Measuring QBR Metric Lineage

Purpose and scope

A method for tracing every QBR metric to source, period, definition, and owner.

What the protocol tests

Lineage coverage is only interpretable with the denominator and exclusions declared.

The record can describe the sampled process, not the whole customer base.

A clean status is not proof that the customer outcome was correct.

Limits and reporting discipline

Methodology

Measuring QBR Metric Lineage asks whether a QBR metric can be reconstructed from the retained source. The unit is one customer success record or review event, not an individual performance rating. The method fixes the observation window, source, exclusions, and decision owner.

The output is a reproducible evidence register, not a customer portfolio benchmark. It does not establish causation or a universal health score.

Collect source, period, definition, and owner for each sampled record. Include routine, escalated, corrected, and unresolved cases. Preserve event time, load time, and source version so a later correction does not erase what was knowable.

Mark missing or conflicting evidence explicitly. Do not merge records or backfill a field from a later state merely to make the sample cleaner.

Two reviewers classify a frozen sample independently and record reasons. Measure lineage coverage and agreement before drawing a conclusion. Resolve disagreements against the written rule and revise ambiguous definitions before expanding.

Separate preparation from judgment. A reviewer can classify and route a question; the named owner decides the exception or changes the control.

Report counts, denominators, missing-data rates, and limitations. A bounded sample cannot prove causation, and a clean record does not prove the customer outcome.

Give the owner a short result, the affected population, one proposed control change, and a retest date. Change one control at a time.