Documentation — Getting Started
Demands turns public complaints into scored, buildable opportunities. This documentation covers how that works, so you can judge how much to trust any given result.
Last updated 2026-08-17
The short version
Twelve connectors pull raw text from places people describe problems. A model reads each document and extracts the specific friction. Near-identical frictions are grouped into clusters. Each cluster becomes one opportunity, scored across seven dimensions.
Nothing is invented. Every stage is traceable back to a document you can open.
The five stages
| Stage | What happens |
|---|---|
| Ingest | Connectors pull raw documents and company records |
| Extract | A model pulls specific pain points out of each document |
| Cluster | Near-identical pain points group together |
| Synthesise | Each cluster becomes one product idea |
| Score | Seven weighted dimensions produce a 0–100 score |
Where to go next
- How scoring works — the dimensions, weights and what each one means
- Data sources — the twelve connectors and what each contributes
- Reading an opportunity — how to interpret a page without over-trusting it