Route optimization 101: more stops, fewer miles

Ask most distributors what route optimisation does and you will get an answer about maps and mileage. That is the part everyone can picture, and it is the smaller half of the problem. A route is not a travelling-salesman puzzle with deliveries attached — it is a day with hard edges: shops that only receive between eight and nine, a stop that genuinely takes twenty minutes, a driver whose shift ends whether or not the last drop happened.

Optimising for distance alone will happily produce a beautiful short loop that arrives at a locked door.

What a route is actually solved against

Five things, and only one of them is distance:

Drive time between consecutive stops — not straight-line distance, which is optimistic by roughly a third in most territories.
Receiving windows. A grocer with a 06:00–10:00 dock is a hard constraint, not a preference.
Service time per stop. A twelve-case drop and a full merchandising visit are not the same stop, and averaging them is how days overrun.
Vehicle capacity in weight and volume, which caps what the sequence can carry before it caps where it can go.
Maximum stops per run, because a theoretically feasible day and a humane one are different things.

One morning against the clock. The grey bars are driving; the solid bars are time actually spent at the door. Two of these stops can only be served inside a shaded window — which is the constraint a shortest-path route will cheerfully violate.

Why the order stops were entered is never the right order

Most route lists are archaeology. A customer was added when they signed, another when a rep picked them up, a third when someone left and their accounts were split across two vans. The sequence records the order things happened, which has no relationship to the order they should be visited in.

That is why resequencing an existing route so often returns more than anyone expects. You are not competing against a good plan. You are competing against an accident.

The same eleven stops, ordered two ways. Same van, same driver, same customers — one of these runs misses two receiving windows.

What the arithmetic actually returns

Take a real-shaped example: eleven stops, twelve minutes of service each, an average road speed of 30 km/h once you apply a road-distance correction to the straight lines. Resequenced, that run goes from 61.4 km to 47.9 km.

Before: 61.4 km ÷ 30 km/h × 60 = 122.8 minutes driving, plus 11 stops × 12 minutes = 132 minutes at the door — 4h 15m.

After: 47.9 km gives 95.8 minutes driving, plus the same 132 minutes at the door — 3h 48m.

13.5 km and 27 minutes, on a day where nothing else changed.

Notice what did not move: the 132 minutes spent actually serving customers. Optimisation cannot compress that, and any vendor implying otherwise is selling you a number you will not see. What it compresses is the 27 minutes of driving between them.

The same saving repeated across a fleet. The eight-van line is an illustration rather than a benchmark — four vans halves it.

Where the money actually lands

Twenty-seven minutes a run is not a rounding error, but it is also not a cheque. It shows up in four places, and only two of them are cash:

Fuel you do not burn, which is the obvious one and usually the smallest.
Paid driver hours — either overtime you stop paying, or hours you stop buying to cover the same work.
Vehicle wear, spread over years and therefore easy to ignore until a replacement cycle arrives early.
One more stop per run, which is the one that compounds. A shorter run has room in it, and room is how a fleet grows without a fleet purchase.

The cheapest truck you will ever add is the half-hour you just freed on the six you already own.

What optimisation will not fix

Worth saying plainly, because it is where most disappointment comes from:

A window you cannot move. If four customers all insist on 07:00, no sequence solves it. That is a commercial conversation, not a routing one.
A territory that needs splitting. If a van is doing 180 km before the first drop, resequencing polishes the wrong thing.
A driver who is right. Someone who has run a territory for fifteen years knows that the loading bay on Fourth Street is unusable before nine. Any tool that will not let them override a sequence will be worked around within a fortnight, and then quietly abandoned.

How to tell whether you have a problem

Four questions, and you can answer all of them from last week's data without buying anything:

Does the same customer sit in a different position on the run depending on who drives it?
Do you know how long each stop actually takes, or is it one number applied to everyone?
How often does a van return with stock that was scheduled to be delivered?
When a route runs long, can you say whether it was drive time, service time, or a window that caused it?

If the fourth question has no answer, the routing is not really the first problem — the measurement is.

Key takeaways

Distance is one of five constraints, and rarely the binding one. Receiving windows and real service time usually decide the day.
Most route orders are an accident of history, which is why resequencing an existing run returns more than people expect.
Optimisation compresses driving, not serving. Time at the door is untouched — treat any promise otherwise with suspicion.
The compounding return is capacity, not fuel: a shorter run has room for another stop.
If you cannot say why a route ran long, fix the measurement before the sequence.

← All articles See how OrderPulse sequences a route →