What 'On-Time Rate' Actually Measures (And What It Doesn't)
On-time rate is the most-quoted KPI in last-mile — and the most quietly fudged. A short field guide to the definitions that matter.
We benchmarked 2,400 real delivery runs across three US metros. Here is what manual dispatch leaves on the table — and how AI clusters recover it.
Last-mile delivery is the most expensive segment of any logistics operation — typically 41% of total shipping cost. Most of that overhead is invisible: empty miles between stops, drivers idling at customer doors, dispatchers manually reshuffling routes when an order changes.
The benchmark
We ran 2,400 real delivery routes through Fleetune's AI optimizer and compared the output to dispatcher-built routes for the same stops. Three US metros, four fleet sizes, mixed pickup-and-delivery workloads.
- Average distance reduction: 27.8%
- Average completion time reduction: 31.4%
- On-time delivery rate improvement: +8 pts (91% → 99%)
- Driver overtime hours: -42%
“We stopped staffing the dispatch desk on Saturdays. The algorithm makes calls we used to argue about in a meeting.”
— Marcus Lee, Logistics Manager
Why manual routing leaks money
Dispatchers anchor on geography they recognize. They cluster stops by ZIP code, not by drive time, and they avoid splitting familiar neighborhoods across drivers. Both moves feel intuitive — and both add miles. Our solver evaluates 8,000+ candidate sequences per route in under five seconds, surfacing combinations no human would patiently consider.
The savings compound. A 28% miles reduction is also a 28% fuel reduction, a proportional emissions reduction, and a meaningful reduction in vehicle wear. For a 12-truck operation, that is roughly $96,000 per year before counting labor.