Fleet managers do not have a data problem. They have a too-much-data problem. Telematics, TMS, ERP, delivery apps: every system generates metrics, dashboards, reports. The result, paradoxically, is that nobody looks at anything carefully, or they end up focusing on the wrong indicators.
Industry research confirms that data overload is the primary pain point for fleet managers. Not because data is useless, but because without a clear hierarchy of indicators, it becomes impossible to separate signal from noise.
This article isolates the 7 KPIs that truly matter in last-mile delivery fleet management, explains how to interpret them and, where they exist, gives the measured values with their sources.
1. Cost per delivery
The most direct KPI, and often the most overlooked in its entirety. Cost per delivery is not just fuel and tolls: it includes vehicle depreciation, driver labor cost, administrative overhead, the potential cost of a redelivery attempt, and the share of fixed costs allocated per stop.
How to calculate it: total fleet costs for the period / number of successfully completed deliveries.
How to read it: there is no single reference value, because it depends on what you count. In a Politecnico di Milano model of Milan, the home delivery round alone costs the carrier €3.05 per parcel (Seghezzi and Mangiaracina, 2023). In our full-cost model of a pharmacy delivery, built only from public sources and including vehicle, driver and unloading time, it comes to €14.17. The trend matters more than the absolute value: compare the figure with your own history and across areas, customers and vehicles.
How to improve it: route optimization is the primary lever. Reducing deadhead kilometers and increasing the number of stops per route lowers the unit cost directly. Those interested in the return on investment of route optimization software will find concrete figures to work with.
2. First-attempt delivery rate
How many deliveries are completed without requiring a second attempt? This indicator measures the operational effectiveness of the entire chain: from address quality to time-window planning, to communication with the recipient.
How to calculate it: deliveries completed on first attempt / total planned deliveries x 100.
How to read it: in parcel deliveries in Italy the first attempt succeeds roughly nine times out of ten. In the second quarter of 2025 the share was 91% according to tracking data from Parcel Perform, which sells shipment tracking software. Every point lost is an avoidable, measurable cost: a second attempt is another round paid for the same goods.
How to improve it: the causes of failed deliveries are often predictable. Incomplete addresses, time windows incompatible with recipient habits, lack of specific access instructions. Those who systematically analyze the reasons behind each failed first-attempt delivery discover recurring patterns that can be addressed.
3. Km driven vs optimal km
The ratio between the kilometers actually driven by the fleet and those that an optimization algorithm would calculate as the ideal route. This KPI measures how far current planning deviates from theoretical efficiency.
How to calculate it: total km driven / km calculated by optimized route x 100. A value of 100% indicates perfect adherence to the optimal plan.
How to read it: there are no measured thresholds for this ratio, and some deviation is normal (traffic, road closures, unexpected events). What matters is the trend in your own history, vehicle by vehicle and route by route: a gap that keeps growing, or that concentrates on a few routes, shows where the plan does not hold.
How to improve it: the first step is visibility. Without an optimization system that calculates the ideal route, this KPI is not even measurable. The second step is analyzing deviations: are they due to traffic, driver choices, or planning that does not account for real conditions? Historical data and predictive models allow many of these variables to be anticipated.
4. Vehicle utilization (%)
What percentage of available capacity is actually being used? A 12-cubic-meter van that goes out with 4 cubic meters of goods represents a fixed cost working at a third of its potential. Likewise, a vehicle sitting idle in the depot is immobilized capital.
How to calculate it: two dimensions to monitor in parallel.
- Volume/weight utilization: volume (or weight) carried / maximum capacity x 100.
- Time utilization: hours or days of actual use / hours or days available x 100.
How to read it: first decide what you measure, the load at departure or the average over the whole round, because the two give very different numbers. Averages measured over the whole trip are low. For diesel vans the European framework for calculating logistics emissions (GLEC) uses a combined load and empty-running factor of 36% (ADEME, Base Carbone). There is no threshold that fits everyone: compare utilization with your own history, and a drop that lasts for weeks signals one vehicle too many.
