Most road freight scorecards measure what is easy to export rather than what changes a decision. Twenty numbers arrive monthly, three of them move, and nobody can say which system produced any of them. The fix is not a bigger dashboard. It is a short list of KPIs where each one has a formula, an owner and a named source system, which in practice means the number comes out of freight visibility software rather than out of a spreadsheet somebody rebuilds every month.
Twelve KPIs cover European road freight properly. Each one below carries its formula, the system that should supply it, and the way it is most often gamed.
Service KPIs
1. On-time delivery. Deliveries within the agreed window divided by total deliveries. The trap is the window definition: a four-hour window flatters the number, a fifteen-minute window destroys it, and comparing two sites with different windows is meaningless. Fix the window before you compare anything. Source: visibility platform arrival events, not the carrier’s self-report.
2. On-time in-full. Deliveries both on time and complete, divided by total. Always lower than on-time delivery, and the gap between the two is the quality of your loading rather than your transport. Where the retailer applies penalties against it, the mechanics are set out in the guide to OTIF in road freight. Source: WMS for the in-full half, visibility for the on-time half.
3. Transit time reliability. Standard deviation of actual transit time on a lane, not the average. The average tells you nothing about whether the receiver can plan. A lane averaging 26 hours with a 2-hour deviation is a better lane than one averaging 24 hours with a 9-hour deviation, and most scorecards rank them the other way round.
4. ETA accuracy. Share of shipments where the predicted arrival fell within a defined tolerance of actual, measured at a fixed horizon such as four hours out. Measuring accuracy at the moment of arrival is self-congratulation. Source: the prediction engine’s own history, compared against gate events.
Cost KPIs
5. Cost per kilometre. Total lane cost divided by kilometres run. The most quoted and least comparable KPI in the industry, because it moves with fuel, tolls, cabotage exposure and empty running. Only compare it within a lane type, never across a network.
6. Cost per pallet or per delivery. The number the commercial side actually needs, because it maps to the goods rather than the vehicle. Track it alongside cost per kilometre: when the two diverge, the loading is changing, not the transport market.
7. Empty running percentage. Kilometres run empty divided by total kilometres. Below 15% is strong for a European road operation, above 25% means there is a structural backhaul problem rather than a planning one. The strategies that move it are collected in the guide to reducing empty miles, and the arithmetic behind the cost sits in the freight calculators.
8. Detention cost per load. Total detention charges divided by loads. Most operations cannot produce this number because arrival and departure timestamps come from drivers rather than from systems. Until it is measured it is absorbed silently, at a scale set out in the guide to detention and demurrage.
Asset and flow KPIs
9. Vehicle fill rate. Volume or weight loaded divided by usable capacity, whichever constrains first. Report which one constrains: a fleet running at 95% weight and 60% volume has a completely different problem from the reverse, and a single blended number hides both.
10. Dwell time at site. Gate-in to gate-out. The KPI that most often exposes that a transport problem is really a warehouse problem, and the one that most needs automatic gate timestamps to be credible. Method in the guide to yard dwell time.
11. Tender acceptance rate. Loads accepted by the first-choice carrier divided by loads tendered. The leading indicator of a rate problem: acceptance falls before the market price shows up on any invoice, typically four to six weeks earlier.
12. Carbon per tonne-kilometre. Emissions divided by tonne-kilometres moved. Now a reporting obligation rather than a nice-to-have for companies inside the CSRD scope, and the one KPI where the calculation method has to be stated alongside the number or the number is not auditable.
Benchmarks, and why most of them mislead
Published European benchmarks are worth using as a sanity check and nothing more, because lane type dominates every one of these numbers. Long-haul international, regional distribution and urban delivery produce ranges that barely overlap: an urban operation at 40% empty running may be performing well, while a long-haul lane at 40% is failing badly.
Rough European road-freight ranges, for orientation only: on-time delivery 92% to 97%, OTIF 88% to 94%, empty running 15% to 25%, vehicle fill 70% to 85%, dwell time 45 to 120 minutes. If your number sits far outside the range, check the definition before you check the operation. Nine times out of ten it is the definition.
Where KPI programmes fail
Three failure modes, in order of frequency:
- No named source system. The same KPI is calculated two ways by two teams and the monthly meeting is spent reconciling rather than deciding. Every KPI needs one system of record, written down.
- Self-reported timestamps. Arrival times supplied by the party who would pay the penalty. This invalidates on-time delivery, detention and dwell simultaneously.
- Targets set from last year rather than from capability. A target the operation cannot physically hit produces gaming rather than improvement, and the first thing gamed is always the window definition.
Start with four: on-time in-full, empty running, dwell time and detention cost per load. Each has a clear owner, each is fed by a system rather than a person, and between them they cover service, asset use and the cost that hides. Add the rest once those four are trusted.
