Ask how many delivery points one depot should serve and you will be given a number. The number is always wrong, because a depot does not serve points. It serves vehicle-days, and a delivery point can consume anything from two minutes of one to the whole of one.
Two operations with three hundred customers each can need one building or three. The difference is not the count. It is how the points sit relative to the gate, and how much work each one takes when the truck arrives.
This matters because the depot is the most expensive answer to a capacity problem, and the only one that also adds stock. Getting to it too early is a mistake that takes years to reverse.
What one depot's day actually looks like
A published Thai study followed a next-day parcel operation with a single collection centre in Phitsanulok on the busiest day of its week. Forty-one customer and franchise points, three closed-body pickups, one working window from 13:00 to 19:00. Every vehicle left the gate at 13:00.
Here is what came back.
| Vehicle | Points served | Distance | Time used | Of the 360 minutes |
|---|---|---|---|---|
| 1 | 4 | 15.86 km | 176 min | 48.9% |
| 2 | 20 | 29.60 km | 204 min | 56.7% |
| 3 | 17 | 120.26 km | 347 min | 96.4% |
Same depot, same afternoon, same six hours. One vehicle managed four points and another managed twenty. The one that managed four drove less than any of them.
Look at what each vehicle was actually limited by, because all three were limited by something different.
Vehicle 1 was limited by volume at one point. It drove for 35 minutes all day. It spent 135 minutes standing at a single customer collecting 451 pieces. Its other three points cost six minutes between them. Four points filled an afternoon because one of them was a warehouse in disguise.
Vehicle 2 was limited by the number of stops. It drove 102 minutes and stood still for 102 minutes. Of that standing time, 60 minutes went on one large customer and the other nineteen points shared 42 minutes, about two minutes each. Twenty points cost less than four did.
Vehicle 3 was limited by distance. It covered 120.26 kilometres, more than the other two put together multiplied by two and a half, and finished with thirteen minutes of its window left.
That last line is the whole answer to the original question. This depot is full at 41 points, but not because 41 is its limit. It is full because of the seventeen points in the far group. Another customer in vehicle 2's cluster would cost about two minutes. Another customer in vehicle 3's would cost a fourth vehicle.
The route that will not improve is the one telling you something
The same study re-planned the day with routing software and compared the result against what the drivers had actually done. Vehicle 1's time fell 41.33%. Vehicle 2's fell 38.18%. Vehicle 3's fell 3.61%.
That gap is the most useful diagnostic in this article, and it costs nothing to run on your own data.
A route that gets substantially shorter when you re-sequence it was badly ordered. That is a planning problem, and planning is cheap to fix. A route that will not improve no matter how you re-sequence it is not badly ordered. Its time is geography, and no amount of scheduling will take it out. The truck is driving to where the customers are.
When the routes that refuse to improve are all pointing the same way from the gate, that is the first honest signal that the territory has outgrown the building. Until then, what you have is a plan that needs work, and a second depot would be buying a solution to a problem you do not have.
The unit to count in
Stop counting points and count drop slots. One depot's weekly capacity is:
vehicles Γ days worked per week Γ drops per vehicle-day
and the number of delivery points that fits behind it is that figure divided by how many visits each point gets in a week.
The middle term is the one that moves, and the Phitsanulok day shows how far it moves inside a single operation. Take eight vehicles working six days and run the three observed route shapes through it.
| Drops per vehicle-day | Drop slots per week | Points at 1 visit a week | Points at 2 visits a week |
|---|---|---|---|
| 4 | 192 | 192 | 96 |
| 17 | 816 | 816 | 408 |
| 20 | 960 | 960 | 480 |
Same eight trucks. Same building. Between 96 and 480 delivery points, a five-fold range, decided entirely by which route shape the territory produces.
This is why a benchmark from another company is worthless here. Somebody who tells you a depot in Thailand serves four hundred points is describing their drop size and their customer density, not yours.
The outer edge of the territory is fixed by law rather than by preference. Section 103 bis of the Land Transport Act limits a licensed driver to four continuous hours from the moment of starting, and after an unbroken rest of at least half an hour to a further period of not more than four consecutive hours, within a twenty-four hour cycle. It prescribes hours and no distance. Whatever the far edge of your territory is, the truck has to reach it and come back inside that structure, and the study's assumed 90 km/h is a speedometer number rather than a door-to-door one. How long a Thai lane really takes door to door is the part that decides where the edge falls.
Before a second depot: check the points are on the right one
An ice manufacturer in Chonburi delivers to more than four thousand customers a day from three distribution centres. A study took 35 delivery points running on six vehicle routes, all of them attached to DC3, and asked a simple question: given the three centres the company already had, which centre should each point belong to?
Ten of the thirty-five were on the wrong one. Nine belonged to DC1 and one belonged to DC2.
The total point-to-centre distance fell from 243.57 to 177.62 kilometres per trip per day, a reduction of 27.08%. At the 3.5 baht per kilometre the study uses for fuel, the 65.95 kilometres removed is about 231 baht a day. Point N13 was the clearest case: 23.975 kilometres from the centre serving it, and 7.881 kilometres from one it was not.
