Why an audit log matters under GDPR
The audit log implements the principle of accountability: the controller must be able to show that it processes data in line with the rules. Article 32 GDPR requires technical and organisational measures adequate to the risk, and a change log is one of the simplest measures that can be shown to an inspector.
In the InnerWeb system the audit trail records changes to selected data together with the author and the time: permits, the taxonomy of dangerous work, training results, automation rules and inspection routes. The list of audited objects is explicit, so the inspector knows what the log covers and what it does not.
Practice 1: audit what a person changes
The selection rule is simple: audited are the data that a person can change. Data written only by the system, such as the heartbeat, location history, notification logs and webhook queues, is outside the trail, because its history is source data. Auditing it would duplicate data without evidential value.
This rule also has consequences for the risk assessment. The audit log answers the question 'who changed it', not 'where the employee was'. The question about position concerns the location history, whose access is set separately, by role and purpose defined in the internal act.
Example from a coverage review
In a coverage review of the trail, 127 entities were compared with 60 audited, and 48 more were brought under audit. The largest of the added tables had 5,825 rows, so the cost of auditing proved negligible. The conclusion for the controller: there is no need to save on auditing small configuration tables.
Share of the time-in-location aggregation table in audit table size
Practice 2: do not audit bulk data without documenting human action
The best example of a trap is the table aggregating time spent in a location. It took 6,331 MB, which is 95% of all audit tables, and it did not document any human action. After excluding it from the audit, the size of the trail fell, and the source heartbeat data remained.
For the inspector the conclusion is twofold. First, the size of the log is also a risk: an oversized log is expensive and slow, so it becomes hard to review. Second, excluding a table does not weaken accountability if the source data remains and changes made by people are still recorded.
Practice 3: exclude sensitive fields from the trail columns
The change log must not become a source of leaks itself. Sensitive fields, such as the password, the confirmation token and the device push token, are excluded from the audit trail columns, so the change log does not store secrets. For own extensions the inspector checks whether new fields of this kind have been added to the exclusion list.
This is consistent with the minimisation principle in Article 5(1)(c) GDPR: the log keeps the fact and time of a change and its author, not data that could be used to take over an account.
Practice 4: set the time zone and check historical entries
Time evidence has value only when it is known which time zone it was recorded in. The audit trail records time in the Europe/Warsaw zone from October 2026. Entries from previous winter seasons are shifted by 1 hour, which must be taken into account in time evidence.
A practical rule: with every extract from the log submitted for an inspection, add a note on the time zone and on any shift of historical entries. A one-hour difference without explanation can be read as an error or manipulation, and an explanation closes the matter.
Practice 5: watch panel performance with large tables
The audit panel pages through many records using OFFSET. For a table with 525 thousand rows this means about 318 thousand rows scanned per page. For large tables, paging by key or limiting the depth is recommended.
For the inspector this means a habit: filter the log by object, person and date rather than reviewing it from the start. An extract for an inspection should have a narrow scope, because this speeds up the work and limits the disclosure of other people's data.
Checklist
Video
Frequently asked questions
Does the audit log show where the employee was?
No. The log records data changes made by people, with the author and the time. Location history is source data from a different mechanism, and access to it is set by role and purpose.
Does the log store passwords?
No. Sensitive fields, such as the password, the confirmation token and the device push token, are excluded from the audit trail columns.
Where does the one-hour shift in older entries come from?
From October 2026 time is recorded in the Europe/Warsaw zone, and entries from previous winter seasons are shifted by 1 hour. Add this information to the extract.
Is recording user sessions used for supervision?
No. Recording clicks and queries (Record and Replay) serves regression testing, is switched on in the settings, and the list of recordings is visible only to the administrator.
Sources
Let's show the audit log on your data
We will go through the list of audited objects together and set the access roles for the log.
Let's talk