Voice of Customer Synthesiser

Feed it a batch of raw customer text — reviews, interviews, tickets, survey responses — and get back a structured report covering JTBD, a verbatim language bank, pain points, delight drivers, gaps, and messaging implications.

What it does

Produces a structured synthesis across six dimensions:

• Jobs to Be Done — at least three primary jobs in the format When I [situation], I want to [motivation], so I can [outcome]
• Language bank — verbatim customer phrases only, preserved exactly in quotation marks across five categories: problem language, desired outcomes, frustrations, evaluation criteria, and product experience
• Pain points and frustrations — each with a frequency rating (common / occasional / isolated) and the most vivid direct quote
• Delight drivers — what customers value, what surprised them, and how often
• Gaps and unmet needs — missing expectations, desired features, and alternative comparisons
• Messaging implications — 4–6 concrete recommendations covering claims to emphasise, exact language to reuse, proof points, and objections to pre-empt

Every response opens by confirming the data type, volume, and source received. If evidence for any dimension is weak, the skill explicitly flags INSUFFICIENT SIGNAL FOR RELIABLE FINDING rather than filling gaps with inference.

When to use it

Use it when you have a batch of raw customer-generated text and need structured insight before strategy or creative work begins. Works from reviews, interview transcripts, NPS comments, support tickets, community posts, and sales call notes. Effective as a Claude.ai Project with source documents attached, or as a one-off analytical task. Use Sonnet for standard volumes; Opus for large batches or deeper analysis. The output feeds directly into persona-builder.

When not to use it

• For quantitative survey data — the skill reads qualitative text, not Likert scales or numerical exports
• Building personas — that step belongs to persona-builder, which is designed to take this output as its input
• Reviewing copy for brand voice compliance — use brand-voice-enforcer for that