GDPR. Article 30 guide.
GDPR Article 30 records versus the real data estate
Test each Article 30 item against the systems behind it, then file every gap with an owner, a decision and a date.
Co-founder and CEO · Updated November 8, 2026

On this page
Reconcile one activity at a time, both ways
Reconcile each processing activity in two directions. Start from the record and check the systems its entry names. Then start from the systems and look for personal data that no activity explains. An activity passes when every named system was checked, every category found is declared and every gap has an owner and a decision.
Applies to the record a controller keeps under Article 30(1) and the record a processor keeps under Article 30(2) of the GDPR. The UK version has its own guide, linked at the end. Roles: the privacy lead keeps the record, system owners confirm their systems, and a database administrator runs the catalog queries.
Time: plan one working session per activity. Start with the activity that holds the most sensitive data, because its gaps matter most.
- List the systems each activity relies on, with an owner
- List the tables and files that hold personal data in them
- Compare categories, recipients, transfers, retention and security with the record
- Search backwards for personal data no activity names
- Log each gap: update the record, delete the copy or justify it
Article 30(1), reconciled for one activity
Records for a customer support activity. Each row shows what the record declares, what the systems show, the gap and the action.
| Article 30(1) item | Declared in the record | Found in the systems | Gap | Action and owner |
|---|---|---|---|---|
| (a) Controller details | Controller, representative and DPO contact details | Not tested against systems | DPO mailbox changed in August 2026 | Update the entry; privacy lead, November 15, 2026 |
| (b) Purposes | Handle customer support requests | crm.customer_export_2025 serves no listed purpose | A copy with no purpose | Delete it, or name the purpose; CRM owner |
| (c) Categories of people and data | Customers; name, email, order history | Also phone, postal address and date of birth in crm.customer | Three categories undeclared | Add them to the entry, or drop the columns |
| (d) Recipients | Helpdesk vendor | An analytics warehouse role readable by marketing | Internal recipient undeclared | Add the recipient category, or revoke the grant |
| (e) Transfers | None | The helpdesk contract lists a US data region | Transfer undeclared | Record the transfer and its safeguard, or change region |
| (f) Erasure time limits | 24 months after last contact | Oldest row in crm.support_ticket is from 2019 | Retention exceeded | Purge, or record the exception; CRM owner |
| (g) Security measures | Role-based access and protected stored data | PUBLIC can read crm.newsletter_signup; emails in plain form | Description doesn’t match the systems | Revoke PUBLIC; decide on protection |
What Article 30 requires of each record
Under Article 30(1), each controller keeps a record of the processing activities under its responsibility. Items (a) to (g) run from contact details to a general description of the Article 32(1) security measures. Items (f) and (g) apply where possible.
A processor’s record under Article 30(2) lists each controller it acts for, the categories of processing, any third-country transfers and a security description. Both records must be in writing, electronic form included, and the organization makes them available to the supervisory authority on request.
Article 30(5) exempts an organization with fewer than 250 employees. The exemption falls away if its processing risks people’s rights and freedoms, isn’t occasional, or includes special category or criminal offense data. Customer support runs all year, so it isn’t occasional processing and the exemption wouldn’t cover it. Recital 82 ties the record to demonstrating compliance.
Five steps from record to systems and back
Steps 2 and 3 carry the catalog queries. Both are read-only and list column names only; no values are read.
1. List the systems each activity relies on
- Owner
- Privacy lead, with each system owner
- Tool or path
- Your record of processing and asset register
- Expected output
- One row per activity and system, with an owner, including copies and exports.
- How to verify
- Every system in the asset register maps to an activity, or goes to step 5 as unexplained.
2. List candidate columns in PostgreSQL
- Owner
- Database administrator
- Tool or path
- psql, as a role that can run catalog queries
- Command
- SQLTested on PostgreSQL 16.15, 17.11 and 18.6, November 9, 2026, as a login with no grants beyond the PUBLIC defaults, against synthetic data.
-- Read-only. pg_catalog lists tables you can't read; information_schema doesn't. SELECT n.nspname AS schema_name, c.relname AS table_name, a.attname AS column_name, format_type(a.atttypid, a.atttypmod) AS data_type, c.reltuples::bigint AS approx_rows FROM pg_attribute AS a JOIN pg_class AS c ON c.oid = a.attrelid JOIN pg_namespace AS n ON n.oid = c.relnamespace WHERE a.attnum > 0 AND NOT a.attisdropped AND c.relkind IN ('r', 'p') AND n.nspname NOT IN ('pg_catalog', 'information_schema') AND n.nspname !~ '^pg_toast' AND a.attname ~* '(email|phone|mobile|birth|dob|address|postcode|first_?name|last_?name|surname|passport)' ORDER BY 1, 2, 3; -- Completeness: the tables this database holds, for your reconciliation count. SELECT count(*) FROM pg_class AS c JOIN pg_namespace AS n ON n.oid = c.relnamespace WHERE c.relkind IN ('r', 'p') AND n.nspname NOT IN ('pg_catalog', 'information_schema'); - Expected output
- One row per column whose name matches, such as crm.customer.email. approx_rows is an estimate, and -1 before the first analyze.
