Ask a transportation company what makes a lane expensive and you will get an answer about the clock long before you get one about the map. Procurement asks the opposite question first. How far is it, how many kilometres, what is the rate per kilometre.
Distance is the cheapest thing in a freight quote to change your mind about. It is a running cost, mostly fuel and tyres, and it is roughly linear: twice as far burns about twice as much diesel. The delivery window is not like that at all. It is a claim on a block of a vehicle's working day at a fixed time, and it decides how many other jobs can share that day.
That is where the money is. Not in the kilometres.
What a window actually reserves
When you ask for delivery between 09:00 and 10:00, the carrier is not pricing an arrival. It is pricing a reservation, and the reservation is longer than the window:
- the travel time to reach you, which has to be planned to the slow case, not the average
- the window itself, because arriving at 09:05 does not release the vehicle at 09:05
- the time on site, which is a fixed amount per stop plus a variable amount per pallet or case
- whatever is left of the day afterwards, which is only worth something if it is enough to do another job
The standard scheduling arithmetic treats unloading as exactly that split, a fixed time per drop plus a rate per case, with the fixed part running from about ten minutes at an easy site to about twenty five minutes at a hard one. And it treats a second trip as only possible when at least two hours of the day are still free. That last rule is the whole cost story in one line. A window positioned so that the vehicle comes back with ninety minutes left has consumed the entire day. A window positioned so that it comes back with three hours left has consumed half of it.
Nothing in that calculation is about distance.
The same customers, five different sets of windows
There is a Thai measurement of exactly this, and it is unusually clean because only one thing was allowed to vary.
An automotive parts plant on the eastern seaboard delivers to 17 customers, all of them within about 12 kilometres of the plant. Total demand is 2,710 trays. Each vehicle holds 500 trays a round, so the arithmetic floor is six rounds. The reason the windows exist at all is worth noting: the customers had increased output, had no room to hold material, and so required delivery in several small drops a day at set times. The window was their warehouse cost, moved onto the plant's transport.
The researchers ran the same 17 customers, the same demands and the same map through the same planning method five times, changing nothing but the set of time windows. Here is what came out.
| Window set | Rounds needed | Distance | Average vehicle fill | Cumulative lateness |
|---|---|---|---|---|
| 1, the original windows | 7 | 130.56 km | 77.4% | 7 h 01 m |
| 2 | 6 | 116.66 km | 90.3% | 0 h 05 m |
| 3 | 6 | 116.54 km | 90.3% | 4 h 11 m |
| 4 | 7 | 136.88 km | 77.4% | 0 h 22 m |
| 5 | 8 | 141.73 km | 67.8% | 0 h 22 m |
Read the middle two columns together. Distance moves 21.6% across the five. The number of vehicle rounds moves by a third, from the six that the load arithmetic demands to eight. Same goods, same customers, same roads, same planner. The windows alone decided whether the operation ran at its theoretical minimum or a third above it.
The fill column says why. Six rounds means average loads of 452 trays out of 500. Eight rounds means 339. The window pattern was not making the vehicles drive further so much as sending them out part empty, and a part empty round costs very nearly what a full one costs.
A bunched window is a capacity problem, not a distance problem
The original window set is the interesting one, because it is what a real customer base actually asks for. The windows averaged 33 minutes, the tightest was 10 minutes and the widest 50. They fell into four bands:
| Band | Customers | Trays wanted in that band |
|---|---|---|
| 08:15 to 08:30 | 2 | 280 |
| 09:45 to 10:15 | 5 | 830 |
| 11:30 to 12:30 | 5 | 620 |
| 14:30 to 15:30 | 5 | 980 |
Look at the 09:45 band. Five customers want delivery inside the same half hour, and between them they want 830 trays. A vehicle holds 500. No amount of clever routing fixes that, because it is not a routing problem. Those five customers are close together, the furthest pair only about 2.9 kilometres apart in a straight line, so distance was never the obstacle. Two vehicles have to be standing at those five gates in the same thirty minutes because 830 is more than 500.
This is the general rule and it is worth stating plainly. A set of delivery windows sets a floor on how many vehicles have to exist at the same moment. That floor is a fleet size, and it is fixed before anyone opens a map. If you want to know why your carrier's rate looks high for a short lane, count the trays your customers want inside your busiest hour and divide by a payload. The same logic decides how much capacity is worth committing to for a normal week.
The shortest route is not the route that hits your window
The same study ran four planning methods over all five window sets. Three of them chase distance in different ways. The fourth, earliest due date, simply sequences the work by deadline and accepts whatever distance falls out.
| Method | Average distance | Cumulative lateness |
|---|---|---|
| Max-Nearest, the shortest | 121.94 km | 29 h 12 m |
| Nearest | 125.91 km | 17 h 14 m |
| Earliest due date | 128.47 km | 2 h 24 m |
| Saving algorithm | 135.83 km | 32 h 33 m |
The shortest plan on the map missed the customers' windows by more than twenty nine hours in total across a day of work. Sequencing by deadline instead cost 5.36% more distance and cut that to two hours and twenty four minutes.
