Data-location evidence checklist

Verify data locations using an editable register for primary records, backups, logs, AI endpoints, support access and migration copies.

Last updated

In short

  • Record storage, replication and administrative access separately for each data class.
  • Require dated evidence tied to the actual service and configuration.
  • An unknown location is an open assessment item, not an approved country.
On this page (6)

Procurement and evaluation · Part of Data residency

Data classes, processors, storage and access locations, retention and dated evidence fields.

Download the data-location checklist

Follow each copy of the data

A data-location review should trace a record through the services that store, process, copy and expose it. Asking where the main database runs is a start. It does not establish where backups, diagnostic logs, search indexes, support attachments or migration exports are held.

The editable register provides one row per data class and service. Its sample entries are illustrative and intentionally contain unresolved fields. They are not a statement about an actual customer’s architecture or a certification of a provider.

Begin with the purpose of the review. A contractual storage-location requirement, an internal access policy and a legal transfer assessment are different questions. Give each an owner so the technical team is not asked to infer a legal conclusion from a diagram.

Collect evidence at the right level

Data path Evidence to request Why a headline region is insufficient
Primary application data Deployment inventory and applicable contract schedule The selected region may cover only specified services
Backups and replicas Storage configuration, retention and restore location Recovery copies can use a different provider or region
Logs and monitoring Destination list and sample redacted events URLs, identifiers or document text may enter diagnostics
Support access Authorized teams, access procedure and audit record Remote access is a separate question from storage
AI processing Model, embedding, search and tool endpoints A local chat interface can call external services
Temporary exports Staging location, access owner and deletion schedule Migration introduces copies outside steady-state diagrams

Prefer configuration evidence tied to the deployed service over a generic region list. A contract establishes a commitment; configuration and audit evidence help establish what is actually operating. Keep both, and record the date because infrastructure and subprocessors can change.

Use a concrete data-flow exercise

Take a hypothetical support ticket containing a document attachment. Follow its creation, application storage, search indexing, notification email, backup, diagnostic capture and eventual deletion. Ask whether a support engineer can download it and where that engineer’s tooling stores the resulting file.

Then repeat the exercise for a failed request, because exception paths often capture more detail than successful ones. A stack trace or support upload can contain data that the main interface deliberately hides. Record uncertainty rather than assuming a path is empty because nobody has documented it.

Acceptance for this exercise is a complete route map with evidence references and named owners for remaining gaps. It is not a requirement to publish sensitive infrastructure details publicly. Store the detailed register in an access-controlled location and share a suitably redacted version when needed.

Compare providers using the same data classes

Pilae’s residency page describes its placement approach. For Microsoft 365, the service data-location documentation links to service-specific commitments and tenant information. Neither should be reduced to a yes/no label without checking scope.

A regional infrastructure service, a managed application provider and an internal datacentre may each satisfy different requirements. Include identity providers, payment or messaging systems, and tools run by your own staff. Keeping an application on premises does not stop an administrator from forwarding a support attachment elsewhere.

Where GDPR applies, processor arrangements and international transfers are addressed in the official regulation. Ask a qualified privacy or legal owner to interpret the relevant requirements for the actual parties and processing. Storage geography, corporate jurisdiction, access and transfer safeguards need separate analysis.

This checklist does not assert that a particular country, provider or encryption feature guarantees compliance. It helps produce the facts needed for a decision. If the buyer requires a contractual restriction, capture the precise data classes and permitted exceptions rather than relying on the word “sovereign.”

Verify changes and deletion paths

Assign someone to review new connectors, model providers, monitoring agents and support tools before they introduce additional destinations. Link the location register to change management so it is updated when architecture changes, not only during an annual questionnaire.

For deletion, distinguish active records, retained backups, audit logs and legal holds. A deleted user interface entry may remain in recoverable copies for an agreed period. Request the applicable retention procedure and evidence appropriate to the risk; do not promise immediate erasure from every medium without understanding the design.

At handover, confirm the register covers disaster recovery and exit, including temporary restored environments and exported records. A useful next action is to complete it for one sensitive workflow, then review the unresolved rows with the responsible operators and privacy owner. That turns a broad residency concern into an actionable placement assessment.

Sources and further reading

Use the linked documentation to confirm the version, edition and contract that apply to your project.

Review our placement evidence

Trace production data, backups, support access and external processing through your proposed architecture.