Data, privacy and regulation
Personal data, least privilege and regulatory reporting.
An insurer knows a great deal about its customers: where they live, when they were born, what they own, their health on holiday, the accidents they caused. The law expects that knowledge to be protected, and the regulator expects the numbers built on it to be right.
Personal data
In the EU, the GDPR governs personal data: anything that identifies a living person, directly (a name) or when combined with other data (a birth date and a postcode). Harbour Mutual's customers table is personal data; so are the adjusters' notes about a claim. Health details, such as a hospital stay abroad on a travel claim, are a special category, with stricter rules.
The principles an analyst meets every day:
- Purpose limitation: data collected to run a policy may be used to run the policy, not for anything that comes to mind.
- Data minimisation: use only the data a task needs. A pricing study needs a customer's age band and region, not their name.
- Storage limitation: keep personal data no longer than needed; old claims files are deleted or anonymised on a schedule.
- Integrity and confidentiality: protect the data against loss and against people who should not see it.
Pseudonymised data, where names are replaced by an id such as customer_id, is still personal data: whoever holds the key can reverse it. Anonymised data, which cannot be linked back to anyone, is not.
Least privilege
Every person and every program should have the access its job needs, and no more. In a database this is done with roles and grants: a reporting tool reads but cannot change; a pricing analyst sees ages, not names; a broker's portal shows the broker's own clients only. Access is granted on purpose, never by default.
| Who | Needs | Should not have |
|---|---|---|
| Month-end reporting | Read every table | Any right to change data |
| Pricing analyst | Age band, region, claims history | Names and birth dates |
| A broker's portal | Their own clients' policies | Other brokers' clients |
| Claims handler | The files of the claims they handle | The whole customer base |
Regulatory reporting
Insurers report to their supervisor (in the Netherlands, De Nederlandsche Bank; in Poland, the KNF; in Germany, BaFin). Under Solvency II they send quarterly and annual templates: premiums, claims and reserves by line of business and country, and the capital that the risks require. The figures must be complete, accurate and reproducible: the insurer must be able to show how each number was derived from its data.
That is where an analyst's SQL matters. A query saved as a view, run the same way every quarter, from data protected by constraints, is an audit trail; a spreadsheet copied by hand is not.
Backups
Confidentiality is one half; availability is the other. Production databases are backed up (for PostgreSQL, with pg_dump for a copy of one database, or continuous archiving to restore to any moment), and restores are tested: a backup never restored is a hope, not a backup.
On Day 10 you will create Harbour Mutual's roles, hide personal data behind a view, and let each broker see only their own clients.
Check your understanding
Answer every question, then check.