What the watch event log records
The wearable device event log is a durable record of what the watch did and what it reacted to. It is kept on the device itself and sent to the server. It contains sensor decisions, the course of the alarm cycle and the results of successive heartbeats, which are the regular check-ins the device sends to the system. After an incident, the sequence of events can be reconstructed instead of guessed from individual positions on the map.
The data protection officer and the IT team look at the same record for different reasons. The officer wants to know which categories of data are created and how long they stay on the server. The IT team wants to know whether the record is complete, whether it can be verified and whether it adds load to the network. This page organises the answers to the questions that come up most often in such discussions.
How the log reaches the server and why it creates no duplicates
The watch does not send every entry separately. The log is sent in batches every third heartbeat cycle, which at a 26-second interval means roughly every 78 seconds. A single batch contains at most 89 entries, and the queue of pending entries on the device is limited to the same 89 positions. This limit protects the battery and the connection, and at the same time it stops the queue from growing without end.
Resending is safe. Uniqueness of an entry is enforced by the pair of user and entry identifier, so sending the same batch again after a broken connection does not create duplicates in the database. For an audit this means that the number of entries in the log matches the number of events, not the number of send attempts.
There is also a precautionary rule on the server side. If even one entry in a batch is invalid, the server rejects the whole batch with status 400. The device does not remove from the queue the entries the server has not stored, so nothing disappears silently. For the IT team this is an important feature: a 400 response is a signal to analyse the format, not a data loss.
Time scale of the watch log
Two timestamps in one entry and the offset trap
Each entry stores two times: the device time and the time the server received it. The difference shows how long the watch was silent before it came back into range. If the difference is a few seconds, the connection was working. If it is a dozen minutes, the event happened during the gap and arrived only with the next batch.
Time zones require care. The device time may be stored in UTC, while some tables coming from the watch keep local time. Comparing them without conversion gives a shift of 2 hours, which is easy to mistake for connection delay. Before any analysis, establish which time zone a given column uses, and only then compare. Practical details are described on the page about device and server time in logs.
What the log does not contain and where the window ends
The log is not an archive. The queue on the device holds 89 entries, so a watch that stays out of contact for longer than about 2 hours loses the beginning of its own failure: the oldest entries are pushed out by newer ones. This is a protective window, not a permanent record. Anyone who needs a longer history must rely on what has already reached the server.
It also does not contain the content of conversations or data from outside the device. It records sensor decisions, the alarm cycle and heartbeat results, meaning technical and event data. Location data about the employee is stored separately in the pulse telemetry, which has no default retention: a cleanup mechanism exists but is switched off until the client decides how long data should be kept. That decision belongs to the data controller and should be recorded in the implementation documentation.
In practice, the data protection officer should ask three questions: how long log entries are kept, who has access to them in the panel and for what purpose they are analysed. The system provides the record for audit, but the purpose and the retention period are set by the organisation.
Case from a live system: 26 minutes without a trace
In one of the working systems, a watch vibrated and operated for 26 minutes without sending anything to the server. The lack of a durable record made it impossible to reconstruct what happened on the device during that time. This case was the direct reason why the event log was introduced as a permanent feature rather than a temporary view.
Conclusion for the IT team: the position on the map alone is not enough to explain an incident. A record of device events is needed, sent in batches and resistant to repeats. Conclusion for the officer: since such a record is created, its retention and scope of access must be agreed before the first dispute about who could view it and when.
Before the log appears: a condition on the host side
Recording the log works only on hosts where the backend with the appropriate endpoint has been deployed. A host without it returns 404, and the watch remains blind for that host: the device sends batches, but the server does not accept them. That is why, after each update, it is worth checking that all environments respond correctly, as described in the checklist for version consistency of hosts after deployment.
It is also worth remembering that the application version on the watch may be recorded in several places and can drift apart. The version from the package manifest reported by the device management system is treated as the source of truth. If the log from a particular watch looks different from the rest of the fleet, the first step is to compare versions, and only the second step is to analyse the entries.
Video
Frequently asked questions
Does the log contain the content of conversations?
No. It records sensor decisions, the course of the alarm cycle and heartbeat results. These are technical and event data of the device.
What happens when a batch is sent twice?
Nothing bad. The pair of user and identifier enforces uniqueness, so duplicates are not created.
After how long offline does the log lose the beginning of events?
After about 2 hours, because the queue holds 89 entries and the oldest are pushed out by newer ones.
Who decides how long telemetry is kept?
The client, as the data controller. A cleanup mechanism exists but stays off until the client decides.
Let us walk you through event recording at your plant
We will show how the watch log looks in the panel and how to set retention in line with your data protection policy.
Let's talk