METHOD
How we know when we’re wrong.
How MOSAIC works inside is a trade secret. The standard every briefing has to meet isn’t. This page sets out what goes in, what comes out, how each call is scored, and where we’ve found the limits of our own sources.
INPUTS
What goes in and what doesn’t
In:
Publicly available information: filings, interviews, speeches, public statements, published writing, and public records (OSINT).
Material a client already holds and chooses to share (HUMINT).
Every source is logged, so each claim can be traced back to it.
Never:
Contact with the subject. The person profiled is never approached.
Faces, voices, or physiological signals. MOSAIC is not SPOT.
Protected characteristics. We don’t infer race, religion, health, sexual orientation, or political affiliation.
Hacked, leaked, or unlawfully obtained data.
WHAT COUNTS AS A CALL
Five rules before anything ships.
A Behavioral Briefing™ is built from calls: how the person is likely to decide, what moves them, and what backfires. Every call has to meet five rules.
1. Specific enough to be proven wrong. “Likes control” isn’t a call. “Won’t commit without a third decision-maker in the room” is.
2. Evidence attached. Each call carries the documented behavior it rests on.
3. Behavior and motive kept apart. What someone did is reported as observed. Why they did it is marked as inference.
4. Language matched to confidence. Every briefing carries a confidence score. Below 70%, findings are stated as tendencies, not predictions, and words like “will”, “always”, and “never” aren’t used.
5. Thin signal isn’t scored. If the record couldn’t have shown a trait, we say “not observable in this record”, never “absent”.
SAMPLE OUTPUT
The rules, on a full briefing.
This is a composite. Peter Crispin isn’t a real person, and no single record was used to build him. The format, the rules, and the limits are exactly what a client receives before MOSAIC’s output gets customized to their needs.
