Every organisation has a retention policy somewhere. Very few have a deletion workflow that works. The gap becomes visible when a customer asks for their data to be removed: one company produces a signed deletion record within a week, another discovers the record exists in four systems, two backups and a marketing spreadsheet.
This buying guide is for the manager who has been asked to formalise retention and deletion, usually because of a privacy obligation, an enterprise customer's due-diligence questionnaire or an audit finding. It covers the requirements to specify, a vendor scorecard, pricing questions, the pilot and the exit.
What you are actually buying
Not a policy document. You are buying three capabilities that have to work together:
- A data map. Where each class of personal data lives, including SaaS tools, analytics, email and backups.
- A rule engine. A way to express retention periods per data class and per legal basis, with exceptions for legal hold.
- An evidence trail. Who approved a deletion, what was deleted, where, when, and what failed.
If a vendor can only produce the first two, you are buying a cleaner, not a control.
Requirements to put in the document
Coverage
- Connectors for your actual systems: relational database, object storage, CRM, support desk, marketing platform, analytics, email and backup storage.
- Deletion inside backups, or a documented expiry path if the vendor cannot reach them.
- Search indexes, caches and data warehouses, which are usually forgotten and always searchable.
Policy and workflow
- Retention rules by data class, with a start event such as account closure or last activity.
- Legal hold that suspends deletion without disabling the rule.
- Approval steps with named approvers and separation of duties.
- Exception handling when a system is unreachable.
Evidence and requests
- A deletion certificate per request that lists systems, record counts and timestamps.
- An immutable audit log of rule changes, approvals and executions.
- A subject-request workflow with identity verification and a tracked response clock.
- Reporting that shows overdue deletions before an auditor finds them.
The workflow in seven steps
- Inventory: list systems, owners and data classes.
- Classify: mark personal, sensitive, financial and operational data.
- Set rules: retention period and trigger event per class.
- Check holds: legal, contractual or investigation holds override the rule.
- Approve: a named person signs off, with the scope visible.
- Execute: delete or anonymise across every mapped system, with retries.
- Verify and record: confirm absence in primary, replica, index and analytics, then store the certificate.
Step seven is where most tools are weakest and where your auditor will spend the most time.
Vendor scorecard
Weights should follow your risk, but this is a reasonable starting point for a mid-market buyer.
- System coverage (25): connectors you need today, plus the two systems you plan to add next year.
- Backup and archive deletion (20): a clear mechanism, not "backups expire eventually".
- Workflow and approvals (15): separation of duties, legal hold, exception handling.
- Evidence quality (15): certificates, audit log export, retention of the audit log itself.
- Subject-request support (10): intake, verification, response tracking.
- Security posture (10): least privilege, encryption, sub-processor list, residency.
- Exit (5): export of policy definitions and a deletion certificate at termination.
Score evidence, not slides. Ask each vendor to show a real deletion certificate from a customer environment with the data redacted.
Pricing questions
- Is pricing per connector, per data source, per terabyte, per request or per seat?
- What does implementation cost, and are custom connectors charged separately?
- Is the audit log included, and for how long?
- Are subject requests metered, and what happens at the limit?
- Is a sandbox included for testing rules?
- What are the exit fees, and how are policy definitions exported?
- Do sub-processors add cost or require separate contracts?
Model a five-year cost that includes an extra connector each year and a rise in subject requests. Per-request pricing looks cheap in a pilot with twelve requests and expensive in year three.
The pilot
Pick one data class and two systems. Dormant customer accounts are a good choice because the rule is easy to explain and the risk of a wrong deletion is manageable with approval.
- Duration: 45 days, including one full cycle.
- Success test: deletion verified in the primary database, a read replica, the search index and the analytics copy, with a backup expiry date recorded.
- False positives: count records that should have been retained because of an open order, dispute or legal hold.
- Manual effort: hours spent by your team per cycle, which is the number that decides renewal.
Exit terms
Retention tools accumulate your most sensitive rules and evidence. Before signing, agree that you can export policy definitions in a readable format, receive a deletion certificate covering the vendor's own copies, and get a defined transition period. If the vendor cannot describe its own deletion on exit, do not trust its deletion of yours.
The Bangladesh obligation map
Three sources usually shape the retention matrix: the Personal Data Protection Act, 2026 for personal data, sector directives from Bangladesh Bank for financial services, and contract terms from enterprise customers. Map each data class to the strictest applicable rule and record the reasoning. Where two rules appear to conflict, document the decision and send it for legal review instead of choosing the convenient reading.
Where deletion fails in practice
- Backups retained indefinitely because nobody owns the expiry calendar.
- Aggregated reports that no longer look like personal data but can be re-identified.
- Team spreadsheets and exports that were never in the data map.
- Free trials and departmental SaaS tools bought outside procurement.
- Email attachments and shared drives holding the same customer record.
- Suppliers who keep a copy under a vague contract clause.
Where this fits in our work
If you need the data map, the retention matrix and the deletion evidence built together, start with Managed Technology & Cybersecurity; the pricing page explains how compliance work is scoped. For a sector example, read how student data privacy obligations change year by year.
Sources
- NIST, Privacy Framework, accessed 17 November 2026.
- NIST SP 800-88 Rev. 1, Guidelines for Media Sanitization, accessed 17 November 2026.
- NIST SP 800-53 Rev. 5, Security and Privacy Controls, accessed 17 November 2026.
- ISO/IEC 27701, Privacy information management, accessed 17 November 2026.
- DataGuidance, Bangladesh jurisdiction overview, 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 *.