← All guides

Surveys and eNPS · 9 min read

Employee survey privacy — segmentation, GDPR and question boundaries

Segmentation fields, small groups, consent and sensitive questions: GDPR principles, national employment rules and safe survey reporting.

Convomint editorial team Updated
Employee survey privacy — segmentation, GDPR and question boundaries

Segmentation fields are background details used to group responses, such as department, tenure band or location. They help investigate different experiences but can also identify respondents. Before adding a field, decide what action it could inform and whether less detail would be enough.

Sources checked on 24 October 2026. This guide distinguishes legal requirements, practical safeguards and quality standards. It supports survey preparation; a lawyer or designated DPO should assess the legal basis and applicable employment rules for a particular deployment.

Anonymity, confidentiality and GDPR

Term Meaning
Anonymity People cannot be identified taking account of means reasonably likely to be used
Confidentiality Access is restricted, although the data may still identify people
Pseudonymisation A code replaces an identifier, but a link to a person may remain
Aggregation Responses are combined; this does not determine the status of source data

GDPR Recital 26 and Article 4 distinguish personal from genuinely anonymous data. Anonymous data falls outside GDPR, but collecting personal data and subsequently anonymising it is still regulated processing. Invitation lists, tokens, IP addresses, timestamps and logs may separately be personal data. A report that is anonymous to a manager does not establish an anonymous process at the supplier.

Do not collect extra identifiers solely to retrieve an otherwise anonymous answer later. Explain that genuinely anonymised responses may no longer be individually retrievable. Relevant GDPR rights apply while data remains identifiable, taking Article 11 into account.

What may you ask?

There is no universal statutory list of questions permitted in every employee survey. Where personal data is processed, define a purpose, necessary scope and appropriate legal basis.

Question type Practical boundary
Task clarity, tools, support and work organisation Usually relevant to an experience survey; still assess the legal basis and scope
Department, location and tenure Use only for planned analyses; prefer broad categories over exact dates or unique job titles
Age, sex, nationality and pay Not inherently the Article 9 categories, but potentially identifying and relevant to discrimination protections; do not collect by default
Examples: health, diagnoses, disability, religion, political views, union membership, ethnicity and sexual orientation Special-category data: an Article 9(2) exception is needed as well as an Article 6 basis and compliance with employment law
Criminal convictions and offences A separate Article 10 regime; ordinary employee consent is not sufficient authorisation
National identifiers, home addresses, family details and plans to have children Exclude from an ordinary experience pulse; a separate HR obligation does not automatically justify survey reuse

Article 9 also covers genetic data, biometric data used for unique identification, philosophical beliefs and sex life. Nationality is not the same as ethnicity. Ordinary data or combinations of fields can nevertheless reveal protected information. Workload questions differ from questions about treatment for depression. Content, context and resulting inferences matter more than a “wellbeing” label. GDPR Articles 5, 6, 9 and 10.

EDPB Guidelines 05/2020, paragraphs 21–23, emphasise the imbalance in employment. Consent can be valid in exceptional circumstances where choice is genuinely free and refusal has no adverse consequences. Where consent is the basis, withdrawal must be as easy as giving it and does not invalidate earlier lawful processing (Article 7(3)). An optional field does not by itself establish lawful processing. Voluntary participation and consent as a processing basis are separate questions.

Where legitimate interests are appropriate, document purpose, necessity and balancing of rights. Article 6(1)(f) does not override special-category restrictions or national employment rules; public authorities cannot use it for processing in performance of their tasks. A generic inclusion or wellbeing objective does not create an Article 9 exception.

Poland-specific example: Labour Code Articles 22¹–22¹b govern employer collection of employee data. Article 22¹a permits consent for additional ordinary data subject to safeguards against adverse consequences for refusal or withdrawal. For special-category data, Article 22¹b(1) permits reliance on consent only where disclosure is initiated by the employee or applicant. A consent box under an employer’s question about a diagnosis, religion or union membership does not demonstrate that condition. A different route requires an applicable legal exception and compliance with employment rules.

