Why Most Names Fail Before They Even Get Tested
I spent three years running naming projects for B2B SaaS companies, and I can tell you that about 70% of the shortlists die in the first round of domain checks. Not because the name is bad. Because nobody bothered verifying availability until the final stage. That will waste you weeks. The actual process is simpler than people make it sound. You pick a semantic field, generate candidate names inside it, filter by pronouncability and trademark conflicts, then lock down what's available. Most teams skip the semantic field step and jump straight into brainstorming random words. That approach produces garbage. It produces things like CloudSyncPro or DataFlowHub, which sound like they were generated by feeding a thesaurus into a blender.
The Art Of Choosing A Memorable Name Why Your Title Speaks Softly Into The Mind
When you choose a name, you are choosing what people say when they need you. That is the entire constraint. Everything else is secondary. If a founder has to spell it out over the phone, you have already lost. If someone hears it in a noisy restaurant and cannot recall it the next morning, you have lost again. The name does not need to be clever. It needs to survive casual conversation. Phonological simplicity is what matters most, and most founders completely ignore this in favor of aesthetic preference. Two syllables, one stress pattern, common grapheme-to-phoneme mappings. Words that follow English pronunciation rules without exceptions. "Kodak" works because it is impossible to mispronounce. "Xero" works because the X reads naturally as Z. "Stripe" works because it is a real word that maps directly to a visual concept. Those are not accidents. Here is a specific problem I ran into that most people do not expect. I was working on a naming project for a logistics platform and we had narrowed it down to four finalists. One of them was Courierly. It passed every test. Domain was available. Trademark search came back clean. Pronunciation was airtight. Then our client's CEO said it out loud in a meeting and immediately realized it sounded exactly like "currier" — someone who carries packages — which was the very company they were competing against. The name was technically fine but pragmatically toxic in their specific market. We pivoted and landed on Relayrd instead, which was available across .com, .io, and .co. Took three additional days and cost about $400 in domain transfers. Still cheaper than a rebrand after launch.
The Working Method
Start with a constraint, not a word list. Define what the name cannot be before you decide what it should be. Here is my checklist from actual projects: - Must be pronounceable on first hearing by someone who speaks English as a second language
- Must return zero active trademarks in Class 9, 35, 41, and 42 (the standard software/business classes)
- Must have an available .com or the founder accepts a significant brand tax for using .io or .co
- Must not share phonetic similarity with any direct competitor in the same category
- Must survive the radio test: if spoken once on a podcast, can the listener find it without spelling Generate 50 to 100 candidates. Filter them down to 10 using the checklist. Run those 10 through a proper trademark database search — USPTO TESS, EUIPO, and WIPO Global Brand Database at minimum. That is non-negotiable. I have seen teams skip the international search and spend six months building a brand only to receive a cease-and-desist from a company in Germany that registered the same name two years earlier.
Get the Full Details
After the trademark clearance, run the remaining names through a domain availability check across at least five extensions. Price out the .com. If the .com costs more than $5,000, factor that into your decision. Many founders treat a premium domain as a sunk cost problem and move on, which is reasonable. But you should know the number before you fall in love with a name.
Counter-Intuitive Things That Actually Matter
Length is not as important as people think. Three syllables is fine if the stress pattern is clear. "Pinterest" has four syllables and nobody trips over it. What actually hurts memorability is inconsistent letter-to-sound mapping. When a name contains a silent letter, an unusual diacritic, or a deliberate misspelling that breaks a common pattern, you are forcing every person who hears it to relearn English phonics in their head. That cognitive load kills recall. Another thing beginners consistently get wrong is assuming that invented words are inherently better. They are not. Invented words like "Zillow" or "Spotify" are expensive to build because they carry zero semantic baggage. You have to teach the market what they mean from scratch. Real words that map to a category — "Stripe" for payments, "Notion" for knowledge management — start with built-in meaning. The tradeoff is that real words are harder to trademark and nearly impossible to find as .com domains. You have to accept that. I once worked with a founder who insisted on a made-up name because he believed it would be easier to trademark. He was wrong. The USPTO rejects invented names frequently when they are too abstract or when they phonetically resemble existing marks. He ended up with a name that was less protectable than his second choice, which was a modified real word. He spent $3,000 in legal fees and four months of prosecution to get a mark that was still vulnerable to a opposition filing. We should have just picked the modified real word on day one.
When This Process Completely Fails
This approach does not work when your market is highly saturated in a single category. If you are opening a coffee shop in Seattle or a fintech app in the US, your semantic field is so crowded that any decent name in that space is either taken or priced out. In those cases, you either go geographically obscure — a neighborhood reference that only locals understand — or you abandon the category entirely and pick a name that lives in a different semantic universe. Both strategies carry risk. The geographic approach limits expansion. The unrelated approach requires significantly more marketing budget to create association. Another scenario where this breaks down is when the founder has strong emotional attachment to a specific word or personal meaning. I have watched three projects die because the founder refused to move past a name that meant something to their childhood but made zero sense to anyone else. No amount of testing changes that. In those situations, the practical move is to pick the best available name and archive the personal one as a codename for internal use only. It happens all the time. If you want a more systematic approach to generating candidates, there are open-source scripts on GitHub that pull from Scrabble dictionaries and apply phonetic filtering algorithms like Metaphone and Double Metaphone. They are rough around the edges but save you about two hours of manual generation. I typically run a custom Python script that cross-references a word list against domain availability APIs and outputs a ranked table. The output is not final candidates — it is a starting pool. You still do the human work of listening to each name out loud and checking it against competitors.
The name is the first signal. It will either open the door or close it before anyone reads your product description. Treat it like the structural decision it is, not a creative exercise.