Two things about this are worth taking seriously.
The measure is depot-to-point distance, not driven route. That is the honest reading of the numbers, and it is still the right measure for this question, because it is the part of the distance that a change of depot actually moves.
The model minimised load multiplied by distance, not distance. A heavy point close to the gate matters more than a light point far away. If you redraw territories on a map by eye, you will get this backwards, because a map shows you distance and hides volume.
Zone boundaries are almost always drawn once, when the second site opens, and then left alone while the customer base moves underneath them. Redrawing them costs an afternoon and no capital. Do it before anything else.
What a second depot actually changes
Only one line of cost falls when you add a site: local delivery. That is because delivery distance splits into stem distance, the run out to the delivery area and back, and drop distance, the running around once you are there. A second depot shortens the stem. It does nothing to the drop.
Everything else moves the other way. Storage cost rises, because two medium buildings cost more than one large one. Primary transport rises, because stock now has to be trunked to two places. Inventory holding rises, because each site needs its own buffer. Systems and management cost rises. Add the curves together and the total has a minimum somewhere, but where that minimum sits does not travel between companies. In the worked example those curves come from, the least-cost point falls at around six to eight sites, and the accompanying text is explicit that the real answer depends on product type, the geographic spread of demand and the service level promised. The shape of the argument transfers. The number does not.
So the question is never "are we full". It is "is the stem distance we would remove worth more than everything else we would add".
The four cheaper answers, in order
Work down this list before pricing a building. Each one is faster to try and faster to undo.
1. Re-attach the points to the nearest depot you already have. Free. Rank every delivery point by volume times distance to each site, not by distance alone, and move the ones that come out wrong. If you have only one depot this step does not exist, which is worth knowing, because it means a single-depot operation reaches the expensive answers one step sooner than everyone else.
2. Change the visit frequency of the far cluster. The far group is expensive because you keep driving there. Halving how often you go halves the stem, and the arithmetic of when less frequent delivery costs less usually clears the far cluster before it clears anything closer in.
3. Add a second departure wave rather than a second site. If what is full is the yard and the loading bays between 05:00 and 07:00 rather than the working day, you have a dock problem wearing a depot costume. A depot fixes distance. It does not fix everybody wanting to load at the same time.
4. Buy the far tail as a lane instead of running it. The seventeen points at the end of a 120-kilometre run are a route, not a region. A transportation company already working that corridor can carry them without you owning anything, and the decision is reversible next month.
The test that says yes
If all four fail, price it properly. The gate is not the transport saving on its own, it is the transport saving against the standing cost of the building.
Krungsri Research puts average ready-built warehouse rent in 2024 at 110 to 230 baht per square metre per month, averaging 165 baht across the Bangkok Metropolitan Region and the Eastern Economic Corridor, and 143 baht in the central region outside it. By province the averages ran from 155 baht in Samut Sakhon and 175 in Ayutthaya up to 200 in Chonburi and Chachoengsao and 230 in Samut Prakan. Occupancy was 88.0%, the highest in years, and is forecast to ease to 85 to 86% through 2027, so there is a little more room to negotiate ahead than behind.
Take a 3,000 square metre satellite at the 165 baht average. That is 495,000 baht a month, or 5,940,000 baht a year, for the floor alone. No staff, no racking, no stock, no systems.
Now convert it into the only unit that lets you compare it with the saving. At an all-in 8,000 baht a vehicle-day, and a vehicle working 250 days a year, the rent alone is:
- 742 vehicle-days a year
- just under three trucks removed from the plan, every day, for a full year
- or about 23,760 baht of saving needed on every operating day
Put the same building in Samut Prakan at 230 baht and the same 3,000 square metres costs 8,280,000 baht a year, which is 1,035 vehicle-days, or 4.1 trucks. The address alone moves the gate by more than one truck's worth of annual saving.
If your best case is that the second depot takes one or two vehicles out of the plan, the answer is no, and it was no before you started drawing floor plans. If your routes show four or five vehicle-days a day spent purely on stem distance in one direction, it is worth a proper study, and where the site goes then becomes the question that decides how much of that saving you actually collect.
What to do this week
- Pull one busy day from the tracking data. For each route, list points served, kilometres, and minutes from gate to gate.
- Split each route's minutes into driving and standing. Three routes from one depot will usually be limited by three different things.
- Re-sequence each route in any routing tool. Note which routes barely improve. Those are your distance problem. The rest are your planning problem.
- Rank every delivery point by volume times distance to each depot you own. Count how many are attached to the wrong one.
- Convert your fleet into drop slots per week and divide by visits per point. That number, not a benchmark, is your depot's capacity.
- Only then price a building, and price it in vehicle-days.
The count of delivery points behind one depot is an output, not a target. It falls out of drop size, customer density and how often you go. Change any of those three and the same building serves a different number of customers next month, without a brick being laid.