An average freight transit time describes a set of recorded movements; it is not a delivery promise for the next shipment. Mean, median and percentile measures answer different questions about that set. Use the same collection and delivery events, time units and reporting population before comparing the numbers.
The following ten durations are fictional teaching data, in hours: 20, 21, 21, 22, 22, 23, 24, 25, 28 and 60. They are not WTM service measurements or an industry benchmark.
| Measure | Calculation | Result for this example |
|---|---|---|
| Mean | Add all ten durations and divide by ten | 266 ÷ 10 = 26.6 hours |
| Median | Average the fifth and sixth sorted values | (22 + 23) ÷ 2 = 22.5 hours |
| Range | Show the smallest and largest observations | 20–60 hours |
| Interpolated 90th percentile, R7 method | Position 1 + 0.9 × (10 − 1) = 9.1; interpolate between positions nine and ten | 28 + 0.1 × (60 − 28) = 31.2 hours |
The 60-hour observation pulls the mean above the median. Neither number is wrong. One summarises the arithmetic total; the other locates the middle of the ordered observations. Showing both makes the distribution easier to understand than reporting only the more attractive number.
Percentiles can differ between calculation methods. NIST’s percentile reference explains alternative interpolation approaches, including the R7 method used above.
For comparison, a nearest-rank method selects position ceiling(0.9 × 10), which is the ninth observation: 28 hours. That differs from the interpolated result because the method differs—not because the shipment records changed.
Record the method with the result. A percentile from ten observations is also a description of a small sample, not evidence of what every future load will achieve.
Group equivalent lanes and services before drawing conclusions. Mixing urgent dedicated freight with scheduled shared freight can obscure the question you are trying to answer. State the reporting dates, shipment count, timezones and the start and finish events.
A collection-to-arrival duration is different from order-to-completed-handoff lead time. Use the freight clocks reference to define those boundaries.
Keep unfinished or unresolved shipments visible. A report containing only completed deliveries can leave its slowest outstanding movements out of the average. Show how many records are unresolved and whether the summary describes completed shipments or another defined population.
Do not remove a real long-duration observation merely to improve the average. Investigate its recorded circumstances, then explain any exclusion rule consistently. Pair the distribution with a separately defined on-time-in-full measure when assessing accepted delivery commitments.
For your next load, send WTM the actual ready time and required receiving deadline. Book against the service you need through time-sensitive deliveries, rather than treating a historical average as a guarantee.