Explore Moveka

Article

Delivery Speed Metrics Explained: Average, Median, Percentile and On-Time Rate

Learn how to read delivery speed metrics without being misled by late outliers, failed orders or route mix. Use average, median, P90 and a clear on-time rate to improve handovers and keep customers informed.

Published 15 Oct 2026 · By Moveka

On this page

See what customers really experience—not just a flattering average. A clear delivery scorecard helps you spot slow orders, tighten handovers and keep customers informed.

Ready to make your next delivery easier to arrange? Start on the Send page to see the options and prices that apply to your parcel.

Get my delivery price

What do delivery speed metrics mean?

  • Average (mean): Add the delivery times in your chosen group and divide by the number of deliveries. A very late order can pull the average upward.
  • Median: Put delivery times in order and take the middle value. It shows the midpoint, so an unusually late order has less influence than it does on the average.
  • Percentile: A point in the ordered distribution. The 50th percentile is the median; P90 shows the point at or below which 90% of the measured deliveries fall. It helps you inspect the slower tail.
  • On-time rate: The share of eligible orders delivered within the delivery window you defined. Its meaning depends on the orders included and the rule for counting them.

No single metric tells the whole story. Read the average, median, a slower-tail percentile such as P90, and on-time rate together—then check the order counts behind them.

How do you measure delivery performance consistently?

  1. Choose the reporting window and the unit you are measuring, such as one order from dispatch to completed delivery.
  2. Write down the on-time rule before calculating: which promised window applies, and which timestamp counts as delivery? Use the same rule each time.
  3. Define the eligible order group. Keep cancellations, failed attempts, reschedules and orders still in progress visible; state how each is treated instead of quietly removing them.
  4. Calculate the average, median, chosen percentile and on-time rate from that same group. For the rate, divide eligible orders delivered on time by all orders in the defined eligible cohort.
  5. Compare like with like. Check results within similar service, route or area, and delivery type before interpreting a network-wide change.
  6. Show the order count beside each rate, and review missing timestamps or status changes before sharing the scorecard.
Your denominator can change the story. If failed attempts or unresolved orders disappear from the eligible cohort without a stated rule, the on-time rate may look stronger than the customer experience.

Worked example: why can the average hide a late delivery?

Imagine delivery times of 1, 1, 2, 2 and 9 hours. The average is 3 hours, while the median is 2 hours. The late order pulls up the average; the median shows the middle delivery. Neither tells you everything, so add a slower-tail percentile and the on-time rate to see more of the customer experience.

Use one consistent percentile calculation method, especially when comparing small groups. Keep the measurement window and included orders consistent too.

Why do delivery speed comparisons go wrong?

  • Averages hide the slow tail: A better average may still leave some customers waiting much longer.
  • Route or service mix changes: A period with a different blend of routes, areas or delivery types is not automatically a fair comparison. Compare similar groups first, then show the overall result with its order counts.
  • Cancelled and failed orders vanish: If you exclude them without showing how they were handled, your rate can tell an incomplete story.
  • Small groups look decisive: A rate without its order count can disguise how little data sits behind it.
  • Inconsistent timestamps muddy results: Different start points, delivery events or status updates can make periods impossible to compare cleanly.

What should you record for each order?

Keep one row per order with an order ID, promised date or window, dispatch or handover time, delivery or attempt timestamps, final status, service type, route or area, and any cancellation or failure reason. A basic handover log helps you reconcile what was dispatched, accepted, attempted and completed.

Use that same routine to keep customers informed: record the latest order status and make updates part of your handover process. When orders leave with a clear record, it is easier to answer “Where is my order?” and act on delays.

What should you do next to improve delivery reliability?

Start with a simple scorecard: average, median, P90, on-time rate, eligible order count, cancellations and failed attempts. Review it on the same basis each time, investigate the slow tail, and turn what you learn into a more consistent handover routine. Moveka is Kenya’s largest courier and one of Kenya’s most deeply connected courier networks. When you’re ready to arrange a delivery, start on the Send page to see the options and prices that apply.

Get my delivery price