 2020 because people had come to see it as a brand. Even then, the Court made clear the protection is narrow – nobody gets a monopoly on the word booking.10

Such a naming strategy works best when you are first or loudest in your category, when your real moat sits elsewhere, and when you can live with others using similar words right next to you. It gets riskier in crowded categories, where the words that match every search query also match every competitor. And it never skips question one: a name you can't register can still collide with somebody else's older mark. So check thoroughly, especially as older registrations sometimes cover names an office would refuse today.

Which route machines prefer, I don't know yet. My working assumption, not a finding: descriptive names match the prompt better, distinctive names are harder to confuse with anyone else. Similarity versus interference, in the terms of my last article. An agent recommends what it can resolve to one company. So whichever route you take, make yourself easy to resolve: the same name string everywhere, supported by the plain descriptor next to it, and structured data on your site that says which entity you are and where else you live. Google's Organization markup with sameAs links is a decent start.11

Stuck between two name options?

That usually means you have two good names, not none. Flip a coin.

Then, go back to building. The cleverer names on our 2006 list never left it. The banal one is still on the shelf.

The checklist for naming your startup

Brief

  • Write one line each:
    • the truest core and thing about the business
    • the feeling you want to convey – or the change in emotions your solution promises
    • the misunderstanding to prevent
  • A yes-territory and a no-territory
  • A budget that might be influence the category you can afford: descriptive, abstract or hybrid

Generate

  • 200+ candidates. Go broad and wild. Leverage LLMs.

Filter to 10-15

  • Story, distinctiveness, sound – say each one out loud ten times
  • Rough collision check: an LLM with search, Google plus your category, TMview, WIPO Global Brand Database

Test

  • Phone test
  • Native-speaker read-aloud for every market in the next three years
  • Pitch test: sounds natural if talked about?
  • Blind top-five ranking by every founder

Secure (or not)

  • Professional freedom-to-operate search on the final two or three
  • Own it if you can or skip registration if no other option is better
  • If you register:
    • All core markets outside the Paris Convention at once. For markets included in the Paris Convention you have a six months priority window.
    • EU marks don't cover Switzerland or the UK.
  • One domain string – a .com in some form if you can, for email's sake – one handle, everything routed through it
  • Same name everywhere and descriptior next to it where possible, plus Organization markup on your site

Two things worth reading next: John Colapinto's Famous Names in The New Yorker, on how Lexicon named BlackBerry and Pentium. And Alexandra Watkins' free SMILE & SCRATCH test – a practitioner's 12-point gut check with strong opinions.

  1. Yorkston & Menon (2004), A Sound Idea: Phonetic Effects of Brand Names on Consumer Judgments, Journal of Consumer Research. Replication results are mixed. ↩︎

  2. Topolinski, Zürn & Schneider (2015), What's in and what's out in branding? A novel articulation effect for brand names, Frontiers in Psychology. Builds on Topolinski et al. (2014) in JPSP; later work disputes the swallowing explanation, not the preference. ↩︎

  3. Paul Graham (2015), Change Your Name. ↩︎

  4. Ahrefs, Top Google Searches, US data, September 2026. ↩︎

  5. PYMNTS Intelligence & Visa Acceptance Solutions, 2026 Global Digital Shopping Index. Survey of 5,841 consumers in the US, Brazil and the UAE, March 2026. ↩︎

  6. InformationWeek on Facebook; Bloomberg on Tesla. ↩︎

  7. Beebe & Fromer (2018), Are We Running Out of Trademarks? An Empirical Study of Trademark Depletion and Congestion, 131 Harvard Law Review 945. Data: all USPTO applications 1985–2016. ↩︎

  8. EUIPO fees: €850 online for one class, €50 for the second, €150 for each class after that. ↩︎

  9. Paris Convention, Article 4. Six months is the treaty's priority period for trademarks (patents get twelve). Later filings can only cover what the first one did, so choose your first classes with that in mind. ↩︎

  10. USPTO v. Booking.com B.V. (2020). ↩︎

  11. Google Search Central, Organization structured data. ↩︎