Where each machine is, without asking over the radio.

In a plant with twelve forklifts, knowing where each one is gets sorted over the radio. It works, and it costs a conversation every time.

What the radio does not sort out is the rest: how long they stand idle, how many kilometres they do empty and which areas they pass through that they should not.

Fitting the fleet with a tracking system is a project. The cameras are already there.

You mark the areas and the crossing points. From then on you know which machine goes where, how many times and how long it stays.

Two things come out of that: the operational one — where there is a free one right now — and the underlying one, which is the one that saves money: how many trips are made empty and which routes repeat.

And the areas a forklift should not enter stop depending on the driver remembering.

In a manufacturing plant, 38% of the forklift trips measured over a week were empty. Half of those went from point A to point B in the morning and from B to A in the afternoon, always at the same time.

It was a consequence of where the intermediate store had been put. Moving it twenty metres took about eighty empty trips a day out of the day.

The reason almost nobody measures this is that the result is taken as read: 'the forklifts are busy all day'. And it is usually true — they are busy. The question is what with.

When it turns out that a third of the trips are empty, the reaction is not usually to buy fewer machines: it is to move an intermediate store, change a replenishment time or stop going down for one thing at a time.

And those are decisions that cost no money, which is what makes them interesting.

A large plant has two different sorts of traffic and usually manages only one. Inside, forklifts and machinery; outside, lorries coming in, waiting, weighing and leaving.

The outside one is what creates queues on the public road and calls from the council. By counting lorries at the entrance and measuring the queue you can regulate arrivals by time slot before the bottleneck exists.

And the inside one is what decides how long it takes to move a tonne from one place to another, which is a cost almost nobody has measured.

At internal junctions, at entrances and on one-way ramps, the usual answer is a traffic light on fixed timings. It works with average traffic and fails with the real thing.

If the light knows what is waiting on each side, it can give way to whoever needs it. And it can give priority by type: the tanker first, the forklift after.

That is not a traffic engineering project: it is the camera already looking at the junction, connected to the traffic light already fitted.

With fleets going electric a new problem appears: there are few charging points and they stay occupied by vehicles that have already finished charging.

By knowing which bay is occupied, by which vehicle and since when, you can raise an alert when a point has been sitting there not charging. It is the difference between having six chargers and having six chargers available.

And the same goes for the lorry park: free bays, length of stay and occupancy by time slot.

With a virtual number per vehicle, the trace stops being '40 forklifts went past' and becomes the route of each one across the shift.

That is where the patterns nobody sees from inside show up: two machines doing the same trip in opposite directions, a route that passes the same point three times, a badly placed intermediate store.

Almost always the conclusion is not to buy another forklift: it is to move something twenty metres.

The virtual number lives inside the covered area. If a forklift leaves the cameras' reach and comes back half an hour later, as far as the system is concerned it may be a different one: linking the two stretches needs a camera on the way.

And number plate reading has its conditions — angle, distance, light — and its data protection rules, which are not the same as counting vehicles. It is a separate installation and it is described that way from the start.