OTIF, Fill Rate and Perfect Orders: Reconcile the Same Twenty Orders
Maintained by IronTandem. Methods, checks and corrections.
Use one inspectable order cohort to derive three service measures, expose overlapping failures, and avoid an invented intersection.
Define the observation before the percentage
A service percentage needs an assessed order cohort, an observation point and a rule for every pass or fail. The downloadable example contains twenty fictional orders and four binary fields: on time, in full, damage-free and correct documents. A value of one means the check passed. Orders 01 through 14 pass all four. Order 15 has a documentation failure, 16 a damage failure, 17 and 18 are incomplete, 19 is late, and 20 fails all four. These flags are illustrative, not a suggested tolerance policy or an industry benchmark.
Worked example: count the OTIF intersection
Eighteen orders are on time: only 19 and 20 fail that check. Seventeen are in full: 17, 18 and 20 fail it. Sixteen orders, 01 through 16, pass both. Observed OTIF is therefore 16 / 20 = 80%. The on-time percentage of 90% multiplied by the in-full percentage of 85% gives 76.5%, which is not the observed OTIF. Multiplying the marginal rates assumes a relationship between the checks that this cohort does not have. Retain the order-level flags so the intersection can be counted rather than inferred.
Reconcile overlapping failures
Four orders miss at least one OTIF condition: 17, 18, 19 and 20. Only order 20 misses both. Inclusion-exclusion reconciles the count: 18 on time + 17 in full - 16 both = 19 passing at least one, so one of twenty passes neither. Adding two late orders to three incomplete orders would count order 20 twice and incorrectly report five affected orders. The same problem arises in operational action lists: fixing a late date does not automatically resolve an incomplete delivery on the same order.
Keep complete-order fill separate from unit fill
If the in-full flag is defined at the same fulfillment point, complete-order fill is 17 / 20 = 85%. The file contains no requested or supplied unit quantities, so it cannot support a unit fill percentage. A one-unit shortage on a large order and a completely missing small order both fail the binary in-full test but have different unit effects. Do not invent a unit rate by substituting order counts. When using the fill-rate calculator, provide real unit totals as well as the complete-order totals, with matching scope and timing.
Add quality checks for perfect-order service
Eighteen orders are damage-free and eighteen have correct documents, but only fourteen pass every check. Perfect-order rate is 14 / 20 = 70%, not a product of four component percentages. Orders 15 and 16 are OTIF successes but not perfect orders. A late complete order, order 19, is also not perfect. This separation helps explain why improving paperwork or damage can improve perfect-order performance without changing OTIF. Use an observed all-checks count when available; separate component totals alone do not identify exactly which orders overlap.
Make period comparisons reproducible
Choose whether on time refers to the customer-requested date, a confirmed date or another documented commitment. State whether in full is assessed at first shipment, final delivery or a specified deadline. Keep excluded cancellations, partial deliveries and revised promises visible. Changing those definitions between months can improve the percentage without improving service. Compare the underlying failed-order list, reconcile it to the reported cohort, and record definition changes alongside the trend. A dashboard number becomes useful when someone can reproduce it from the same records and explain which operational failures remain.
A compact view of the twenty-order cohort
| Orders | Count | Failed checks |
|---|---|---|
| 01-14 | 14 | None |
| 15 | 1 | Documents |
| 16 | 1 | Damage-free |
| 17-18 | 2 | In full |
| 19 | 1 | On time |
| 20 | 1 | All four |
Reproduce the example
The sample file contains fictional figures only. Keep the dates and assumptions in the walkthrough when comparing its result.