Framework · Measurement · 3 min read
GDPR-safe tracking plan
A table that links every analytics event to the question it answers, with properties that contain no personal data, a retention limit and an owner, so you can measure the system and never individuals.
- 1Event
One user action in the tool, named clearly.
booking_completed - 2Question
The design question this event answers. No question, no event.
"What is the completion rate?" - 3Properties
Only the context you need. No names, IDs or free text.
app, screen, duration_bucket - 4Retention
How long raw events are kept before deletion or aggregation.
Raw events 13 months - 5Owner
Who answers for the event, its purpose and its access.
UX team
Review the plan with the DPO and works council before anything is tracked.
Use it when
- A product team wants to "add analytics" to an internal tool and nobody has defined what they want to learn.
- You need approval from the data protection officer (DPO), legal or the works council before you can track usage.
- Stakeholders worry that usage data will be used to rate people's performance.
How to use it
- Start from the questions. Write the 3–5 design questions you need answered, such as "Where do bookings begin?" or "Which fields cause errors?". Use a Goals-Signals-Metrics chain to get there. Output: a short list of questions tied to a goal.
- Map each question to the fewest events. For each question, name the event or events that answer it. Drop any event that answers no question (purpose limitation and data minimisation). Output: an event list where every row has a question.
- Strip personal data from properties. Keep only context such as app, screen, entry point, field ID or a duration bucket. No user IDs where they are not needed, no names, no free text, IP anonymisation on, and report only aggregates. Output: properties with no PII.
- Set retention and an owner. Give every event a retention limit (for example raw events 6–13 months, aggregates longer) and a named owner. Choose self-hosted or EU-hosted analytics, such as Matomo, where you can. Output: the completed tracking plan table.
- Review it early with the DPO, legal and the works council. Bring the plan: events, purpose, retention, access. Lead with purpose limitation and with what you will never collect. In some countries, such as Germany, a formal agreement with the works council is usually needed first. Output: approved plan, or a list of changes.
- Use alternatives where tracking is not allowed. Moderated usability tests, diary studies, Contextual inquiry, support-ticket analysis, surveys and server-side logs aggregated by IT. Output: a research plan that answers the same questions.
Worked example
An internal operations platform wants to know why shipment bookings are slow. The UX lead drafts this plan and takes it to the DPO and works council.
| Event | Question it answers | Properties (no PII) | Retention | Owner |
|---|---|---|---|---|
| booking_started | Where do bookings begin? | app, screen, entry_point | 13 months | UX |
| booking_completed | Completion rate and time | duration_bucket, steps_count | 13 months | UX |
| validation_error_shown | Which fields cause errors? | field_id, error_type | 13 months | UX |
The works council asks whether duration_bucket could show who is slow. The lead answers that there is no user ID, so the data cannot be linked to a person, and reports are aggregated per screen.
Watch out for
- Tracking everything "just in case". It breaks data minimisation and answers nothing. Every event needs a question.
- Hidden identifiers. A user ID, an email in a URL or a free-text field turns system data into people data. Check every property.
- Involving the works council last. If they hear about tracking after it is built, you lose trust and often the data. Bring the plan before you build.
- Reporting on individuals. Even once, it confirms the fear of surveillance. Report per app, screen or task, never per person.
Template
Copy this into your wiki and fill one row per event before you talk to the DPO.
Tracking plan: [app / flow] Goal:
Purpose statement: We measure the tool's usability, not people's performance.
Hosting: self-hosted / EU-hosted IP anonymisation: on
We will never collect: names, user IDs, free text, ...
| Event | Question it answers | Properties (no PII) | Retention | Owner |
|-------|---------------------|---------------------|-----------|-------|
| | | | | |
Access (who can see the data):
Reviewed with DPO on: Works council on: Legal on: