The technical half.
What the app measures, where each step runs, what leaves the smartphone and what never does. If you only want to know what the app does, the features page is the shorter read.
A phone in your hand moves nothing like a phone in your pocket.
You get rosters, clock-on and worked time like any other workforce platform. What you also get is a measurement nobody else in this category takes, and it is possible because a pocketed phone is strapped to a swinging leg while a held phone is being deliberately steadied by an arm.
The app samples the accelerometer and the gyroscope 5 times a second and sorts each window into pocketed, held, off or unknown. All of that happens on the smartphone. No camera, no microphone, and no app or screen content is read at any point.
| Signal | Pocketed | Held | Ratio |
|---|---|---|---|
| Accelerometer swing, m/s² | -10.5 to +11.2 | -2.6 to +3.1 | about 4 to 1 |
| Gyroscope swing, rad/s | -4.6 to +3.2 | -0.7 to +0.9 | about 5 to 1 |
| Dominant frequency | once per stride | twice per stride | about 1 to 2 |
| Pattern | periodic, stride-locked | irregular, non-gait | n/a |
Between the sensor and the site manager.
5 steps. The important thing about them is where each one runs: the first 3 happen entirely on the worker's own smartphone, and the fourth sends only derived sessions rather than the signal they came from.
Read
The accelerometer and the gyroscope are sampled 5 times a second, on the device, for as long as the shift is running.
Classify
Each window is sorted into pocketed, held, off or unknown by a classifier that runs on the smartphone and never sends the raw signal off it.
Buffer
With no reception the results are held on the phone, so a dead spot in a cutting, a shed or the back paddock costs nothing and never shows up as missing time.
Sync
When signal returns the phone uploads derived sessions only: start, end, state and duration. The underlying motion trace is never transmitted.
Surface
The site manager opens one screen and sees the day: who is on site, who is where, how the shift actually ran, and what needs a decision.
Steps 1 to 3 run on the worker's smartphone. Step 4 is the only moment anything leaves it, and what leaves is a list of state changes rather than a signal anyone could reconstruct a person's day from.
The app, the dashboard and the database.
Configuration travels down to the smartphone roughly every 10 minutes, so a workplace can retune thresholds for a vibrating excavator cab or a quiet nursery without shipping a new version of the app. Derived sessions travel up, and only up.
On the phone
The app
Worker, iOS and Android
Sensors at 5 Hz. Accelerometer, gyroscope, screen power state.
On-device classifier. Pocketed, held, off or unknown.
Shift gate. Runs only between clock-in and clock-out.
Local buffer. Holds sessions through a blackspot, so no signal never reads as off.
In between
The backend
Australian-hosted, encrypted in transit
Threshold vector. Per-workplace sensitivity, synced to devices.
Session ingest. Accepts only hardware-verified devices.
Rostering. Shifts, changes, timesheets.
Live push. Alerts to the dashboard as they happen.
In the browser
The dashboard
Site manager and administrator
Today. Who is on site, and where each crew is.
The day. Hours, breaks and what changed from the plan.
Weather and alerts. Heat, rain, wind and storm risk.
Reports. Timesheets, incidents, consent log.
Underneath
The database
Managed cluster, Australian region
Derived sessions. Start, end, state and duration. No raw sensor stream.
Versioned consent log. Timestamped and exportable.
No media of any kind. No camera, no microphone, no app or screen content.
Simplified from PocketMode's provisional patent figure 1 and the internal system architecture document, version 1.0, 17 August 2026. Integrity checking, push messaging and error reporting are omitted here for clarity.
The crew and the day, not a ledger on each person.
This is the line the product is built on, and it is worth stating plainly. The dashboard answers questions about the work: who is on site, where the crews are, what the conditions were, how long the day ran, what changed from the plan and what was reported. Alongside that it gives a phone distraction report across the working day.
What the dashboard shows
- Who is clocked on, at which site, on which task
- Hours worked, breaks and the timesheet for approval
- Consecutive days and cumulative hours, so fatigue is visible before it is a problem
- Conditions recorded against the shift, and alerts when they turn
- Changes to the published roster, with who made them and when
- Incidents and near misses, including the ones reported anonymously
- A phone distraction report across the working day
What it does not show
- No location history outside the shift, and none between clock-off and clock-on
- No camera, microphone, screen content or app usage, ever
- No raw motion trace. It never leaves the smartphone
- No transcript, and no minute-by-minute account of what anybody did
- Nothing collected on a break, and nothing collected before or after the shift
- Nothing that could be used to discipline someone without an independent reason to look
The classifier exists to make the record honest, not to audit a person. Knowing a phone was pocketed rather than held is how the app can tell worked time from a gap without asking anyone to fill in a form. It is a property of the record rather than a verdict on anybody, and the employer view is built so it stays that way.
If a record could be used to discipline someone without an independent reason to look, it is the wrong record. That test is what the privacy page sets out in full.
Every control on a site is a rule nobody measures.
Worksites and farms alike have had phone rules for the better part of 2 decades, and not one of those rules was written on the back of a measurement of how much phone use was actually happening, or whether the rule changed anything.
A traffic management plan establishes where the plant goes. An induction establishes what somebody was told. A gate reader establishes who walked in. None of them establish whether anyone was looking up when the tractor reversed.
Verification asks about work as done.
The hazards that injure outdoor workers are mostly created on the day the roster is written rather than on the day the injury happens, which is why a rostering system is a safety instrument whether or not anybody calls it one.
| Hazard | The rostering decision that creates it |
|---|---|
| Heat and ultraviolet exposure | What time the shift starts, and how long it runs |
| Fatigue | How many days in a row, and how compressed the roster is |
| Lone working | Who is allocated to which crew, and who is on their own |
| Psychosocial harm | How predictable the roster is, and how much notice people get |
| Struck by plant or vehicle | Who is on foot, at what time, and alongside what plant |
Drawn from the hazard mapping in the employer edition. This is an argument from the evidence, not a measurement.