Alerts That Get Acted On
A telematics system can generate hundreds of notifications a day. The design that keeps them useful is mostly subtraction.
Running it · Procedure
Every deployment starts with alerts enabled and ends with them ignored. The failure is in the design rather than in the people.
Why they stop working
Volume. Forty a day is not a set of alerts; it is a feed.
No addressee. An alert to a shared inbox belongs to nobody.
No action attached. A notification that does not say what to do produces a decision each time, which is expensive.
And no consequence for ignoring one, so ignoring becomes the norm within a month.
The three categories
Act now: a safety-relevant fault, a vehicle immobilised, a temperature excursion in a refrigerated load.
Schedule: a service threshold reached, a recurring code, a consumable low.
Review later: everything else, delivered as a weekly summary rather than individually.
Most fleets have everything in the first category and then discover nothing is.
Designing each one
Who receives it, by name or by role that is actually staffed.
What they should do, stated in the alert.
What happens if they do not, which for the first category means an escalation.
And how it is closed, so an open alert is visible.
An alert without all four is a notification, not a control.
Thresholds
Set from your own data rather than from defaults.
A default deceleration threshold tuned for a car fires constantly on a laden van.
A default idle threshold flags every refrigerated vehicle.
Run for a month with alerts logged but not sent, then set the thresholds where the volume is manageable, then switch them on.
Reviewing them
Quarterly: which alerts fired, how many were acted on, how many were false.
An alert with a high false rate gets retuned or removed, not tolerated.
And an alert nobody acted on in three months should be turned off, which is the hardest and most useful decision in this area.
The measure
Alerts per vehicle per week, which should be small.
Proportion acted on, which should be high.
If the first is large and the second is low, the system is producing noise and the fix is subtraction rather than better filtering.
Log silently for a month first
The sequence that prevents alert fatigue.
Enable alerts into a log, not to people.
Count the volume by type.
Set thresholds from your own distribution rather than from defaults tuned for a different vehicle.
Then switch them on. A month of patience buys a system people still read in a year.
Reproduce the workflow
For another way to make this requirement testable, consult compare this planning case. Reproduce the case with real roles, codes, failures and recovery steps.