These Polish conditions are not a universal rule for every country. GDPR Article 88 allows specific national employment provisions. Check local restrictions and any required involvement of employee representatives. Do not reuse survey groups for discriminatory decisions or retaliate against respondents. A global platform or EU hosting does not remove local responsibilities.

Design safe segmentation and reporting

  1. Plan useful breakdowns before collection. Remove fields with no clear decision use.
  2. Prefer broad bands. Do not automatically combine department, age, tenure, location and role; assess each combination and what other reports reveal.
  3. Distinguish headcount, respondents and valid answers to each question. A large department may supply only two answers to one item.
  4. Combine sensible categories or withhold results when disclosure risk is too high. Hiding one cell is insufficient if totals and other cells reveal it.
  5. Check exports, comments, filter changes and repeated waves. One new team member may become identifiable by comparing reports.

Five or ten responses is an organisation’s reporting rule, not a GDPR minimum. No single number guarantees anonymity or reliable measurement. Higher-risk results may need withholding even above a threshold. These recommendations do not claim that the platform automatically blocks every unsafe combination.

Example: a department has 12 respondents, including 11 specialists and one manager. Publishing the department total and specialist result can reveal the manager’s answer despite hiding their row. WP29 Opinion 05/2014 discusses identification risks. See calculating and interpreting group results.

Before collection and after responses arrive

  • Roles and suppliers: establish the actual controller and processor roles, access and an Article 28 agreement where applicable. Buying software does not transfer controller responsibility.
  • Participant information: explain purpose, basis, recipients, retention period or criteria, rights and contact details. Consider Article 13 for direct collection and Article 14 when adding information from HR records.
  • Retention: distinguish invitations, responses, comments, exports and anonymous aggregates. There is no universal statutory retention period for every survey. Delete or effectively anonymise data when its justified purpose ends, accounting for applicable legal obligations.
  • Comments and AI: minimise personal details, restrict access and check supplier purposes, including reuse for their own models. Assess any transfers outside the EEA; EU hosting alone does not settle Chapter V requirements.
  • Risk: assess whether a data protection impact assessment is needed and conduct it before processing if high risk is likely (Article 35). Not every simple survey requires a DPIA.
  • Decisions: do not turn a pulse into an automatic dismissal or pay ranking. Solely automated decisions producing legal or similarly significant effects are subject to Article 22 restrictions.

Explain rights appropriate to the legal basis, including objection where legitimate interests apply. Erasure is not an unconditional right in every process.

If a comment contains health information, an allegation about a named person or an urgent issue, restrict circulation and follow the appropriate procedure. Do not publish it in the team report. An ordinary survey is not a substitute for a reporting channel; tell employees where to raise matters requiring a response.

Laws, standards and benchmarks differ

GDPR is EU law; national employment provisions add context. Technical standards are generally voluntary, although a contract can create specific obligations. A reference to a standard is not certification of a platform or validation of a questionnaire.

Document Relevant scope
ISO 20252:2026 Service requirements for market, opinion and social research; replaces the 2019 edition
ISO 10667-1:2020 and Part 2 Work-related assessment services: client and provider responsibilities respectively
ISO 30414:2025 Human capital reporting, not a ready-made eNPS methodology
AAPOR Survey methodology and transparency guidance, not employment legislation

These descriptions use the standards’ public scope statements, not a compliance audit against their full texts. A benchmark is different again: comparative data requiring a documented population, period, instrument and calculation.

Prepare your survey

Use the question and segmentation template, then set group-reporting rules before launch. Create an account and review the available settings. Match participant information to the actual configuration and established legal basis.

Marking a field mandatory sets the conditions for submitting the form; it does not by itself create a legal duty to participate. Distinguish insufficient observation, an omitted optional question and refusal to participate. Do not record refusal as a negative rating or automatically as insufficient observation.

Sources and status

The EDPB lists Guidelines 02/2026 on anonymisation as under public consultation. We do not present that draft as binding law or final guidance.

Convomint

Prepare a survey

Prepare a survey or review, collect perspectives and organise next steps in Convomint.

Try it for 14 days → Working with clients? Explore the partner programme. →