Glossary
Behavioral Metadata
Data about how work happens, not what the work is. Metadata is who communicates with whom, how fast, across which channels, at what cadence. From the connected systems Zoe reads patterns, not paragraphs: connections are read-only and scoped, what is extracted is discarded by default, and when the case is sealed the workspace is destroyed at 30 days with a certificate. (The deck, data-room files and the Founder Call transcript are stored for the case; the connected systems are not.) The patterns turn out to be more predictive than the content, which is why the method survives the privacy conversation instead of losing it.
Why it matters
Behavioral metadata predicts outcomes three to six months before financial signals surface. Communication graph collapse precedes customer churn; decision velocity slowdown precedes missed forecasts; meeting load spikes precede founder burnout and key-person departure. Metadata is the early warning system. It's also privacy-preserving and objectively measurable, unlike cultural surveys or interview callbacks. For PE investors running 40-plus portfolio companies, behavioral metadata is the only scalable way to detect deterioration before it compounds into value destruction.
How Zoe reads it
Behavioral metadata is the fuel of all nine Zoe dimensions. It includes Slack activity graphs and response latency, GitHub commit and PR patterns, HubSpot and Salesforce pipeline movement, QuickBooks activity, and cross-team communication density. The analysis is on cadence, latency, and network topology: patterns, not paragraphs. Message bodies are never stored, what is extracted is discarded by default, and personal identifiers are stripped at ingestion. Every data point is metadata: who, when, with whom, how often.