Data minimisation is usually written as a principle and implemented as an afterthought. A form asks for a date of birth because the template had a field. A campaign page stores a phone number in three systems. An old export sits in a shared folder because deleting it requires someone to decide.
The principle is simple: collect what the purpose requires, keep it only as long as the purpose requires, and let only the people who need it reach it. The work is in the specifics. This checklist turns the principle into controls a web team can own, evidence a reviewer can inspect, and a test that shows whether deletion actually happens.
Start with a field inventory
Before changing any form, list every personal field the application handles. Include fields collected by embedded tools, support widgets and analytics, not only the main signup form.
| Field | Why it is collected | Required or optional | Where it is stored | Retention |
|---|---|---|---|---|
| Example: mobile number | Order updates and delivery coordination | Required for delivery | Application database, SMS provider | Until order closure plus the agreed support period |
| Example: date of birth | Age eligibility for a regulated product | Required only for that product | Application database | Remove when eligibility is verified |
If a field has no stated purpose, remove it. If the purpose is "might be useful later", that is a reason to defer collection, not a reason to store it.
The working checklist
| Area | Control | Owner | Evidence |
|---|---|---|---|
| Collection | Every field has a documented purpose and a required justification | Product owner | Field inventory with reviewer sign-off |
| Collection | Optional fields are genuinely optional and do not block checkout | UX lead | Form test and screenshot |
| Collection | No sensitive field is collected unless the purpose requires it | Product owner | Data protection review note |
| Forms | Privacy notice names the purpose, retention and sharing for each field group | Legal or privacy lead | Published notice and version date |
| Storage | Personal data is not copied into analytics, logs or test fixtures | Engineering lead | Code review and log sample |
| Storage | Backups follow the same retention policy as live data | Infrastructure lead | Backup policy and restore test |
| Use | Each system has a documented role and a data-sharing agreement | Vendor manager | Contract register |
| Use | Access is role-based, with a current list of who can export | Application owner | Access review record |
| Sharing | Third-party scripts are limited to those with a named purpose | Marketing lead | Script inventory |
| Sharing | Exports are approved, logged and given an expiry date | Data owner | Export log |
| Retention | Each data class has a retention period and a deletion method | Records owner | Retention schedule |
| Retention | Dormant accounts are reviewed on a fixed cycle | Operations lead | Dormancy report |
| Deletion | A verified request removes live data, backups and derived copies within the stated window | Privacy lead | Deletion record and test result |
| Access | Administrative accounts use unique identities and strong authentication | Security lead | Account inventory |
| Logging | Logs exclude personal data unless the purpose requires it | Engineering lead | Log schema and sample |
| Change | A new field cannot ship without a purpose and retention entry | Product owner | Definition of done in the ticket |
| Vendors | Every processor is listed with its data category and review date | Vendor manager | Processor register |
Retention is a decision, not a default
Most applications do not have a retention problem because someone chose to keep data forever. They have it because nobody chose anything. The signup table grows, the export folder grows, and the backup schedule copies all of it. A retention schedule fixes that by naming a period and a method for each data class, then assigning the job of enforcing both.
Start with the classes that carry the most risk: identity documents, payment references, support conversations and marketing consent. For each, ask three questions. What purpose does this data still serve? What is the consequence of deleting it now? Who is allowed to approve an exception? The answers become the schedule.
Deletion also needs a method. A status flag is not deletion. Removing a row from the live table is not deletion if the same record remains in an export, a cache or a vendor system. Write down every place the record can exist, then test the path end to end.
How to verify the checklist
- Form audit. Open the five highest-traffic forms and map every field to the inventory. Any unmapped field is a finding.
- Database sample. Inspect a sample record and list every table that contains the same person. Copies that nobody owns are candidates for removal.
- Deletion request. Submit a test deletion and trace it through the application, backups, exports and vendor systems. Record the elapsed time.
- Export review. Ask for the last three exports and check who approved them, where they were stored and whether they were deleted.
- Permission test. Sign in as a support user and attempt to reach records outside the role. The expected result is a refusal, logged.
Run the deletion test at least twice a year. A process that worked during the project may not survive a change of staff or vendor.
Common mistakes
- Treating the privacy policy as the control. The policy describes intent; the form and the database determine the outcome.
- Minimising the signup form but leaving the support widget collecting everything.
- Deleting the live record and leaving the archive, export and CRM copy untouched.
- Letting analytics tools capture form contents through session recording or event tracking.
- Giving every engineer a production export permission because it is convenient.
- Setting a retention period with no job, calendar reminder or owner to enforce it.
Metrics worth reporting
- Number of personal fields without a documented purpose.
- Systems holding a copy of customer data, compared with the processor register.
- Median time to complete a verified deletion request.
- Dormant accounts past the retention threshold.
- New fields shipped without a retention entry, which should trend to zero.
Where our team fits
We review forms, data flows and access patterns, and implement retention and deletion routines under Managed Technology & Cybersecurity. Scope and starting points are on the pricing page. For the procurement side of the same subject, read data minimisation in web applications: a buying guide.
Sources
- NIST Privacy Framework, accessed 17 November 2026.
- UK Information Commissioner's Office, Guide to the data protection principles, accessed 17 November 2026.
- Regulation (EU) 2016/679, General Data Protection Regulation, accessed 17 November 2026.
- DataGuidance, Bangladesh jurisdiction overview, accessed 17 November 2026.
- OWASP Application Security Verification Standard, accessed 17 November 2026.
Responses (0)
No comments yet. Add the first one below.
Leave a response
Your email address will not be published. Required fields are marked *.