How to optimize delivery routes: a practical guide for fleets

Every morning, in hundreds of warehouses, someone opens a spreadsheet and starts building the day’s delivery routes. They know the customers by heart, know which roads to avoid, remember who needs their goods before 10am. It works, but it’s slow, hard to reproduce and almost impossible to verify: nobody really knows whether those routes are the most efficient possible, or whether every day burns avoidable kilometres and hours.

Optimizing delivery routes means moving from relying on experience alone to actually measuring. This guide covers what that means in practice, which constraints genuinely matter, what changes when you deliver in Italy, and how to move from manual planning to a system that computes optimal routes in minutes. It also covers the questions that come up less often and are worth more: how many vehicles you really need, what a customer costs to serve, whether a fixed route still holds up.

What does optimizing delivery routes mean

Optimizing delivery routes means assigning stops to vehicles and ordering them in the sequence that minimizes an objective (usually distance, time or cost) while respecting every operational constraint. It’s not simply “finding the shortest path between two points”: it’s solving simultaneously which vehicle serves which customers, in what order, and at what time, across dozens or hundreds of stops.

It’s a mathematical problem known as the Vehicle Routing Problem (VRP). Its difficulty grows explosively with the number of stops: with just 20 deliveries the possible combinations already exceed what a person can evaluate by hand. That’s why manual optimization, beyond a certain scale, never produces the best solution: it produces an acceptable one, which is a different thing.

The constraints that actually matter

The difference between a theoretical path and a route that works lies entirely in the constraints. Serious optimization considers them all at once:

• Time windows: the customer only receives between 8 and 12, or has several slots in the day.
• Vehicle capacity: weight, volume, number of parcels, refrigerated compartments.
• Multiple warehouses: departures and returns from different depots.
• Priorities and mandatory sequences: urgent deliveries, pickups before drop-offs, fixed points.
• Urban restrictions: low-emission zones (LEZ/ZTL), time-based access limits.
• Driving and rest times: for fleets subject to the EU Mobility Package, routes must respect the limits of Regulation 561/2006.
• Available resources: vehicles and drivers actually on duty that day, owner-operators included.

Ignoring even one of these constraints produces routes that look great on paper but fail in reality, which is exactly why generic mapping tools fall short in real distribution.

In practice the constraints that hurt most are often more specific than these: the van with three separate temperature compartments, goods picked up in the morning that have to go back out in the afternoon, a transfer that happens in a lay-by rather than at the depot. Those are the cases collected in the OptivoRoute constraint library.

What changes in Italy

Italian distribution adds a few constraints that decide whether a plan survives its first morning, and that generic mapping tools rarely model:

• Each city writes its own rules for limited traffic zones (ZTL): access hours, permits tied to the individual vehicle, size limits in historic centres. The shortest route on the map can be one that a given van is not allowed to drive. The rules changing urban deliveries in 2026 are collected in a separate guide.
• Many distributors run mixed fleets: company vehicles paid by the hour next to owner-operators (padroncini) paid per delivery or per day, who often start and end their route away from the depot. The two costs have to sit in the same calculation, as in route planning with owner-operators.
• From 1 July 2026 the EU driving-time rules also reach vans above 2.5 tonnes used in international transport and cabotage, so for part of the light fleet the driving limits enter the plan for the first time.
• Whole sectors run on fixed routes: pharmaceutical wholesalers serving pharmacies on daily rounds, retail chains with dock slots booked at the distribution centre. Those routes drift as customers and volumes change, which is the case described in fixed-route optimization in pharmaceutical logistics.

These are the cases to test first on any tool, using the hardest day of the month rather than an average one.

Manual planning vs automated optimization

The point isn’t to replace the planner: it’s to free them from the mechanical work of building routes by hand, so they can focus on exceptions and data. It’s the shift we describe in moving from Excel to automated route planning.

