Naming: our field guide from brainstorm to trademark check

Every naming project starts with the same scene. A founder announces the name "just came to her in the shower", the room goes quiet, and someone's trademark search finds it registered to a supplements company by Thursday. The shower name is dead by lunch. This is not the tragedy it feels like. Names that survive companies are rarely found; they're funnelled — generated at embarrassing volume, screened with increasing brutality, and sold to a committee with unusual care.

We've named coffee companies, banks and things we can't tell you about yet, and the process holds across all of them. Here it is, end to end — the version we run inside our brand practice.

Phase one: brief the axes, not the adjectives

Most naming briefs are lists of vibes: "modern but warm, premium but approachable". Every name ever made fits that. A useful brief defines axes — spectrums the name must locate itself on:

Descriptive ↔ invented. "Hearthbrew Coffee" tells you what it is; "Brassfern" makes you ask. Both are defensible; the choice depends on whether discovery or distinctiveness is the bottleneck.
Founder-led ↔ system-ready. A consultancy selling one person's brain wants the founder's name. A product that will outlive its team doesn't.
Serious ↔ playful. Not "some of both" — pick a position within one notch of centre.
Global ↔ local. A wine label that never leaves the Mornington Peninsula can be gloriously parochial. A SaaS that sells into Germany and Japan has pronunciation homework.

We also write the no-list in this phase — the sounds, words and associations the client hates, said out loud and agreed on. The no-list saves more time than any mood board, because naming arguments are usually unspoken vetoes colliding in week six.

Deliverable from this phase: one page. Axes marked, no-list agreed, one sentence stating the name's job ("help a first-time buyer remember us over the two bigger roasters"). If this page isn't signed, we don't generate.

Phase two: generate obscene volumes

Amateur naming generates twelve names and falls in love with four. Professional naming generates three hundred and kills them procedurally. The volume isn't waste; it's how you exhaust the obvious. The first fifty names in any brief are everyone's first fifty names — same roots, same suffixes, same botanical metaphors. Names fifty-one onwards are where the territory gets interesting.

Our generation techniques, per axis:

Metaphor mapping — from the name's job, branch three unrelated domains (for a ledger: navigation, weaving, geology) and mine each for roots. Unusual metaphors produce names competitors can't follow you to.
Constructive — blends, clippings, compounds. "Northwind" + "ledger" thinking. Brutal hit rate but the survivors own themselves completely.
Found language — verbs, nouns, obsolete trade words, cartography terms, the names of things in sheds. Real words carry warmth constructs can't buy.
Sound-first — start from phonestheme. Short front vowels read quick (bit, flip); long vowels read slower and warmer (slow, hearth). If the brief says "calm", we generate for sound before meaning.

Everything goes into one shared sheet, no authorship attached. Anonymity is load-bearing: people defend names they made and undersell names they didn't.

Phase three: screen like you mean it

Three rounds, each cheap at the top and expensive at the bottom:

Round 1 — the smell test (300 → ~60). Gut reads against the axes, in-house, fast. Anything that mispronounces at first glance in English dies here. Anything a stakeholder will never say out loud in a sales call dies here. We read the whole list aloud — names live in the ear, not the eye.

Round 2 — the hazard round (60 → ~12). Linguistic checks in target markets (the classic crash-test: say it in a German and a Japanese accent; check obvious slang collisions); domain reconnaissance (exact .com availability is noted but not fatal — more below); Google, app-store and social-handle squatting checks. We also do the meanness pass: what will a hostile subreddit do to this name? If the answer is immediate and funny, kill it now and save the PR team a decade.

Round 3 — legal (12 → ~6). Trademark screening with actual IP counsel, in the actual classes of goods, in the actual markets. We are designers, not lawyers, and we say so in writing. Clients occasionally push to skip counsel on a "small" brand. That conversation is short: the most expensive rebrand is the one a cease-and-desist chooses for you.

Phase four: reality tests on the shortlist

Six survivors get road-tested before anyone falls in love:

The phone test. "Thanks for calling, this is ___." If reception can't spell it back, the name carries a tax forever.
The email test. hello@___. addresses read aloud to a stranger, spelled back unprompted.
The sentence test. Drop each name into five real sentences: a referral ("You should talk to ___"), a complaint, a funding announcement, a support ticket, a domain upsell. Names are verbs-in-waiting ("just ___ it") — check they don't break grammatically or worse, work too well as one.
The cold-context test. The word alone on an invoice footer, an app icon, a conference lanyard. No logo, no colour, no story. This is where meaning-free invented names wobble and founder-beloved cleverness shows its seams.

Phase five: presenting so it survives committee

How names die isn't screening; it's presentation. Deaths we no longer cause:

Death by email. Never send a list of names by email. A list invites ranking; ranking invites picking the least-offensive name, which is how you get "Solutions" in your wordmark.
Death by nakedness. A name shown as bare 12pt type on a slide is judged as type. Show each shortlisted name in situ: on a phone screen, a packet, a door, a sponsorship board. People can't hear a name's potential without help.
Death by democracy. Do not open the shortlist to a company-wide vote. Votes optimise for the explicable; the best names take a paragraph to explain and a week to love.

Our format: present three, never more. Each with its axis position, its story in ninety seconds, its real-world mockups, and its known weaknesses stated by us before anyone else finds them. Stating the weakness inoculates: "this name will be misheard as X about once a month" said by you is diligence; discovered in the meeting it's an ambush. Then — and this took us years to learn — send the client away for a week. Names need to be lived with, and decisions made in the room are made about the presentation, not the name.

On domains: stop grieving the .com

The perfect .com for any decent two-syllable name has been parked since 2004 and is owned by someone with a P.O. box and a dream. So: decide what the domain is for. Brand trust and email deliverability need something clean and unambiguous; they do not need the bare word. hearthbrew.coffee and hearthbrew.co carried a subscription brand past seven figures. Pick the name for the name; pick the domain for clarity; never trade a great name for a mediocre exact-match.

Key takeaways

Brief axes, not adjectives — and write the no-list early.
Generate hundreds. The interesting territory starts after the obvious is exhausted.
Screen in three rounds: smell test, hazards, then real trademark counsel. Skipping counsel is how rebrands happen to you.
Road-test the shortlist: phone test, email test, sentence test, cold-context test.
Present three names in context, state weaknesses first, and let the client live with them for a week. Never email a list; never hold a vote.
Don't sacrifice the name to the domain. The domain's job is clarity, not purity.

FAQ

How long should naming take? Four to six weeks including legal screening on a shortlist of three. Faster is possible; cheaper-and-faster usually means the legal round got skipped, which isn't a saving so much as a loan.

Can we keep a name that's weak but has heritage? Often yes — inherited equity outweighs awkwardness more often than founders think. Screen it the same way (especially the hazard round) and fix the pronunciation problems with wordmarks and verbal identity rather than a rename.

Descriptive or invented — how do we actually decide? Look at the discovery channel. If strangers must find you by category (search, shelf, app store), a descriptive spine helps enormously. If you'll be referred person-to-person or out-spend competitors on brand, invented names compound better. Most of our recommendations land on a construct with a descriptive echo — unfamiliar enough to own, familiar enough to file.

Who owns the final decision? One accountable human, advised but not governed by the committee. Naming by consensus produces the name nobody hates, which is the name nobody remembers.