State of Lone Worker Safety · FY26

A year of duress alarms, measured.

20,433 real activations. 815 confirmed emergencies - one every 10.7 hours. Every event logged by the Duress monitoring centre from 1 July 2025 – 30 June 2026, published because nobody else in this industry shows their numbers.

Published 2026-07-22 · Free to cite with attribution · Methodology at the end

7sMedian operator pickup
94.2%Answered within 30 seconds
815Confirmed emergencies
1m 43sMedian case resolution
Response

When the alarm goes, a human answers in seconds.

Across every activation - real, false, day, night - the median time from alarm to a monitoring operator on the case was 7 seconds. 71.6% were answered inside 10 seconds; 99% inside 52.

0–5s
13.5%
5–10s
52.8%
10–15s
16.7%
15–20s
5.8%
20–30s
5.2%
30–60s
5.9%

Share of activations by time to operator pickup, FY26. Failover-redundancy drills excluded.

The finding that matters

Risk concentrates exactly when nobody is watching.

22%of activations happen after hours (6pm–6am)
34%of confirmed emergencies happen after hours
17%of activations happen on weekends
25%of confirmed emergencies happen on weekends

An alarm raised after hours is roughly twice as likely to be a confirmed emergency as one raised during the working day. The hours when a supervisor-based check-in system is weakest - evenings, nights, weekends - are the hours when the alarm is most likely to be real.

Triggers

1 in 6 alarms was raised by the system, not the worker.

17.7% of activations were automatic - fall detection, missed check-ins and expired safety timers firing when the worker couldn't, or didn't, press anything.

Device button press
48.7%
App (slide-to-alert, triple tap)
29.7%
Fall detectionauto
9%
Missed check-inauto
5.8%
Safety timer expiredauto
2.9%
Integrations + other
2.4%
Phoenix button
1.2%
The honest number

93% of alarms weren't emergencies. We answered every one anyway.

Of 20,433 activations, 93% resolved as false alarms or could not be verified. That number is the price of the 815 that were real: a monitoring model that screens harder, calls back later, or waits for confirmation is a model that eventually misses one. Median time to a human, false alarm or not: 7 seconds. No other provider in this category publishes this figure.

Patterns

When alarms actually happen

By hour of day

0
6
12
18

Activations peak mid-morning and early afternoon - the shape of community visits, store floors and field work. The after-hours tail is smaller but carries double the emergency rate.

By month

Jul
Aug
Sep
Oct
Nov
Dec
Jan
Feb
Mar
Apr
May
Jun

Volume grew through the year - June 2026 ran 10% above July 2025 - with the expected summer-holiday dip across December and January.

Method

How these numbers were made

  • Source: the complete emergency-event log of the Duress monitoring centre, 1 July 2025 – 30 June 2026. Every activation, no sampling.
  • Pickup time is measured from alarm raised to a monitoring operator opening the case.
  • Test activations are excluded from activation counts. Internal failover-redundancy drills (identifiable by design as pickups over 60 seconds) are excluded from all figures.
  • Customer identities are anonymised; figures are aggregates across all monitored organisations in Australia, New Zealand and the UK.
  • Figures are point-in-time as at publication and will be refreshed in the FY27 edition.
Questions

About this report

Where does this data come from?
Every number is drawn from the Duress monitoring centre's complete emergency-event log for 1 July 2025 – 30 June 2026: 20,433 real activations. Test activations and internal failover drills are excluded. No sampling, no estimates.
Why publish the false-alarm rate?
Because it is the honest cost of never missing a real emergency. 93% of activations turn out to be false alarms or unverified - and every single one was still answered by a human in a median of 7 seconds. A system that filters harder misses the 815 that were real.
What counts as a confirmed emergency?
An activation where the operator verified a genuine emergency and escalated - dispatching emergency services, executing the escalation plan, or staying live with the worker until resolution. De-escalated live situations (512 in FY26) are counted separately.
Can I cite these figures?
Yes, with attribution to "Duress, State of Lone Worker Safety FY26" and a link to this page. For media enquiries or the underlying methodology, contact trav@duress.com.

This is what answering in 7 seconds looks like, at scale.