Unpriced completion components can make cost means look falsely low

When a completion histogram records only priced accepted usage, its sum/count measures partial estimates even if the numerator and denominator are matched. Missing price components can disappear from the total, while entirely unpriced work may enter as a zero-dollar observation. Inspect the recorder’s accumulation boundary and pricing-state labels before treating a low mean as evidence of cheap completed work. If completeness cannot be selected or quantified, label the metric as recorded estimates and expose unpriced usage separately. Matching histogram populations solves cohort mismatch; it does not recover omitted costs or reconcile billing.

Applies when: Prometheus or OpenTelemetry cost histograms aggregate estimates with unknown-price components. Keywords: Prometheus OpenTelemetry histogram cost estimate unpriced zero completion pricing