A disaster recovery plan is only useful when your SaaS team can execute it under pressure. This guide provides a repeatable SOC 2 readiness approach for selecting scenarios, documenting RTO and RPO targets, running a tabletop exercise, testing backup restoration, tracking remediation, and preserving evidence an auditor can evaluate.
Overview
SaaS disaster recovery testing should demonstrate more than the existence of backups or a written business continuity policy. The goal is to show that your team understands critical services, can make decisions during a disruption, and has tested whether systems and data can be recovered within business-approved targets.
For SOC 2 readiness, recovery testing is usually most useful when it connects three things:
- Defined expectations: Which systems must be restored, in what order, and within what timeframe?
- Demonstrated capability: Can the team follow the recovery procedure using available credentials, tools, documentation, and vendor support?
- Traceable improvement: Were failures, delays, assumptions, and follow-up actions recorded and resolved?
Your exact testing approach should reflect your architecture and audit scope. A single-region application, a multi-region platform, and a SaaS product dependent on several external providers will have different recovery risks. Document those assumptions instead of adopting a generic cloud business continuity test.
Start by mapping responsibilities between your company and its cloud providers. The cloud compliance shared responsibility matrix can help identify which recovery tasks belong to your team, which belong to a provider, and which need evidence from both.
What to track
1. Select realistic disaster scenarios
Choose scenarios that could interrupt availability, integrity, or access to customer data. Test one scenario in depth rather than discussing every possible disaster superficially.
| Scenario | What to examine | Useful evidence |
|---|---|---|
| Accidental deletion or corruption | Point-in-time recovery, restore permissions, data validation, and customer impact | Backup configuration, restore logs, validation notes, issue record |
| Cloud region or availability failure | Failover sequence, DNS or routing changes, dependencies, and monitoring | Runbook, change records, monitoring output, timeline |
| Ransomware or malicious encryption | Isolation, identity control, clean recovery point, and incident coordination | Incident plan, access logs, immutable backup settings, exercise notes |
| Critical vendor outage | Vendor communication, workaround, service dependency, and recovery responsibility | Vendor plan, contract contact details, status updates, decision log |
| Credential or secrets compromise | Revocation, replacement, privileged access, and safe restoration | Access review, rotation record, approval trail, test results |
Include service dependencies such as identity providers, DNS, payment systems, queues, object storage, databases, observability platforms, and customer notification channels. A restore test that excludes a required dependency may produce an overly optimistic result.
2. Record RTO and RPO assumptions
Recovery time objective (RTO) is the target time to restore a service or capability. Recovery point objective (RPO) is the maximum acceptable period of data loss, expressed as a point in time. These are business and technical planning targets, not merely settings in a backup console.
Use an RTO and RPO worksheet for each critical service:
| Service or data set | Business owner | RTO target | RPO target | Recovery method | Last test | Result |
|---|---|---|---|---|---|---|
| Production database | Product or engineering | Document target | Document target | Snapshot, replica, or backup restore | Date | Pass, partial, or fail |
| Object storage | Operations | Document target | Document target | Versioning or backup recovery | Date | Pass, partial, or fail |
If no approved target exists, mark the value as an open decision. Do not replace an unknown requirement with an arbitrary number.
3. Run a tabletop exercise
A tabletop exercise tests decisions and coordination without necessarily changing production systems. Invite the incident lead, engineering or infrastructure owner, security representative, customer support, communications, and a business owner. Include vendor contacts when a provider is part of the recovery path.
A practical 60-to-90-minute agenda is:
- Briefing: State the scenario, scope, assumptions, and success criteria.
- Initial response: Ask who declares the incident, who has authority, and how the team confirms the event.
- Containment: Discuss access restrictions, isolation, traffic management, and preservation of evidence.
- Recovery: Walk through the runbook, dependencies, restore points, validation steps, and escalation paths.
- Communications: Identify internal updates, customer messaging, vendor coordination, and approval requirements.
- Debrief: Record decisions, blockers, untested assumptions, and assigned actions.
