Observation & Confidence
Absence is not zero.
Most tools quietly fill gaps with estimates. Orbit does not. This page defines the states an Orbit finding can hold and how each one should be read.
—How certain is this?
Evidence states
Four ways to read a finding.
Observed
The evidence was captured directly. The finding states what was seen and when it was seen.
Unobserved
The evidence could not be captured during this analysis — a page was unreachable, blocked, or absent.
Reported as unobserved. Never scored as zero, and never silently omitted.
Interpreted
Two or more observations are related to explain a likely consequence. The underlying observations remain attached.
Pending
The subject is configured but has not yet been observed. No claim is made until it has been.
Recency
Every finding carries its age.
- Observations are dated at capture, and findings show which capture they came from.
- Where a subject has history, movement between captures is shown instead of a single static figure.
- A stale subject is marked as stale rather than presented as current.
- Re-running an analysis produces a new dated record; it does not overwrite the past.
Why it matters
Confidence is a product decision.
- Estimated data looks confident and quietly misleads. Orbit prefers a visible gap.
- Decisions made from clearly-sourced evidence can be defended and revisited.
- When Orbit does not know something, saying so is the more useful answer.
Say what was seen. Say what was not. Never confuse the two.