Naming a company: our process from blank page to trademark

In early 2024, three founders walked into our Surry Hills studio with a product that modelled emissions for local councils, a funding round that needed a brand, and a working name they all quietly hated. The working name — the kind of compound-verb startup noise that ships on pitch decks and dies there — had been chosen in a hurry fourteen months earlier. Now it was on contracts, in the codebase, and in the mouths of forty customers. Renaming it was going to hurt. They wanted to know exactly how, and how much, and whether the pain was survivable.

This is that story, start to finish: four months from blank page to the name they signed. The mechanics of our naming practice — brief axes, generation volumes, screening funnels — are documented in our naming field guide. What a field guide can't show is the texture of one real project: where it snagged, what the founders fought about, and why the winning name was on the kill list twice before it won.

Weeks 1–2: the brief is a treaty

The client — who would eventually become Meridian Climate — had a classic founding-team problem: three founders, three private naming philosophies, zero shared criteria. One wanted gravitas ("we're selling infrastructure to governments"). One wanted warmth ("council officers are people, and they're exhausted"). One, honest soul, wanted whatever would get the .com.

So the brief phase wasn't a questionnaire; it was a negotiation. We ran a half-day session built around a single exercise: the conference test. Imagine your company on the lanyard at the three events that matter — a local government summit, a climate-tech demo day, an engineering meetup. Read the name aloud in each room. What must it say? What must it not say? This surfaces positioning as theatre instead of adjectives, and it works because founders will argue about a lanyard in a way they won't argue about a brand pyramid.

Out of that session came the axes, agreed and signed:

• Institutional but not bureaucratic — credible to a council CFO, not boring to a 28-year-old analyst.
• Geographic but not parochial — the data is local by nature; the ambition isn't.
• Real-word or nearword — no coined syllable salad. This audience is paid to be sceptical of marketing.
• Speakable by a phone operator — the name would be read aloud in council call centres, on radio, at podiums. If it needed spelling, it failed.

And the no-list, which mattered just as much: no "green", no "eco", no "sustaina-" anything (the category is lousy with them), no Greek gods, no founder surnames, nothing that a bored procurement officer could make a joke about.

The brief ended with one sentence — the name's job: "Make a sceptical council officer believe, before the demo starts, that these people take data as seriously as she does." That sentence decided everything that followed.

Weeks 3–5: volume, then silence

Generation ran for three weeks across three unrelated metaphor domains we chose from the job sentence: navigation (the work is about orientation), textiles (weaving datasets together — the founders kept saying "the fabric of the city"), and instruments (measurement as an act of care).

The first hundred names were everyone's first hundred. Compass-this. Atlas-that. Loom, Warp, Weft. Lovely words, all spoken for, all carrying the specific staleness of a domain the whole category already mines. The back third of the list is where we found the interesting territory: obsolete cartography terms, surveyor's words, the names of things in an instrument maker's catalogue. Three hundred and forty names, all told.

Then the most important week of the project: nothing happened to the list for five days. Generation and evaluation must be separated by time, or the team falls in love with its own cleverness in real time. When we reconvened, we scored blind — numbered list, no discussion, each founder plus both our strategists marking every name against the four axes. Blind scoring is unglamorous and essential. In-room discussion is where the most assertive founder's taste eats the list.

The blind cut took 340 to 41.

Weeks 6–9: the gauntlet

Screening is where naming earns its fee, and where amateurs lose their companies. Each survivor went through four filters, in order of cheapness:

Linguistic screen (two days). Our Singapore and London teams read every name aloud, then checked it against the obvious traps: meaning and sound in Mandarin, Cantonese, Malay and the major European languages; unfortunate homophones; how it abbreviates; what it sounds like said fast on a bad phone line. This killed nine names, including one beautiful surveyor's term that abbreviates to something no council would ever say into a microphone.

Domain and handle reality (two days). Not "is the exact .com free" — that test kills every good name on earth — but "can we build a defensible address" : the .com held by a domain parker at a sane price, an available .com.au, a clean .earth, consistent social handles. We refused to let perfect-domain thinking run the project; a great name at get- or -hq or .io beats a mediocre name at a perfect dot-com, and this audience knows it.

Trademark screen (a week, with IP counsel). This is the filter nobody should run alone. Counsel graded every name not binary-safe-or-unsafe but on risk — clear, arguable, or asking for a dispute — across the relevant classes. Twelve survivors went in; six came out. Two deaths genuinely stung. One of them was Meridian's first-moment rival, a name from the textile domain we all loved, tangled with an existing analytics firm in a near class. Kill it. A name you have to defend is a tax you pay forever.

