Semantic Integrity
Units, Currency, and Time: The Hidden Semantics That Cause Bad Answers
Units, currency, and time basis are often implicit, but AI needs them explicit.
TL;DR
- • Hidden units create wrong comparisons.
- • Make units and time basis explicit in names and metadata.
The problem (layman)
- • Measures can represent dollars, percentages, or counts without stating which.
- • Currency conversions or time windows are embedded in logic.
Why it matters
- • AI may compare incompatible values or summarize incorrectly.
- • Business decisions rely on correct unit semantics.
Symptoms
- • Percentages treated as counts in explanations.
- • Global revenue compared to local currency costs.
Root causes
- • Unit details captured in documentation, not in the model.
- • Currency normalization done inconsistently.
What good looks like
- • Units and currencies encoded in measure names.
- • Time windows explicitly documented and standardized.
How to fix (steps)
- • Append unit and currency suffixes (e.g., Revenue_USD).
- • Standardize time windows in metric definitions.
- • Add metadata notes about conversion rules.
Pitfalls
- • Assuming the visualization implies the unit.
- • Mixing multiple time windows in one calculation.
Checklist
- • Every KPI includes unit and currency in metadata.
- • Time windows are explicit and documented.
- • Conversions are centralized.