Early-Warning & Change Detection
Detecting meaningful shifts as they happen and connecting them to a response.
At a glance
- Produces
- Detections tied to agreed thresholds, and a defined route from a detection to a response.
- Depends on
- A baseline, an expected range, and a decision owner named in advance.
- Fails when
- Nobody has agreed what a detection will mean, so alerts accumulate without consequence.
Early-warning systems and change detection methods monitor streams of evidence for departures from expectation that are meaningful rather than incidental, then connect those detections to decisions about how and when to act. Detection without a response threshold is surveillance, not a warning system.
How we approach it
The hard part of early warning is not detection. It is agreeing in advance what a detection will mean.
- Indicators are designed with a baseline and an expected range, so that a departure can be recognised as a departure.
- Statistical detection is paired with interpretive review. Automated alerts are treated as questions posed to an analyst, not conclusions.
- Thresholds and triggers are agreed with the people who will respond, before the system goes live, and are documented with the reasoning behind them.
- False positives are expected and planned for. A system that never alerts is a system that will be ignored when it matters.