Business Clusters
Ideas grouped by Business Genome similarity (k-means over the 34-metric profile). Clusters without a name yet are still browsable.
No clusters yet.
Frequently asked questions
How is a cluster different from a category?
A category is a label a human assigned to an idea. A cluster is computed purely from the numbers - k-means grouping over the full 34-metric profile - so two ideas can land in the same cluster despite having completely different category labels, simply because their operational, market, and risk profiles happen to be similar.
Why does a cluster description sound like a template?
Clusters get a human-authored name and description once someone reviews the real output and writes one - until then, a cluster shows an automatically generated description built from its own real strongest and weakest group averages, so it's never blank, just not bespoke prose yet.
Why does a cluster sometimes have only one or two ideas in it?
K-means always assigns every idea to some cluster, even an unusual one with no close neighbors yet. A cluster this small is flagged as an outlier rather than hidden - it's more likely an artifact of the current clustering run than a meaningful group, and may merge into a larger cluster next time clustering re-runs.
Do clusters change over time?
Yes - clustering re-runs periodically as more ideas get scored, and k-means results can shift between runs (this implementation uses 5 restarts per run to reduce, not eliminate, that variability). A cluster's membership and centroid today aren't guaranteed to be identical after the next run.