Tactical Delivery Planning: Why Balanced Routes Cost You Kilometres

If your customers are visited on a rhythm rather than on demand, you are in a periodic delivery business. Bottled water, coffee, hygiene services, gas cylinders, linen: the products differ, the planning problem is the same. Every customer has a frequency, every frequency has to land on a day, and every day has to fit in a van. Most planning tools solve that by balancing. Every day gets the same number of stops, every week the same number of hours, every driver the same load. It looks fair and it looks efficient. It is neither, and the reason is geometry rather than software.

This article is about the tactical layer of planning: the structure you decide once and then live with for a year or more. Not which stops go in which order tomorrow morning, but which part of the map belongs to Monday, which part of Monday belongs to week two, and where the boundaries fall. Get that layer right and the daily routing software has an easy job. Get it wrong and no amount of daily optimisation will bring the kilometres back.

Start with a square

When I sit down with a planning team I draw a square and call it the service area. Then I split it into five regions, one per weekday: four corners and a rectangle in the middle. That is the whole first step, and it is the step most tools skip because they start from the customer list instead of from the map.

The lines between the regions are the easy part, once you look at the map instead of the spreadsheet. You follow whatever a driver would not want to cross twice in one day: a highway, a canal or river, a park, a forest, a railway line. Those features already divide the area into natural pieces, and customers on either side of them are further apart in driving time than they look on the screen. A boundary that follows the highway costs nothing. A boundary that cuts through a town centre costs a crossing every single day.

Notice what the square does not do. It does not make the five days equal. The middle rectangle, closest to the depot and usually the densest, is the shortest day in kilometres. The corners are the long ones. A balanced tool would immediately start moving customers from the corners into the middle to even that out, and in doing so it would start sending the Wednesday van into Monday’s corner. The square says: leave it. Short days are not a problem to be solved.

Then split a day into weeks

A weekday region is still too big for one van in one day, because not every customer in it is due every week. So the second step is to split each day into weeks, and here the frequency mix decides the shape. If most customers in the region are visited every two weeks, the region is split in two halves: week one takes the west, week two the east. If most are on a four-week rhythm, it is split in four quarters. The majority frequency sets the grid; you do not try to serve three frequencies with three different grids at once.

The customers who do not fit the majority, typically a small number on a weekly rhythm, are simply visited on the way. They are few, they are spread out, and their extra distance is small next to what the split saves. What matters is that the bulk of the stops on any given day sit inside one half or one quarter of the region. The van drives to that part once, works it, and comes home.

The highway with three exits

The reason this matters is easiest to see on a single road. Picture a highway leaving the depot with three exits along it, customers to the left and right of each. A balanced plan spreads each week’s stops along the whole stretch, because that is what keeps the weeks equal. So every week the van gets off at exit one and somehow has to track its way to exit three and back, or the reverse. It reaches the far end every single week.

The plan I put in forces the van to work one side of the highway between two neighbouring exits. Week one: exit one to exit two, north side. Week two: the same stretch, south side. Weeks three and four: exit two to exit three, north side and then south. Both plans serve the same number of stops per week. The stops are still balanced. But half the weeks now end at the second exit, which means the van does not need to drive to the furthest exit fifty percent of the time. That single fact, repeated across every day and every region of the square, is where the kilometres go.

Watch the counters rather than the vans. Both plans finish the same six stops every week. The balanced van is still on the road when the other one is already back at the depot, and after four weeks the gap is about a fifth of all driving in this toy example, whether the balanced van crosses over at exit two or drives the full circle round the edge roads. In a real service area the stretch is not one road but a whole corner of the square, and the effect compounds with the day split above.

Why planning tools still prefer balance

If the geometry is this simple, why do so many long-term planning solutions still optimise for an even week? A few reasons, and none of them are stupid.

• Balance is the easiest thing to compute and the easiest thing to defend. Stops per day and hours per driver are visible on a dashboard on Monday morning. Kilometres per stop only show up in the fuel invoice a month later, attributed to nobody.
• Balance is what the people in the room ask for. Drivers compare their days with each other, dispatchers get the complaints, and an uneven week feels unfair even when every day is shorter than before. A tool that promises equal days sells itself in the demo.
• Tools start from the customer list, not the map. A solver that is handed ten thousand stops and asked to make twenty equal days will happily interleave regions, because nothing in its objective tells it that crossing the canal costs more than the straight-line distance suggests.
• Daily re-optimisation hides the structural cost. If the routing software rebuilds every route each morning, the day always looks locally optimal. Nobody sees that the optimal route for a badly cut region is still a long one.
• Balance survives change without anyone touching the plan. When a customer joins or leaves, a balanced system rebalances quietly. A territory plan needs someone to decide, once in a while, that a boundary has to move. That is a governance job, and tools are not good at those.

Balance the stops inside a region, not the kilometres across the week. An uneven week of short days beats an even week of long ones every time.

What it did in practice

These are my own numbers from the periodic delivery operations I have planned this way as a consultant, not an industry benchmark. Across those projects the average distance driven per van fell by between 15 and 30 percent compared with the balanced plan it replaced. Fuel consumption fell by more, between 20 and 40 percent.

The gap between the two numbers is not a mystery. Kilometres on the highway and on regional roads are the cheap ones: a truck at steady speed burns far less per kilometre than one pulling away from a kerb. The expensive moments in a delivery route are stopping and accelerating back up to speed, and the compact plan removes a lot of them. When a day’s stops all sit in one half of one region, customers on the same street or at the same location end up on the same day, so there are fewer actual stops to make and fewer stop-and-go cycles between them. Fewer kilometres, and a better kind of kilometre.

The stop counts did not change. The customers did not change. The vans and drivers did not change. What changed was which part of the map each of them was pointed at on a given day, decided once, and then left alone.

How to apply it

1. Draw the square before you open the customer list. Five regions, boundaries on the highway, the water, the parks and the forests. Argue about the lines on a printed map, with the drivers in the room.
2. Find the majority frequency per region and split the day by it: two halves for a two-week rhythm, four quarters for a four-week one. Let the minority frequencies ride along.
3. Accept that the days will not be equal. Measure kilometres per stop and fuel per stop instead of stops per day, and report those to the same people who used to see the balance.
4. Give the tactical plan to the daily routing tool as a constraint, not as a suggestion. Its job is to order the stops inside today’s half or quarter, not to redraw the map every morning.
5. Revisit the boundaries once a year, or when a region’s customer base has clearly shifted. Moving a boundary is a decision, so make it one, with a date and an owner.

None of this needs new software. It needs a map, a square, and the willingness to let Wednesday be shorter than Monday. The tools will balance whatever you give them. Give them a region instead of a week, and they will balance that.

Further reading

A few pieces from tool vendors and researchers that circle the same ground, for anyone who wants the vocabulary the software uses for it.

PTV Logistics: route planning systems for tactical transport planning

Descartes: fixed route planning, why repeatable master routes pay off

Felt: logistics route planning, including two drivers crossing the same bridge minutes apart

Territory design for the multi-period vehicle routing problem with time windows (Computers & Operations Research)