otif-analysis

OTIF requires order-level rows with requested delivery date, actual delivery date, and completeness information. An order counts as OTIF only if it arrives on or before the requested date with all lines or full quantity delivered, judged at order level rather than averaged across lines.

High scores usually reflect measurement choices: tolerance windows, promised-date anchors that hide sales padding, cancelled orders leaving the denominator, and line-level averaging. Computing the metric ladder from tolerant to strict definitions exposes how much each choice inflates the reported KPI.

You need order_id, requested_delivery_date, promised_delivery_date, actual_delivery_date, completeness info (lines or quantities ordered vs delivered), and a status or cancelled flag. Useful segmentation cuts include carrier, region, customer, and product family.

Only the promised-date rungs of the ladder are computable, and the analysis cannot detect sales padding between requested and promised dates. The Skill flags this explicitly and recommends capturing the requested date going forward.

Quietly removing cancelled orders from the denominator flatters the metric. The analysis counts cancelled orders during validation, states how they are treated, and includes a ladder rung with cancellations kept in the denominator to show the effect.