Technical reference
Data and calculations
Not every number in Fluorgan represents the same thing. Some values come directly from reviews; others are model estimates, derived calculations, or planning projections.
Public data model
Public documentation is a versioned projection of product contracts; editorial text is not the source of the math.
Observed, estimated, derived, and projected
Observed is evidence recorded in the system; estimated comes from a model; derived calculates over existing data; projected looks at future work.
Current value and history
Current asks about state now; historical asks what was recorded or known at that moment.
Time, dates, and timezone
Timezone changes civil-day grouping when the contract uses a local calendar; it does not change the original review instant.
Windows, samples, and eligible population
Each metric declares its own window and population; there is no single Fluorgan window.
Evidence and confidence
A value should be read with evidence, coverage, and public rule; volume helps, but it is not the only part of evidence.
Gaps, null, and invalid denominator
Absence of observation is not zero; null is not coerced to 0 for UI convenience.
Versioned public formulas
Public formulas come from the PublicMetricRegistry and carry version, unit, window, population, and limits.
Forecast is not an event
Forecast projects future work, but does not create a review, future memory, or completion guarantee.
Rounding and display
Rounding is presentation; calculations use the appropriate internal value.
When do numbers change?
Confirmed reviews, time passing, availability, card versions, and new evaluations may change specific readings.
Privacy boundary
How we calculate metrics and how we process personal data are related but different questions.
Metric family reference
Retention, Readiness, Workload, Card Health, Recall Integrity, and Consolidation have their own types and limits.
Interpretation limits
Numbers do not capture all knowledge outside the system, fatigue, outside help, future performance, or exam results.
How to verify a metric
Check question, type, window, population, evidence, state, limits, and the specific reference.