Then the researchers applied it for real, on measured road distances rather than coordinates. The result was 222.79 kilometres, one customer late, and that one by nine minutes. The number that matters to a buyer is the last one they report: the share of deliveries the plant had to hand to hired-in outside vehicles fell from 59.99% to 29.41%.
That is the mechanism in a sentence. A window you cannot hit turns your own capacity into somebody else's invoice. Sixty per cent of that plant's work was going to bought-in trucks, not because it lacked vehicles but because its plan could not get its own vehicles to the right gate at the right minute. Nothing about the geography changed to fix it.
What the invoice actually counts
If the window is expensive because of vehicle rounds, then the bill should move with rounds and not with kilometres. A second Thai case, a transport operator studied over six months, shows exactly that.
The operator had been running one vehicle per customer per trip. After grouping customers into shared trips:
| Before | After | Change | |
|---|---|---|---|
| Trips per month | 87 | 49 | β43.68% |
| Kilometres per month | 5,810.75 | 4,279.50 | β26.35% |
| Transport cost per month | 162,500 baht | 91,500 baht | β43.69% |
The cost fell 43.69% and the trips fell 43.68%. Distance fell only 26.35%. Work out the cost per trip and the reason is obvious: 1,867.82 baht before, 1,867.35 baht after. It did not move at all. Every baht of a 71,000 baht monthly saving came from making 38 fewer trips, and the kilometres were incidental.
Cost per kilometre, meanwhile, appeared to improve, from 27.97 to 21.38 baht. That is not a saving, it is arithmetic: an unchanged trip charge spread over more kilometres, because a grouped trip covers 87.3 kilometres where a single-customer trip covered 66.8. Anyone reporting this operation in baht per kilometre would have announced a 24% gain and missed the 44% one.
So when you are told a tight window costs more, the honest version of the claim is not "we will drive further". It is "fewer of your drops will fit in one trip, and trips are what we sell".
Half of your window is spent at your own gate
The other end of the window is the one procurement forgets, because it is not on the invoice.
Researchers at Chulalongkorn University measured a Thai manufacturer's busiest warehouse from the moment a driver joined the queue to load until the vehicle left. The average was 139 minutes. Fifty five per cent of trucks were there more than an hour. The actual loading was 23.9 minutes for a flatbed and 44.3 minutes for a container vehicle, so most of those 139 minutes was neither loading nor queueing at the dock: it was everything else in the visit.
Then they tested four fixes. The one that worked on the queue worked spectacularly, cutting the wait at the loading point from 14.9 to 1.1 minutes for flatbeds and 21.7 to 1.3 minutes for container vehicles. It cut the total visit by 3%.
That is the useful and slightly deflating finding. Adding dock capacity removes the queue and leaves the visit. If your outbound gate consumes two hours and twenty minutes of a vehicle's day, then a customer window three hours away is a whole day, and no window you negotiate at the far end will change it. There is a matching argument at the receiving end, where waiting time charges are the price of the same problem seen from the carrier's side.
What to do with this, in order
1. Put every window you promise on one clock and count the load in each hour. Not the number of customers, the payload. If any hour holds more than one vehicle's worth, you have committed to a second vehicle at that hour, and you are paying for it whether or not anyone has told you.
2. Ask each customer for two numbers, not a target time. The earliest they can physically receive, and the latest they will accept. Those two are what a planner needs. A "please deliver at 10:00" is a window of zero width, and it gets priced as one.
3. Express windows as a handful of codes rather than as individual times. All day, morning only, afternoon only, closed at lunch. Six codes covers most operations. The gain is not tidiness, it is that a code is negotiable and a time on a purchase order is not.
4. Test whether the window is real. In the Thai plant above, the windows existed because the customers had no storage. That is a real reason and it will not move. Plenty of windows have no reason behind them at all beyond the last person who set them, and those are free capacity sitting in your own paperwork.
5. Price it instead of arguing about it. Ask your carrier to quote the same lane three ways: a fixed 30 minute window, a two hour window, and any time that day. The spread between the three is the number you have been trying to guess. Most buyers never see it because they never ask for it, and a carrier cannot volunteer it without looking like it is upselling.
6. Fix your own loading gate before you buy a tighter window at the far end. Two hours at your dock is two hours off the other end of the day, and it is the half you control.
7. Plan the window against the slow case, not the average. A window is a promise about a distribution, not about a journey, and the gap between a map's answer and a real door-to-door time is where the promise breaks.
The one line worth keeping
Distance is a running cost and it behaves itself. A delivery window buys a block of a vehicle's day at a named hour, and it decides how many drops can share that day.
That is why the same 17 customers in the same 12 kilometres needed six rounds under one set of windows and eight under another. Nobody moved. The clock did.