Confidence-convention mismatch

data_quality

A ~67%-confidence radius and a CEP50 radius differ enough that pooling them unconverted yields containment estimates wrong by a factor of two or more.

Detection

Mixed sources with no declared convention; empirical containment far from the nominal rate.

Mitigation

Require the convention as mandatory metadata
Normalize all radii to one confidence level before pooling
Prefer vendors that publish a containment calibration

Read the full article: IP Lookup To H3 →

Affected conversions

Point+radius → H3 approximate
IP lookup → H3 (probabilistic crosswalk) approximate