Promise measures and disruption risk

A useful measure converts a competitive priority into something observable in orders. If the promise is delivery complete and on time, the on-time rate alone is insufficient: an order can arrive short. A shipped-quantity rate is also insufficient when delivery is late, documentation incorrect or goods damaged. ASCM's SCOR framework treats a perfect customer order as satisfying delivery, completeness, documentation and condition requirements together against the customer promise. Count the numerator as orders satisfying every condition in the same cohort and the denominator as all included orders. Do not add the component rates or multiply them as if independent; the actual intersection requires order-level evidence. Do not silently mix order lines with orders, or a shipment-month cohort with an order-entry-month cohort, since that changes the denominator.

Separate measurement from diagnosis. A lower perfect-order rate is an outcome, not a cause. Partition nonperfect orders into actionable failure groups, preserving overlaps so one order is not counted twice. Match the response to the source: a component shortage differs from a constrained test station or wrong paperwork. Record likelihood, consequence and detection time, then choose a proportionate buffer. Safety stock for one component can soften supply interruption but uses capital and may conceal a record error; a second supplier may reduce one exposure but needs qualification, quality and real capacity; spare testing capacity does not correct export documents. A sensitivity case should show what an action changes and what it does not. Do not call an improved metric causal proof unless the comparison uses consistent periods, order mix and audit rules.

Connect the metric to strategy without making it a target detached from cost and flexibility. A policy that blocks shipments with wrong documentation may initially improve documentation while reducing on-time delivery; watch both components and their intersection. If the customer's promise uses the original requested date, do not silently redefine 'on time' by moving the commitment date after a delay just to raise the rate. Record the original and revised promise and why it changed, then use a declared metric rule. Distinguish an outcome indicator such as perfect orders from an early signal such as the share of component deliveries confirmed before the production schedule freezes. The early signal permits intervention but does not guarantee the outcome when testing or transport fails. In risk review, assign the signal an owner, intervention threshold and review date rather than a general instruction to 'increase inventory'.