How PatternLab works —
and what it does not.
This page documents PatternLab's methodology, scope, limitations, and appropriate use. It is maintained publicly and updated when our approach changes.
A pre-research intelligence tool
PatternLab is a tool for surfacing probabilistic experience patterns before human research begins. It synthesizes aggregated public user experience signals into structured archetypes, hypotheses, and research questions — so teams can enter interviews, surveys, and usability studies with sharper frames rather than empty assumptions.
It is designed to support the thinking that happens before research, not to replace the research itself. Its outputs are starting points — not findings.
Experience pattern synthesis
When a user submits a concept, PatternLab sends the concept name, description, target audience, and known assumptions to Claude — Anthropic's large language model. Claude then reasons across publicly known user experience patterns relevant to the concept domain and returns a structured brief.
All synthesis is performed at the time of submission. PatternLab does not store, train on, or learn from user-submitted concepts.
How to read a PatternLab brief
Every PatternLab output is explicitly probabilistic. The following explains what each element means and how it should be used.
Archetypes represent clusters of experience tendencies — not demographics, not segments, not personas. They describe how a type of experience tends to unfold, not who has it. An archetype named "The Workflow-Frustrated" describes a pattern of behaviour, not a category of person.
High, Medium, and Low confidence reflect the relative strength of supporting patterns. High confidence means the pattern is well-represented in public experience data. Low confidence means the pattern is speculative and requires validation before acting on it. Confidence is not a measure of accuracy — it is a measure of evidence strength.
Hypotheses are testable propositions derived from archetypes. They are structured to be falsifiable — meaning human research should either confirm or disprove them. A hypothesis is not a finding. It is a question with a provisional answer.
Structured questions are organised by type: objections, confusion, risk, adoption, and comparison. They are designed to be taken directly into user interviews or surveys. They represent the gaps in current knowledge surfaced by the archetypes.
Recommended research methods are context-specific suggestions based on the archetypes generated. They are not prescriptive. Research teams should apply their own professional judgement about which methods are appropriate for their specific constraints.
Explicit scope boundaries
These boundaries are not disclaimers. They are design decisions. PatternLab is built to support better research, not to replace it.
When to use PatternLab
How we maintain this document
This methodology document is versioned and reviewed quarterly. When our approach changes — whether in synthesis method, output structure, or scope — this page is updated and the version number incremented. Previous versions are retained internally.
Get in touch
If you have questions about how PatternLab works, concerns about a specific output, or feedback on this methodology, we want to hear from you. Responsible use matters to us — and that includes hearing about cases where the tool falls short.