Design a Prompt Set Around Customer Decisions
Cover category discovery, solution comparison, and brand verification without relying only on prompts that already name your brand.
On This Page
Your prompt set determines which part of brand visibility you observe. Start with customer decisions before designing groups and metrics.
Collect Real Questions
Review sales conversations, support records, on-site searches, and product evaluations. Remove personal information while keeping the need, constraints, and decision context.
Instead of asking only whether a brand is good, preserve the conditions customers actually care about: budget, region, team size, data boundaries, and the task they need to complete.
Separate Three Intent Groups
| Intent | Question pattern | Observation |
|---|---|---|
| Category discovery | What approaches address this need? | Whether a brand enters the candidate set naturally |
| Solution comparison | How should these approaches be compared? | Recommendation conditions and comparison dimensions |
| Brand verification | What does this product support? | Factual accuracy and source material |
A mention has a different meaning when a prompt already names the brand. Avoid interpreting every intent through the same rate.
Control Variables Without Suggesting an Answer
Do not insert the selling points you want a model to repeat. Fix language and market, use a few reasonable variations for an intent, and record the wording.
Replacing the entire prompt set changes the observation scope. Keep a version and a reason for each change instead of treating old and new samples as one continuous trend.
Maintain the Set
Merge duplicates, remove irrelevant questions, and add emerging customer needs. The goal is useful decision coverage and interpretable results, rather than the largest possible prompt count.
Next, establish a baseline and examine mentions, recommendations, and citations separately.
