Geofences

A geofence turns continuous tracking into discrete events: this vehicle entered this site at this time and left at that one. For most operational purposes that is the whole answer.

Why it is the right default

It answers the common questions directly: did the engineer attend, when did they arrive, how long was the visit.

It collects far less. No route, no positions between sites, no record of the journey.

It is easier to justify, because the data maps exactly onto the stated purpose.

And it is easier to explain to the workforce, which matters more than it sounds.

Sizing them

Larger than the building. A fence drawn tightly around a property will miss arrivals when accuracy degrades, and produce disputes about attendance that did happen.

Sized for the environment: tighter in open country, considerably wider in a city.

With a dwell requirement, so that driving past a site does not register as a visit.

Tested with real journeys before it drives anything that matters, because the failure mode is silent.

The events worth recording

Entry and exit, with times.

Which fence, so the site is identified.

Nothing between fences, which is the point.

Optionally a daily distance total, which covers mileage without a route.

Where geofences go wrong

Too many. A fence around every customer address in a large book turns into a system nobody maintains, and stale fences produce wrong records.

Fences around places that are not customer sites — a depot, a café, a competitor's premises — which changes the purpose from attendance to observation.

Fences around workers' homes, which happens by accident when a home address is in the customer list and is difficult to explain afterwards.

Review the fence list annually and remove what is not a current operational site.

The boundary question

A geofence answers "were they there". It does not answer "what were they doing", and systems that add dwell-time scoring to fence events have crossed from attendance into performance measurement.

Dwell time varies with the job, the traffic, the customer, and whether the first attempt failed. Treating it as a productivity figure produces exactly the pressure that makes people rush work.

Use fence events for attendance and billing. If you want to understand why jobs take different lengths, ask the people doing them.

What to check in a product

Can fences be defined without a continuous position feed, so the device reports only events?

Is dwell configurable, to avoid drive-past triggers?

Can the raw positions be discarded once the event is derived, on the device or on arrival?

A product that cannot do the last one is a continuous tracker with a fence feature, which is a different proposition from what you were buying.

Test each fence with a real visit

Before relying on any of them.

Approach from the normal direction, park where the driver parks, leave the way they leave.

Check the event fired, once, at the right time.

Check nothing fired when passing on the adjacent road.

Sites with two entrances, long driveways or underground parking need individual attention, and those are exactly the ones that generate disputed visits later.

A concrete product reference

When translating this principle into a buying test, the healthcare workflow provides a concrete workflow reference. Verify the current behaviour in a trial and judge it against the purpose and limits described above.

Independent reference

For an external point of reference, see Google Maps Platform. Its mapping documentation and services provide a useful reference for understanding geospatial implementation choices.