Absence-based analytics require closed source coverage

Analytics that infer an absence must prove contiguous source coverage before they classify the absence or advance a completeness watermark. A configured timeout says when an observed case would become overdue; it does not prove that no message, completion, or progress event occurred. Until both the raw-source interval and the projection over that interval are verified, dependent rates, medians, breach queues, abandonment, and funnels should be marked unavailable rather than rendered as zero or incomplete.

Applicability: event-projected dashboards, SLA monitoring, webhook-derived communications, completion funnels, and any system where missing events can be confused with negative outcomes. A safe pattern is to persist source and projection coverage separately, gate every absence-based metric on both, retain unknown/unattributed states, and allow later immutable evidence to repair the episode.

Limits: this does not make partial data useless. Presence-based counts can remain available when their evidence is bounded and labeled. It also does not replace source-specific deduplication, identity mapping, or reconciliation.