A model can generate hundreds of product names, but volume is not the finish line. The useful outcome is a small shortlist you can explain, test, and investigate. This naming sprint takes a fictional customer feedback tool from a rough brief to three candidates without letting fluent suggestions masquerade as market research.
1. Write the naming brief before requesting names
Start with the product, audience, and promise. Add constraints that would otherwise surface only after several disappointing rounds. For our example, the product collects customer feedback and turns repeated requests into a prioritized product backlog.
Prompt: Act as a product naming partner. Our audience is small software teams. Our promise is turning scattered customer feedback into clear priorities. The name should feel practical, approachable, and credible. Prefer one or two English words that are easy to pronounce. Avoid invented spellings, grand claims, and names that suggest a survey platform. Restate the brief and identify missing information before suggesting names.
Answer the model's questions. In particular, clarify whether you are naming a company, a standalone product, or a feature inside an existing brand.
2. Define naming territories before generating candidates
Ask for conceptual directions rather than an undifferentiated list. This prevents the session from becoming twenty variations on the same obvious metaphor.
Prompt: Propose five distinct naming territories for this brief. For each, explain the central idea, the intended emotional signal, and one risk. Do not generate names yet.
Possible territories include listening, navigation, sorting, shared decisions, and useful signals. Choose three that fit your positioning. Reject territories that imply capabilities the product does not have.
3. Generate a bounded batch through SwiftRelay
Use a fixed batch size so reviewing candidates remains manageable. SwiftRelay is a practical option for running this workflow through an OpenAI-compatible SDK, with per-token billing. Sign up for SwiftRelay to try the exercise with the small gift credit available to new accounts.
For an OpenAI-compatible SDK, set the base URL to https://api.fastapirelay.com/v1 and choose a model such as Claude Sonnet 5.5. Keep the brief and selected territories together in the conversation.
Prompt: Generate eight names for each selected territory. For every candidate, give its territory and a one-sentence rationale. Avoid duplicates and near-duplicates. Treat every name as an unverified creative suggestion; do not claim trademark or domain availability.
4. Score names against explicit criteria
Define the scorecard before asking the model to rank its own work. Otherwise, a persuasive rationale can make a weak name look stronger than it is.
- Clarity: Does the name suggest the right kind of value?
- Recall: Could someone remember it after hearing it once?
- Pronunciation: Can a reader say it without coaching?
- Distinctiveness: Does it avoid generic category language?
- Room to grow: Will it still fit if the product expands?
Prompt: Score each candidate from one to five on these criteria. Explain the weakest score and flag any obvious ambiguity. Return the six strongest candidates, but do not treat these scores as evidence of customer preference.
5. Challenge the shortlist instead of polishing it
Ask for objections before requesting more variations. A hypothetical name such as Signal Basket might sound approachable but also suggest data collection rather than prioritization.
Prompt: Critique each shortlisted name as a skeptical buyer. Identify likely misreadings, awkward spoken phrases, and misleading product expectations. Then propose one revision per candidate that addresses a specific weakness.
Keep the original alongside each revision. A smoother-sounding alternative is not automatically a better strategic fit.
6. Run human checks outside the model
Search for existing products, companies, relevant trademarks, domains, and social handles yourself. A model's lack of recognition is not clearance. Consult a qualified trademark professional when appropriate before committing.
Show three finalists to people resembling your intended buyers. Ask what each name suggests, how they would pronounce it, and which they remember later. Avoid explaining your preferred answer first.
7. Record the naming decision and its tradeoffs
Finish with a short decision note: the chosen candidate, its positioning rationale, unresolved checks, and the strongest rejected alternative. Save the brief and prompts too. Your next naming sprint should begin with what you learned, not another request for fifty clever words.