
Proprietary detection
A school moves as one body. Individual movement inside it is almost impossible to isolate by watching, which is exactly the problem at enterprise record volume. Our detection layer reconciles independent systems against one another and surfaces the combinations that cannot all be true.
Candidate lifecycle
A candidate surfaces from reconciliation across independent systems.
An analyst confirms it is a real combination and not a data artefact.
Alternative explanations are tested against the records that would support them.
A second, independent source supports the same conclusion.
Exposure is calculated deterministically, with stated inputs and a duplicate-exposure check.
A named analyst traces it to source, and a reviewer agrees. Anything else is closed.

What we do not publish
We do not publish detection logic, thresholds, weighting or rule sets, and we do not expose them to third parties. What a client receives is the finding, the calculation, the exposure and the preserved original evidence behind every material statement.
Where models are allowed
Document extraction, timeline construction and narrative drafting, all logged with the model, version and task recorded against the matter.
Final calculations, substantiation decisions and signoff remain with named analysts and reviewers.