Founders typically approach naming from the wrong direction. They pick a name they like, fall in love with it, and then discover six weeks later that the .com is taken, a competitor has a similar name, or that the name is impossible to say correctly the first time you hear it.
The result is either a pivot to a second-choice name (which carries the emotional baggage of not being the first choice) or a launch with a name that's already partially compromised.
The fix is a naming process, not a naming instinct. A good process surfaces the best candidates early and filters out the fatally flawed ones before you're attached to anything. Here's how to build one.
Before generating a single name, answer these questions:
What is the primary emotion? Is your brand trying to feel trustworthy, exciting, calm, authoritative, playful, premium? Different phonetic and semantic choices serve different emotional registers.
Who is the primary audience? A B2B SaaS for enterprise security teams requires a different naming register than a consumer fitness app. Enterprise names trend toward single-word coined terms or credibility-signalling word pairs. Consumer names can carry more personality and playfulness.
What category shorthand do you want? The best names contain a signal of the category without being generic. "Airbnb" says accommodation and air travel without being "Hotel Booking Corp." "Stripe" says payment, movement, and simplicity without being "Payment Solutions Inc."
What must it not be? Are there sounds or words associated with competitors you want to avoid? Are there cultural or linguistic markets where the name will need to work?
Write these answers down before generating candidates. They will be your filter.
The biggest mistake in naming is working from too small a pool. If you have five candidates, you'll compromise. If you have fifty, you can afford to discard forty and still have a strong shortlist.
Generation techniques worth using in combination:
Compound words: Join two words that each relate to your brand character. This is the most common technique for startup names and the source of most of the great two-word brand names (Instagram, Facebook, Snapchat).
Coined words: Modify existing words phonetically — add or drop a letter, merge two words, borrow a suffix from another language. Spotify, Tumblr, Flickr.
Metaphor words: Pick a single word that captures a metaphor for what your startup does. Stripe, Slack, Notion, Linear.
Semantic field mining: List every word associated with your core concept, then every word associated with those words. Look for unexpected combinations. Many of the best names come from two or three steps removed from the obvious.
Use PhonoPair's generator to explore multi-word combinations phonetically as you go — it will surface combinations you wouldn't have thought to evaluate.
For each candidate, run it through four rapid filters before going deeper:
1. The phone test: Call someone, say the name without spelling it, and ask them to text it back to you. If they can't spell it from sound alone, the name will create friction in every word-of-mouth referral, press mention, and spoken advertisement for the life of your company.
2. The repeat test: Say the name twenty times in a row. Some names that sound fine once become grating with repetition. This will also tell you whether the name trips up your own tongue — if it does, it will trip up everyone else's.
3. The shout test: Imagine you are in a noisy bar. Someone asks what your company is called. Can you say the name once, clearly, and have a reasonable chance of being understood? Short, high-contrast names survive this. Long, consonant-cluster-heavy names don't.
4. The global scan test: Run a quick search across major languages to check for unintended meanings. This is especially important for invented words that closely resemble real words. A five-minute search can save you an expensive rebrand.
Your intuition about what sounds good is not reliable, because it's calibrated to your own accent, native language, and familiarity with the words you've been staring at for weeks.
Phonetic analysis tools like PhonoPair give you an objective measure of how well two words flow together — covering consonant transitions, vowel compatibility, stress pattern harmony, syllable ratio, and pattern effects like alliteration and rhyme.
Run your shortlist through phonetic analysis and look at three things:
Pillar 1 (Phonetic score, 0–100): Anything above 70 is strong. Anything below 50 warrants careful consideration — it doesn't mean the name is wrong, but you should understand exactly why the score is low.
Language structure adjustments: Check whether the adjective-noun ordering is natural, whether stress patterns create awkward rhythm when spoken together.
Semantic patterns: Alliteration, rhyme, and semantic complementarity are all positive signals. Phonetic clashes and awkward transitions are negative signals.
Names that score well objectively are not always the most creative — but they are more likely to be the ones that survive contact with customers.
Domain checking and trademark screening should happen together, not sequentially — they're both binary filters, and you don't want to spend time on trademark research for names whose domains are gone.
Domain priorities: The .com is still the most credible TLD for a startup, but .io, .co, and .ai have all established legitimate precedent. If your target .com is taken but the holding is a parked domain (no active site), acquisition is often possible — factor this into your timeline.
Trademark basics: A full trademark search requires a lawyer and is worth the investment before you file. As an early-stage filter, search the USPTO database (TESS) for identical and phonetically similar marks in your target classes. A name that appears clear in this search might still have conflicts — but one that shows obvious conflicts should be dropped immediately.
The combination: Ideally, your chosen name has a .com available (or acquirable at reasonable cost) and returns no obvious trademark conflicts. Names meeting both criteria are genuinely rare, which is why the generation step requires a large candidate pool.
Once you have a shortlist of 3–5 names that have passed all filters, test them with people who represent your target customer — not friends and family who will tell you what you want to hear.
The most useful tests are:
Perception mapping: For each name, ask respondents to rate it on a set of scales (professional/playful, trustworthy/risky, expensive/affordable, modern/traditional). Check whether the name scores where you intended it to.
Association test: For each name, ask what industry or product comes to mind. If a name meant to signal a fintech company consistently suggests a cleaning product or a pharmaceutical, that's information you need before launch.
Recall test: Show respondents the name once, distract them for two minutes, and ask them to write it. A name that's hard to recall after a single exposure will struggle in a world of short attention spans.
The goal isn't a name that wins a vote. It's a name that consistently lands where you intend it to.
Great startup names are not accidents. They're the output of a disciplined process that generates enough candidates to have genuine options, filters rigorously before attachment is formed, and validates with data rather than gut feel alone.
A name chosen this way will serve your company for decades. The process takes two to three weeks done properly — a negligible investment against the cost of a rebrand at Series A.
Start generating and analysing startup name candidates
Open Generator