How to improve it: the key is intelligent matching between orders and vehicles. A planning system that simultaneously considers volume, weight, time windows and delivery zones can maximize the load of each vehicle, reducing the number of vehicles needed.
5. Fuel consumption per km
With fuel typically representing 25-30% of a fleet’s operating costs, monitoring consumption per kilometer is not optional. It is a necessity. And for those managing mixed fleets with electric vehicles, the KPI extends to energy consumption per km (kWh/km). Data granularity depends on the tracking mode you adopt: an OBD device reads consumption when the ECU exposes it, a CAN bus integration accesses richer parameters on heavy vehicles, while the manufacturer’s cloud returns real consumption on modern and electric natively-connected vehicles without any onboard hardware.
How to calculate it: liters (or kWh) consumed / km driven. Monitor by individual vehicle and by individual driver.
How to read it: there is no reference value that fits every van, because consumption varies a lot with the model, the load and the type of route. The comparison that works is internal: vehicles of the same model on the same routes, and each vehicle against its own history. A lasting gap points to a maintenance, driving style or route planning problem.
How to improve it: three main levers. The first is route optimization (fewer km = less fuel). The second is driver training (harsh acceleration, high speeds, idling). The third is preventive maintenance (tire pressure, engine condition). For fleets transitioning to electric vehicles, CSRD reporting requirements already mandate granular tracking of these metrics.
6. On-time delivery rate (%)
In a market where delivery windows are becoming increasingly tight, punctuality is no longer a “nice to have.” It is a contractual requirement in B2B and a retention factor in B2C. This KPI measures the fleet’s ability to meet agreed time windows.
How to calculate it: deliveries made within the agreed time window / total deliveries x 100.
How to read it: there are no measured thresholds that separate B2B from B2C. The definition matters more than the number. Decide first what counts as on time (inside the agreed window, or within a margin of minutes) and always measure the same way, because two rates calculated differently cannot be compared. The figure to worry about is the one getting worse in your own history.
How to improve it: punctuality depends on planning quality. Overestimating or underestimating travel times, not accounting for traffic variability by time of day, allocating too many deliveries per route: these are all planning errors that a system based on historical data and real-time updates can correct.
7. Average time per stop
How long does a driver take on average to complete a single delivery, from the moment of arrival at the address to the moment of departure? This KPI is often underestimated, but it has an enormous impact on fleet productive capacity.
How to calculate it: sum of stop times / total number of stops. Exclude regulatory breaks.
How to read it: field surveys give the order of magnitude. In parcel deliveries in central London the service time is about 4 minutes per customer (Allen et al., 2018). Deliveries to shops and offices with vans and trucks take longer. In Seattle the average dwell time is about 17 minutes (Kim et al., 2021). In the station district of Luxembourg City it is 18 minutes (LIST). Values that grow in your own history point to process problems (paper documents, searching for the recipient, complicated access points).
How to improve it: time per stop is reduced with better information available to the driver (access instructions, recipient contact, delivery notes) and with digitization of processes at the delivery point. A digital Proof of Delivery system, for example, eliminates paper form completion and speeds up stop closure.
From measurement to action
The value of these 7 KPIs does not lie in measurement for its own sake. It lies in the ability to turn numbers into operational decisions. A fleet manager who consistently monitors these metrics knows where to intervene, with what priority, and can measure the effect of every change. The ability to read these data correctly is also one of the most reliable criteria for identifying who can do well as supervisor or fleet manager: those who face a number and ask questions are coordination candidates; those who only use it to justify themselves, less so.
The practical advice is to start with the KPIs that are getting worse the most in your own history. If the first-attempt delivery rate is at 85%, that is the priority. If one vehicle uses much more fuel than the others of the same model, start there. You do not need to improve everything simultaneously: you need to know where you are losing the most and focus resources on that point.
The data is there. The difference is made by those who read it the right way.
In the glossary: Fleet utilization rate · TCO