The phrase test (continuous). Every survivor had to survive being spoken in sentences: "Meridian flagged a 14 percent variance in the fleet data." "Have you moved to Meridian yet?" "Meridian's support team called back in an hour." Names are verbs-in-waiting and sentence-sitters. Several pretty names proved unusable furniture — nothing wrong with them at a podium, everything wrong with them in a Tuesday email.

Week 10: the test we almost didn't run

Here's the honest part of the story. With six finalists, we proposed what we always propose: testing with eight to ten people drawn from the actual audience — council sustainability officers and climate-data analysts — structured as recall and fit interviews, not preference votes. The founders baulked. It felt slow, and worse, it felt like handing the decision to strangers.

We ran it anyway, with an explicit frame: you are not asking them to choose your name. You are asking whether your choice communicates. That's the line that saves naming research from death-by-focus-group. Preference tests produce beige; comprehension tests produce confidence.

The interviews killed the front-runner — a single striking word that eight out of nine participants misheard as a different, far more common word on the phone simulation. Mishearing is unfixable. And they rehabilitated "Meridian", which we'd nearly cut for being "a bit hotel chain". Participants heard it as serious, geographic, precise — exactly the axes — and, the detail that settled it, used the word "trustworthy" unprompted twice.

Meridian won not by being loved but by being correctly understood. For a company selling rigour to sceptics, that's the entire game.

Weeks 11–16: committing properly

Signing the name is not the end; it's the midpoint. The back half of a naming project:

• Trademark filing in the right classes, immediately, before the name appears anywhere public. Announcements wait on filings, always.
• The verbal system around the name. The descriptor line ("climate intelligence for local government"), the pronunciation cue (everyone should say it the same way by week one), the voice rules that make Meridian sound like Meridian in a support email — this is where a name becomes a brand voice instead of a word.
• Migration, the unglamorous spreadsheet: 240 rows covering every surface where the old name lived, sequenced so customers heard it from the founders first, support macros second, invoices last. Code names take months to fully rot out; plan for it, and let the codebase finish what the brand started.

Total elapsed time: four months. Cost of doing it again in two years because the first name was rushed: everything, plus the confusion.

What we'd tell ourselves in week one

Renaming is a positioning project wearing a vocabulary costume. The real work of the Meridian engagement was three founders agreeing what they were selling and to whom. The name is just where that agreement became irreversible.

Fall in love with the criteria, never the candidates. Every painful moment in this project — the textile name we lost to trademark, the front-runner that failed the phone test — was painless in hindsight because the axes did the deciding. Teams that skip the brief end up in month-long name-attachment grief. Teams that sign the brief get to be surprised by their own good judgement.

If you're staring down a blank page, our brand practice runs this process in eight to sixteen weeks depending on how many countries your name must survive. Or start the conversation before the funding round forces it.

Key takeaways

• Treat the naming brief as a treaty: signed axes, a written no-list, and one sentence stating the name's job. No signature, no generation.
• Generate at volume across unrelated metaphor domains; the good territory starts after everyone's obvious hundred.
• Separate generation from evaluation by a week, and score blind — the in-room discussion is where assertiveness eats judgement.
• Screen in order of cheapness: linguistic, domain reality, trademark with counsel (graded by risk, not vibes), then the phrase test.
• Test for comprehension, not preference, with people from the real audience. Preference testing makes beige; mishearing is unfixable.
• The project is half done when the name wins: filings first, then the verbal system, then a row-by-row migration plan.

FAQ

How long does naming actually take? Eight weeks is our floor for a single-market brand; Meridian took sixteen because the trademark landscape was genuinely crowded and the migration was consequential. Anyone promising a name in a fortnight is selling you a brainstorm with an invoice attached.

Should we just use the founder's name? Sometimes — when the person is the product and always will be. The failure mode is a company that outlives its founder carrying a name that no longer fits the org chart. Put founder-naming through the same axes as everything else; it occasionally wins honestly.

Can we crowdsource it? Do a team vote? Crowdsourcing is a preference test with worse data. A team-wide vote optimises for offend-nobody, which is how airports get named "Connect". Solicit the team's objections — they're marvellous at finding linguistic traps — but keep the decision with the people who signed the brief.

Do we really need the .com? No, and the belief that you do has killed more good names than trademark law. You need a defensible, findable address: exact-match .com at a sane aftermarket price, or a country/TLD combination your audience finds natural. What you mustn't do is pick a name where someone else owns the obvious address and gets your type-in traffic forever.

What does it cost when it goes wrong? The visible costs are reprinting and re-filing. The real cost is the eighteen months of explaining, the SEO authority you abandon, and the quiet doubt a rename plants in customers who were just getting comfortable. Rename once, or never.