How to optimize delivery routes in 6 steps

1. Collect stop data: addresses, time windows, quantities, any notes. A perfect format isn’t needed: modern systems clean raw data, and with Magic Order in OptivoRoute you bring them in through a chat in whatever format they arrive, Excel, PDF, photo or text, and it asks when a value is missing.
2. Define the real constraints: vehicle capacity, depots, priorities, urban restrictions, driving times.
3. Set the objective: minimize distance, total time, number of vehicles or cost. These are different objectives and lead to different routes.
4. Run the optimization: the engine computes the best assignment and sequence within the constraints.
5. Simulate and compare: try alternative scenarios (one more vehicle, a different window) and compare the impact on km and cost before deciding. Public tenders call for the same exercise before the bid is signed, because in the transport plan of a technical bid vehicles and kilometres turn into points.
6. Execute and measure: send the routes to drivers, track progress and compare planned vs actual to improve the next round.

The last step closes the loop: without measuring the real result, optimization stays a theoretical exercise.

From tomorrow’s routes to fleet decisions

The value of route optimization is usually read on tomorrow’s plan: fewer kilometres, less planning time. The same engine answers three questions that come up less often and weigh more on the income statement:

• How many vehicles do we really need? Take one out of the scenario, recalculate, and check whether the routes still fit the time windows and the drivers’ shifts. If they do, the saving is the vehicle’s full annual cost and, potentially, a driver’s.
• What does it cost to serve this customer, or to open this area? The answer needs the full cost of the stops, not the marginal cost of adding one to a route that leaves anyway. A decision on a customer or a delivery frequency commits a vehicle and a driver for hours.
• Do our fixed routes still hold up? New customers, shifting volumes and changed windows pull them away from optimal a little at a time. Recalculating them every month and changing only where it pays recovers what has drifted.

The answer is usable because it comes with the routes. A spreadsheet can say that two vehicles fewer will do; an engine that respects your constraints says it by showing the routes that work without them, stop by stop.

In Italian distribution this is where most of the money is, because the cost of a delivery is mostly time rather than kilometres. In our model of a pharmacy delivery built only from public sources (Ministry of Infrastructure cost tables, the national collective agreement, Federfarma and ISTAT), between 49% and 64% of the full cost is time. Decisions that remove hours or vehicles therefore weigh more than those that shave kilometres. On OptivoRoute the three horizons run on the same engine and the same data.

What to look for in route optimization software

Not all software is equal. For real-world distribution, the criteria that make the difference are:

• Raw data import: being able to upload your Excel file as-is, without reformatting, lowers the adoption barrier; the same goes for an order that arrives as a PDF, a photo or plain text.
• Industry-specific constraints: grocery, pharma, food & beverage have different rules; the software should adapt to you, not the other way round.
• Urban awareness: LEZ/ZTL must enter the calculation, not be patched in by hand afterwards.
• Scenario simulation: being able to evaluate decisions (one vehicle fewer, a new client, a fixed route to redraw) on your own data and constraints, compared with today’s plan, before making them.
• Execution and proof of delivery: driver app and digital POD to close the cycle.

To go from criteria to a choice, our guide on how to choose route optimization software in Italy covers the families of tools, the constraints to test first and twelve questions to ask a vendor.

It’s exactly the approach of OptivoRoute, our route optimization software: bring the orders in through Magic Order, as Excel, PDF, photo or text; the AI builds the routes within your real constraints, and you execute them with the driver app.

How much do you actually save

The first saving is time: planning goes from half a day to a few minutes. On costs, the order of magnitude comes from the reference literature on the Vehicle Routing Problem. Real-world applications in North America and Europe show that planning with computerized procedures generally saves between 5% and 20% of transportation costs (Toth and Vigo, SIAM 2002). Where your fleet falls in that range depends on volumes, constraints and how good your starting planning is; the figures Optivo publishes are on the results page.

For a first estimate on your own numbers—vehicles, stops, km and planning hours—try the delivery route savings calculator: 60 seconds, result as a range, no personal data required. For the full quantification method, see our deep dive on the ROI of route optimization software.

Frequently asked questions

The next step

Want to see optimization applied to your real routes? Discover OptivoRoute or book a demo with your data.