Persona Builder

Builds evidence-grounded marketing personas from research inputs — with every attribute tagged as evidenced, inferred, or assumed, so you can see exactly how much weight to put on each one.

What it does

Produces a structured persona profile across ten dimensions — each attribute tagged with one of three confidence levels:

• [EVIDENCED] — directly supported by the research inputs
• [INFERRED] — evidence-based inference, not directly stated
• [ASSUMED] — archetype assumption that requires validation

The profile covers: segment name and label, situational context (role, pressures, time horizon), stated vs. underlying goals, frustrations and failed attempts, information and influence landscape, evaluation criteria, verbatim language and messaging angles, and objections with the underlying concern each one signals.

If evidence is thin, the output is labelled PROVISIONAL PERSONA and lists what additional research is needed. For multi-persona work, a comparison table highlights the differences that most affect targeting and messaging decisions.

When to use it

Use it at the strategy stage, after research has been gathered. Accepts VoC syntheses, interview notes, CRM observations, sales notes, and segment data. Designed to sit directly downstream of voice-of-customer-synthesiser — paste that output in as the primary input. Works best in a Claude.ai Project with research documents attached. Use Sonnet for a single persona; Opus for multi-persona differentiation work. The output feeds into content-brief-generator.

When not to use it

• Without any research input — the evidence/inference tagging system has nothing to work with, and invented demographics are explicitly blocked by the skill's rules
• For ad platform audience definitions — this produces narrative personas, not targeting parameters
• For synthesising raw customer feedback — run voice-of-customer-synthesiser first, then bring the output here