Start with the job the name must do
Write one sentence that states who the app serves, what painful job it handles, and what changes after it works. Add the first acquisition channel and the likely expansion path. A name for an app discovered through spoken creator videos faces different pressure than one sold through enterprise demos.
Use this brief to judge names. Without it, short and clever options tend to win even when they say nothing useful or trap the product inside its first feature.
Do not force the whole pitch into the name
The store listing, subtitle, screenshots, and description can explain the category. The name should carry a clear idea, be easy to recall, and look distinct beside nearby apps. A generic category phrase may be clear but hard to own. An opaque coined word may be ownable but expensive to teach.
Generate both types, plus metaphorical and outcome-led directions. Compare them against the same brief instead of deciding in the abstract.
Run the spoken-name test early
Say each finalist in a normal sentence, then ask another person to type what they heard. Test dictation and automatic captions if video, podcasts, or word of mouth matter. A name that repeatedly becomes a common first name or another brand will leak demand every time someone hears it.
Also ask people to spell the name without seeing it. One small correction is manageable. A different spelling from every listener is a recurring acquisition cost.
Check the places users will search
Search the major app stores for exact and close names in the same category. Then check the domain endings you would actually launch, plus relevant package or developer namespaces such as GitHub, npm, or PyPI. Keep direct competitors and active same-category products as blockers, not minor score penalties.
- Record exact matches, close spellings, and phonetic matches separately.
- Treat a registry not-found result as time-stamped evidence, not a purchase guarantee.
- Open the finalist at a registrar before announcing the name.
- Leave unsupported social-platform checks as manual tasks.
Choose in context
Put the final two or three names into an app icon, store result, spoken recommendation, and plain-text message. Score clarity, recall, pronunciation, distinctiveness, channel fit, and expansion width. Keep domain and collision evidence beside the score rather than blending every fact into one number.
The founder still has to want to say the name for years. Use that judgment after the factual blockers are visible, not before the checks begin.