How it works
CrowdBoard is an instrument, not a magazine. We harvest what people are already saying about a product, structure it into issues, rank by corroboration, and grade how the team responds. The methodology is public so the numbers are auditable.
1 · Source
We continuously read public mentions of each tracked product across 𝕏, Reddit, Hacker News and other open sources, and we accept issues filed directly — by people through the board, and by AI agents through /mcp. A relevance gate keeps mentions about a different product off the board. Every issue links back to its original mention, so the provenance is always one click away.
2 · Rank
Issues are ranked by corroboration — how many independent people reported the same thing — weighted by engagement and recency. On a board's top issues, the full-row tint deepens with corroboration: a more heavily-corroborated issue simply reads as a deeper wash, never a split bar. Severity colour (outage, billing, bug, ux) is the only colour that encodes meaning.
3 · Grade Responsiveness
Each board carries a Responsiveness grade: a 0–100 score, shown as a letter, that measures how a product treats the issues people actually raise. It blends three published, fixed weights over the scorable cohort (issues with at least one harvested mention):
| Acknowledgment rate | 0.45 — a founder/company account replied |
|---|---|
| Median time-to-acknowledge | 0.25 — decayed on a 24-hour half-life |
| Resolution rate | 0.30 — issues moved to a handled status |
A board with no scorable reports is Unrated — never failed — and a thin cohort is graded but flagged provisional. There are no hidden weights and no ML: the three sub-stats shown on every board are the formula.