How NameJury reaches its top picks.
A focus group is only as honest as the people inside it. We built a 27-cell reviewer matrix, a 4-dimension scoring rubric, and a domain check that ignores everything except whether a registrar will actually sell you the name. No vibes. No vanity scores. No tomorrow-I-promise lists.
Why 27 reviewers, not 5.
Five reviewers score the same way — that's not a focus group, that's an echo. The 27-cell matrix is a 3 × 3 × 3 cross-product of three axes. Each cell represents a specific buyer archetype with different sensitivities, references, and tolerance for risk. A name that lights up only one column is a tell — and that's information you cannot get from a single chatbot persona.
- ·First-time
- ·Repeat
- ·Serial
Risk tolerance and pattern-recognition load shift dramatically across founder lifecycle.
- ·Tech / SaaS
- ·Consumer / DTC
- ·Services / B2B
Naming conventions are not universal — a wellness-DTC name reads differently than a developer-tools name.
- ·Playful
- ·Professional
- ·Minimalist
Captures the texture/gravitas trade-off — same word, three completely different reactions.
3 × 3 × 3 = 27 persona cells. Every candidate name is scored in every cell, every time.
Scoring rubric.
Each persona scores each candidate on four orthogonal dimensions, 0–10 integer per cell. We do not blend dimensions into a single composite — premium and clarity often pull in opposite directions, and the spread is the signal.
Does the name signal “I can hand this company money”?
Does the name feel scarce, hard-won, not template?
Does the name hint at what the product does — or at least not mislead?
Would a buyer remember it tomorrow, after a single mention?
For each dimension we sum across 27 personas: 0–10 per cell × 27 cells = 0–270 per dimension. Normalize to a 0–27 reported score by dividing by 10. Net score is the unweighted average across the four dimensions — published alongside the raw spread so you can see which axis is dragging.
Porkbun-truth domain check.
Every candidate is checked against Porkbun's live registrar data. We trust the registrar because the registrar is the only party that can actually sell you the domain at a real price.
- · Live availability of .com, .ai, .co, .app
- · Registrar price — standard, premium, or aftermarket
- · Owner status — available, parked, or actively used
- · Defensive bundle recommendation, per brand
- · Squatter ask prices on parked pages (theatre, not market)
- · Near-miss spellings — if you have to spell it, it's gone
- · Domain-search SaaS that pings stale WHOIS once a day
- · Anything we can't verify by hitting the registrar API live
Trademark screen.
We screen each candidate against USPTO live filings, Google top-results, App Store, and Product Hunt for naming collisions. We return a 3-tier severity (clear / soft conflict / hard conflict) plus the receipts.
This is a screen, not a legal clearance opinion. A clear result here does not mean a name is registrable, defensible, or non-infringing. For binding clearance — especially before filing your USPTO application or making a meaningful brand investment — retain a trademark attorney. NameJury surfaces the obvious blockers so you can rule out the broken names early and brief counsel on the survivors.
Live trademark conflict screen · 8 sources1
1Surfaces filed conflicts. Not legal clearance. Consult a trademark attorney before filing.
Availability check — the defensibility matrix.
For the recommended brand, we generate an 8-row checklist of the properties a competitor would need to claim to dilute your position. Each row is a single action you can take, today, with the link to the right registrar or platform.
Each row is stamped clear · caution · blocked · not_checked · owned.
Built inside a venture studio.
NameJury is run by Felipe Baytelman. It was extracted from the naming workflow inside a venture studio that has shipped a portfolio of consumer and B2B brands — among them LodgeProof (the short-term-rental damage-proof startup) and AdmitCompass (the college-admissions guidance app). The internal version of this method picked those names. The productized version is what you're looking at now. Same rubric, same registrar truth-test, same disrespect for vibes-based shortlists.