A measure is a definition first
Most logistics reports show percentages: on time, in full, damage-free. Percentages are persuasive and easy to misread. Two organizations can report "95% on time" and mean entirely different things, because one counts against the promised date and the other against the first date the customer asked for, or because one counts deliveries and the other counts order lines. A KPI that cannot be reproduced from a written definition cannot support a decision.
For each measure below, the definition names the event being counted, the timestamp used and the population in the denominator. The definitions are authored for this article as examples. They are not an industry standard.
Five measures, defined
| Measure | Numerator | Denominator | Timestamps and notes |
|---|---|---|---|
| On-time delivery | Deliveries handed over at or before the agreed delivery time | All deliveries due in the period | Handover time from the POD; agreed time from the booking. State which time zone and whether a window counts as on time at its end. |
| Order accuracy | Orders delivered with the right items and quantities | All orders delivered in the period | Count orders, not lines, or state otherwise. Exceptions are taken from the receipt record. |
| Damage incidence | Deliveries with at least one package recorded as damaged | All deliveries in the period | Damage must be noted at receipt to count; later claims are tracked separately. |
| Dwell time | Elapsed hours between arrival and departure of a unit at a facility, summed | The number of units that left in the period | State start and stop events, such as gate-in and gate-out, and exclude units still on site. |
| POD completeness | Deliveries with a POD recorded within the stated time after handover | All deliveries in the period | State the time allowed and what counts as complete: signature, quantity, condition. |
Why the denominator is the argument
If a late delivery is removed from the denominator because the customer asked for a change, the on-time rate rises without any improvement in operations. If damage is counted only when a claim is filed, the incidence falls while the damage continues. A reader seeing only the percentage cannot know. Writing the denominator next to the number turns the KPI from a claim into a calculation anyone can check.
A hypothetical scorecard
Hypothetical scorecard, invented for illustration.
In an invented month: 200 deliveries were due. 188 were handed over by the agreed time. On-time delivery is 188 ÷ 200 = 94%. Of the 200, 3 had a damaged package noted at receipt, so damage incidence is 3 ÷ 200 = 1.5%. 194 of the 200 had a POD recorded within 24 hours, so POD completeness is 97%.
The same month reported against a different definition could look different. If the 12 late deliveries included 5 where the customer changed the date, a report that removes them gives 188 ÷ 195 = 96.4%. Both numbers can be honest. They answer different questions, and the scorecard should say which one it is.
Data-quality traps
- Timestamps entered by hand, later, from memory, rather than captured at the event.
- Time zones and daylight changes mixed across systems.
- Orders split or merged between booking and delivery, so numerator and denominator count different things.
- Exceptions recorded in a free-text note that no report reads.
- A POD recorded days after the handover, which makes both completeness and timeliness uncertain.
From measure to decision
A KPI earns its place by changing what someone does. Before adopting one, write the decision it informs: which carrier to review, which site to coach, which customer instruction to clarify. If no decision follows from a measure, it is reporting for its own sake. The GS1 Traceability Standard describes recording events and the data about them, which is the raw material any of these measures depends on.
Choosing which measures to keep
Five measures are enough to start, and fewer can be better. Pick the ones that match a decision you actually face. A team that struggles with disputes at delivery gains more from damage incidence and POD completeness than from dwell time. A team that runs a warehouse will want dwell time and order accuracy first. Adding a measure has a cost, because someone has to collect the data, check it and explain it every period.
Reviewing a number in a meeting
When a figure is presented, ask four questions before discussing it. What is the denominator? Which timestamps were used? What was excluded, and why? How does this compare with the same definition last month? If nobody can answer the first three, the discussion is about a number nobody can check, and the meeting should begin by fixing the definition.
Write the definition at the foot of the report and keep it fixed for at least a quarter. A definition that changes every month produces trends that mean nothing, and the temptation to adjust it after a bad period is exactly the reason to write it down first.
Small samples
A percentage computed on a handful of deliveries swings widely. One late delivery out of ten moves on-time delivery by ten points. Show the count next to the percentage, and avoid ranking sites or carriers on a month in which one of them handled very few shipments.
Limits of this guide
The definitions are examples to be adapted to your contracts and systems. No target values are suggested because appropriate targets depend on cargo, route and customer. The numbers are invented and the scorecard describes no real operation.