4. Perform a backup-restore test
Use a non-production environment whenever possible. Confirm that the test is authorized, isolated, and safe for customer data. A basic backup restore test procedure should include:
- Select a backup or recovery point and record its timestamp, source, retention status, and integrity information.
- Confirm the people, permissions, tools, and vendor documentation required to restore it.
- Restore the data or service into an isolated environment.
- Record start and completion times to calculate actual recovery duration.
- Validate application startup, authentication, key workflows, data completeness, and integrations.
- Compare the recovered state with the expected RPO and document any missing or inconsistent data.
- Clean up the test environment and protect or delete test data according to your retention rules.
Capture screenshots, command output, tickets, logs, configuration exports, and validation results where appropriate. Evidence should make the test understandable to someone who did not participate.
Cadence and checkpoints
Set a recurring schedule rather than treating disaster recovery as a one-time SOC 2 project. The right cadence depends on system criticality, change volume, contractual commitments, and risk. A practical tracker can use the following checkpoints:
- Monthly: Review backup job failures, missed recovery points, retention exceptions, monitoring alerts, and changes to critical dependencies.
- Quarterly: Confirm recovery owners and contacts, review RTO and RPO assumptions, inspect open findings, and conduct a focused restore or tabletop test.
- After major changes: Retest when you change regions, databases, deployment architecture, identity systems, backup tooling, key vendors, or recovery runbooks.
- Before an audit period: Confirm that evidence is complete, timestamps are clear, test results are approved, and overdue remediation has an owner and due date.
- Annually: Reassess the full business continuity plan, including scenarios, dependencies, contact lists, communication paths, and recovery priorities.
Keep an evidence register with the test date, scenario, participants, systems in scope, expected targets, observed results, exceptions, and links to supporting records. The SOC 2 audit evidence collection calendar provides a useful model for assigning recurring evidence tasks and owners.
How to interpret changes
A passing test does not mean the recovery program is finished. Compare each result with the previous test and investigate meaningful changes.
- Recovery time increased: Check for larger data volumes, slower dependencies, changed permissions, manual steps, or an outdated runbook.
- Recovery point worsened: Review backup frequency, failed jobs, replication lag, retention settings, and the time at which the incident was assumed to begin.
- A dependency was unavailable: Record whether the dependency had a documented workaround, provider commitment, or unresolved single point of failure.
- Validation failed: Determine whether the issue affected data integrity, application functionality, access control, or only the test procedure.
- Participants were uncertain: Treat unclear ownership or approval authority as a control weakness, even if the technical restore succeeded.
Classify findings by impact and urgency. Assign each one an owner, due date, corrective action, and retest condition. A useful issue log includes: issue ID, scenario, observation, affected system, risk, immediate workaround, permanent remediation, owner, target date, status, and closure evidence.
Do not quietly edit failed results into a pass. Preserve the original record, explain the cause, document the decision, and show the retest. Transparent evidence is more useful for improving resilience and preparing for assurance activities than an unexplained success label.
When to revisit
Revisit this disaster recovery testing plan on a monthly or quarterly cadence, and whenever the architecture or service relationships change. The trigger is not limited to a production outage. A new database, backup provider, cloud region, identity platform, subprocessor, deployment model, or customer data flow can invalidate previous recovery assumptions.
Before the next test, take these practical steps:
- Export a current service and dependency inventory.
- Confirm RTO and RPO targets with business owners.
- Check backup success, retention, immutability, and restore permissions.
- Choose a scenario that exercises a known risk or unresolved finding.
- Invite participants and assign an independent note-taker.
- Prepare the tabletop agenda and restore test procedure.
- Open an issue log before the exercise and assign remediation owners afterward.
- Store the final results with supporting evidence in a location accessible to the compliance team and audit stakeholders.
For broader SOC 2 preparation, connect recovery findings to your SOC 2 controls mapping matrix and recurring evidence calendar. If a recovery event involves unauthorized access or personal data, coordinate the response with your incident procedures and applicable privacy obligations. Repeating this cycle—plan, test, record, remediate, and retest—turns disaster recovery from a document into an operating capability.