User Persona Builder

---
name: User Persona Builder
description: Synthesizes interview notes, surveys, support tickets, and analytics into evidence-based user personas that drive design decisions, with every claim traced to source data. Use when someone asks "build personas from our research", "turn these interview notes into personas", "how many personas do we need", or "are our personas evidence-based". Do NOT use for sales-targeting ideal customer profiles with filterable firmographics and buying committees - use icp-persona-builder instead. Do NOT use for extracting jobs-to-be-done from research - use jtbd-extractor instead.
---

# User Persona Builder

Build personas that designers and PMs actually consult - grounded in observed behavior, not demographics invented for a slide. A persona is a research artifact, not a character. The costly failure this prevents is the decorative persona: a stock photo, a name, an age, and a made-up hobby that no design decision ever touches, which quietly teaches the team that research outputs are theater.

## Operating procedure

### Step 1: Gather inputs and enforce the evidence floor

Ask for the source material: interview notes or transcripts, survey results, support tickets, analytics, session recordings. Then apply the evidence rules:

- **Minimum for defensible personas: 8-12 interviews** across the target population; below 5, refuse to call the outputs personas - label them proto-personas (assumption maps) and recommend research first, offering to design it (pair with interview-guide-builder).
- Surveys and analytics alone cannot produce personas - they show what and how much, never why. They corroborate interview-derived clusters; they do not replace them.
- If there is no research at all, say so plainly and offer a research plan. Never fabricate a persona from team assumptions and present it as evidence.

Also collect: the decisions the personas must inform (onboarding? pricing? feature priority?), and any existing segmentation to reconcile against. Label assumed context as a guess.

### Step 2: Find patterns, not averages

Read across all data and tag recurring behaviors, goals, and pain points. Cluster users by **behavior and motivation, not by age or job title** - two 28-year-old marketers can need opposite products, and a 24-year-old and a 55-year-old with the same workflow belong in the same persona. Averaging produces a user who does not exist; clustering finds users who do.