Bad startup names cost money in three ways: ad CTR, support tickets ("how do you spell that?"), and rebrand legal. The good news is most of the cost is avoidable with a checklist that takes ten minutes per name.
Pronounceability. the phone test. Read the name out loud to a stranger over a noisy phone call and ask them to spell it back. If they get it wrong twice, your support team will spend the rest of the company's life correcting it on intake calls. Names with a clear vowel skeleton (Stripe, Notion, Loom) survive this test. Cute spellings (Lyft, Tumblr) work only when distribution is so strong that ad spend covers the spelling lesson.
Search collision. the SERP test. Google your shortlist in an incognito window. If page one is dominated by another company, a movie, or a generic product, plan for a paid SEM budget on day one. The phrase "Cluely" had no SERP collision in 2024 and ranked instantly. The phrase "Pulse" had hundreds. Same length, same vibe, completely different distribution cost.
Trademark. the USPTO test. Run a free search at tmsearch.uspto.gov for the exact mark and for phonetic equivalents. A B2B SaaS that uses the same mark as a small brand of dog shampoo can usually coexist; a B2C app that collides with another B2C app cannot. Coexistence is class-based, but you should still know what you are walking into.
Social handles. the platform sweep. Check X, Instagram, TikTok, LinkedIn, and YouTube. If three of five are taken by squatters, plan to pay for the handles or pivot the spelling. The cost of buying a handle from a squatter is often $50 to $500. Tolerable. The cost of making @yourbrand_app vs @yourbrand a long-running compromise is invisible but real.
The five-second test. the elevator. Tell five non-tech people the name and ask them what the company does. If three of five guess close to right, the name is doing free work for sales. If five of five guess wrong, the name is fighting the brand and you will pay for the fight in every cold email and pitch.
The .com test. your generator did this. Even in 2026 .com still wins clicks on cold traffic. .io is fine for developer tools where the audience self-selects. .ai is fine if the brand is genuinely AI-first. Generic startups should fight for .com or accept the conversion penalty.
The team test. Three people on the founding team should be able to say the name without wincing. Founders who hate their own brand for the first six months never recover the conviction.
The future-proof test. Read the name with a future product attached. "Loom for video" is fine; "Loom for audio" still works. "Notion for notes" is fine; "Notion for forms" stretches but holds. If the name boxes you into a single feature ("VideoCalc"), be sure that feature is the company.
Run all eight in fifteen minutes per name. Skip none. The compound interest of a strong name shows up in CAC three years later, by which time you cannot fix it without burning a year of brand equity.