- How to verify
- The count query equals the number of tables the system owner expects. A shortfall means the role sees too little.
3. List candidate columns in SQL Server
- Owner
- Database administrator
- Tool or path
- SSMS or sqlcmd, as a login with VIEW DEFINITION on the database
- Command
- SQLFrom Microsoft Learn documentation, page updated October 19, 2026; not yet tested by Sylox.
-- Read-only. Without VIEW DEFINITION, sys.tables lists only objects you already -- have a permission on, and the query returns a short list with no error. SELECT s.name AS schema_name, t.name AS table_name, c.name AS column_name, ty.name AS data_type FROM sys.columns AS c JOIN sys.tables AS t ON t.object_id = c.object_id JOIN sys.schemas AS s ON s.schema_id = t.schema_id JOIN sys.types AS ty ON ty.user_type_id = c.user_type_id WHERE c.name LIKE '%email%' OR c.name LIKE '%phone%' OR c.name LIKE '%mobile%' OR c.name LIKE '%birth%' OR c.name LIKE '%address%' OR c.name LIKE '%surname%' OR c.name LIKE '%first%name%' OR c.name LIKE '%last%name%' ORDER BY s.name, t.name, c.name; -- Completeness: SELECT COUNT(*) FROM sys.tables; run it as a sysadmin and compare. - Expected output
- One row per column whose name matches, for SQL Server 2022 and later.
- How to verify
- The sysadmin count equals the count your login sees. If it doesn’t, ask for VIEW DEFINITION and run the query again.
4. Reconcile item by item and log the gaps
- Owner
- Privacy lead
- Tool or path
- A reconciliation sheet shaped like the table above
- Expected output
- One row per Article 30(1) item per activity: declared, found, gap, action, owner and date.
- How to verify
- Every gap has an owner, a decision and a date, and the sheet states what was not checked.
5. Search backwards for personal data no activity names
- Owner
- Database administrator, with the privacy lead
- Tool or path
- The query output, shared drives, exports and test databases
- Expected output
- A list of tables, files and copies with personal data and no activity in the record.
- How to verify
- Each item is assigned: add it to the record, delete it, or justify it in writing.
Where the unexplained copies sit
Records drift because copies are made for good reasons and rarely written down. Reporting data lands in a warehouse, exports land in shared folders, and test databases are refreshed from production. Each copy holds personal data under a purpose the record may not name.
Look first in analytics warehouses, shared-folder exports, test and development databases, team spreadsheets and backups kept past their erasure limit. For each one, ask which activity it belongs to and whether the entry describes it.
What this method misses
A column-name search finds columns named like identifiers. It misses a column called col7 and free text that holds personal data, so pair it with a content scan or a sample of values. It reads column names, not the data, so it can’t confirm categories on its own.
Purposes, lawful basis, recipients’ contracts and transfer safeguards are decisions and documents. Systems can contradict them, as rows (d) and (e) show, but they can’t confirm them. SaaS sources need their own export or admin report, and replicas and restored copies need a run of their own. This page is general information, not legal advice.
IRIS gives the system side of this reconciliation, in the sources you connect. Your privacy team keeps the record. Capabilities vary by edition.
The location of each category of personal data: IRIS produces part of this, in the sources you connect.
Owner, purpose and retention recorded for each discovered table or file: IRIS produces part of this, as your stewards record them in the IRIS catalog.
Tables and files a scan couldn’t read, with the reason: IRIS produces part of this, for the tables and files the scan account can list.
The Article 30 record itself: purposes, lawful basis, recipients and transfers: Not produced by IRIS.
Explore discoveryVerify one activity, then read the UK version
Run steps 1 to 5 for the activity with the most sensitive data. The check passes when each named system was queried, the count query matches the owner’s expectation and the gap sheet has no row without an owner.
The UK keeps the same Article 30 record, and its regulator expects it to rest on a data map. ICO accountability and data mapping evidence covers that test.
Official sources
- GDPR Article 30: records of processing activities, EUR-Lex. Opened November 8, 2026.
- GDPR Article 32: security of processing, EUR-Lex. Opened November 8, 2026.
- GDPR Recital 82: record of processing activities, EUR-Lex. Opened November 8, 2026.
- PostgreSQL 18: pg_attribute catalog, PostgreSQL Global Development Group. Opened November 8, 2026.
- SQL Server: sys.columns, Microsoft. Opened November 8, 2026.
- SQL Server: metadata visibility configuration, Microsoft. Opened November 8, 2026.
Start with a free assessment of one database you choose.
Get a free assessmentArticle 30 questions
Yes, unless the Article 30(5) exemption applies. It covers organizations under 250 employees whose processing is occasional, low risk and free of special category or criminal offense data.
Article 30 sets no interval. Our recommendation is once a year for each activity, and again after a new system, a migration or a change of vendor.
No. Article 30(2) asks a processor for each controller it acts for, the categories of processing, transfers and a security description.
No. A scan can show which categories sit where. Purposes, recipients and lawful basis are decisions that your privacy team records.
Protect your brand. Start with your data.
Start with one database. See what personal data it holds, who can reach it and what the scan couldn’t read.


