Private Client · Europe
Securing a SaaS Platform After a Near-Miss: Our SOC 2 Journey
They needed SOC 2 to win enterprise deals and had 90 days to prepare. We built their entire security program from scratch and passed the audit with zero findings. Here's the 90-day plan.
Measured results
- Passed SOC 2 Type II audit with zero findings
- Implemented role-based access control across all systems
- Enabled encryption at rest and in transit for all data
- Built incident response plan with 24-hour detection target
Challenge
The blocker
A SaaS platform processing sensitive customer data needed SOC 2 Type II certification to win enterprise contracts. They had 90 days to get ready for the audit and no formal security program in place.
Solution
The approach
We built a complete security program from scratch: access controls, encryption, monitoring, incident response, and documentation. They passed the SOC 2 audit with zero findings.
Execution story
The CEO called us on a Monday. A large enterprise customer — one that would double their annual revenue — required SOC 2 Type II certification before signing the contract. The customer gave them 90 days to get certified. The company had no formal security program, no documented policies, and no access controls beyond "everyone is an admin."
Ninety days. Zero findings. Here's how we did it.
The starting point
We assessed where they were on day one:
| SOC 2 Requirement | Status |
|---|---|
| Access controls | None (all 30 employees had admin access to everything) |
| Encryption at rest | None |
| Encryption in transit | Partial (HTTPS on the website, not on internal APIs) |
| Audit logging | Minimal (application logs only, no access logs) |
| Incident response | No plan |
| Vulnerability management | None |
| Data retention policy | None |
| Background checks | None |
| Security training | None |
| Vendor management | None |
Everything needed to be built. The audit was in 90 days. The audit covers the previous 3-12 months of operations, which meant we needed the controls in place by day 30 to have 60 days of evidence.
The 90-day plan
| Phase | Days | Focus |
|---|---|---|
| 1 | 1-30 | Implement all technical controls |
| 2 | 31-60 | Operate controls, collect evidence |
| 3 | 61-90 | Audit, remediation, certification |
Phase 1: Technical controls (days 1-30)
Access controls (days 1-5)
The first and most important change was removing admin access from everyone. We implemented role-based access control (RBAC) across all systems:
| Role | Count | Access |
|---|---|---|
| Admin | 3 (CTO, Head of Eng, Security Lead) | Full access to all systems |
| Engineer | 12 | Access to code repositories, staging environment, production logs (read-only) |
| Support | 8 | Access to customer data (read-only), no access to infrastructure |
| Sales | 4 | Access to CRM, no access to customer data or infrastructure |
| Finance | 3 | Access to accounting system, no access to customer data or infrastructure |
We used Okta for SSO with MFA enforced for all users. Service accounts were moved to short-lived credentials managed by a secrets vault.
This was the hardest change politically. Twelve engineers who had admin access to production were now read-only. We handled this by explaining the business context (the enterprise deal required it) and by making the access request process fast (engineers could request temporary elevated access with manager approval, auto-expiring after 4 hours).
Encryption (days 6-10)
We enabled encryption at rest for all data stores:
- PostgreSQL: TDE (Transparent Data Encryption) enabled on the RDS instance
- S3: Server-side encryption with KMS-managed keys
- EBS volumes: Encrypted with AWS KMS
- Backups: Encrypted with separate KMS keys
Encryption in transit was already partially in place (HTTPS on the website). We added TLS to all internal API calls, enforced HTTPS on all endpoints with HSTS, and rotated all TLS certificates to ensure none were older than 12 months.
Audit logging (days 11-15)
We implemented comprehensive audit logging:
- AWS CloudTrail: Enabled in all regions, logs sent to a dedicated logging account
- Application logs: All authentication events, data access events, and configuration changes logged
- Database logs: All DDL statements and data access logged
- Access logs: All access to production systems logged through a bastion host
Logs were sent to a SIEM (we used Datadog) with a 12-month retention period. We set up alerts for:
- Failed login attempts (potential brute force)
- Access to production outside business hours
- Data exports larger than 1,000 records
- Configuration changes to security groups or IAM policies
Vulnerability management (days 16-20)
We implemented vulnerability scanning:
- Infrastructure scanning: AWS Inspector on all EC2 instances
- Container scanning: Trivy in the CI pipeline, builds fail on critical vulnerabilities
- Dependency scanning: Snyk on all repositories, weekly reports
- Web application scanning: OWASP ZAP in CI, runs on every deployment
We also ran a penetration test in week 3 to identify any gaps before the audit.
Incident response (days 21-25)
We wrote an incident response plan that covered:
- Detection: How incidents are detected (alerts, user reports, third-party notifications)
- Triage: Who assesses severity and decides on response
- Containment: Steps to contain the incident (isolate systems, revoke access)
- Eradication: Steps to remove the threat
- Recovery: Steps to restore service
- Post-incident: Root cause analysis, documentation, process improvements
We ran a tabletop exercise in week 4 to test the plan. The exercise revealed two gaps (no clear escalation path for after-hours incidents, no pre-drafted customer communication templates) that we fixed before the audit.
Policies and documentation (days 26-30)
The auditor needs documentation. We wrote:
- Information security policy
- Access control policy
- Data retention and disposal policy
- Incident response policy
- Change management policy
- Vendor management policy
- Employee onboarding and offboarding checklist
- Security awareness training materials
Each policy was reviewed and signed off by the CEO. The documentation lived in a dedicated Notion workspace that the auditor could access.
Phase 2: Operate and collect evidence (days 31-60)
For 60 days, we operated the controls and collected evidence that they were working:
- Monthly access reviews (who has access to what, is it still appropriate)
- Daily log reviews (security alerts reviewed and dispositioned)
- Weekly vulnerability scan reviews
- Change management tickets (every production change documented)
- Incident response drill (one simulated incident, full response)
The evidence was organized in a structured folder that mapped directly to the SOC 2 criteria:
| SOC 2 Criteria | Evidence |
|---|---|
| CC1 (Control Environment) | Security policy, board approval, org chart |
| CC2 (Communication) | Training records, policy acknowledgments |
| CC3 (Risk Assessment) | Risk register, quarterly risk reviews |
| CC4 (Monitoring) | Alert logs, access reviews, vulnerability reports |
| CC5 (Control Activities) | Change management tickets, access request tickets |
| CC6 (Logical Access) | RBAC matrix, MFA enrollment, access reviews |
| CC7 (System Operations) | Uptime monitoring, patch management records |
Phase 3: Audit (days 61-90)
The auditor (a Big 4 firm) conducted the audit over 3 weeks. They reviewed the policies, interviewed key personnel, tested the controls, and sampled evidence.
The audit found zero exceptions. Every control was designed correctly, operating effectively, and supported by sufficient evidence.
The numbers
| Metric | Before | After |
|---|---|---|
| Admin access | 30 employees | 3 employees |
| MFA coverage | 0% | 100% |
| Encryption at rest | None | All data stores |
| Audit logging | Minimal | Comprehensive, 12-month retention |
| Incident response plan | None | Documented and tested |
| Vulnerability scanning | None | Continuous (infra, container, dependency, web) |
| SOC 2 audit findings | N/A | 0 |
What we learned
SOC 2 is not about being perfect. It's about being consistent. The auditor doesn't expect zero security incidents. They expect:
- You have controls. Documented, approved, and communicated.
- The controls work. Evidence shows they're operating as designed.
- You can prove it. Logs, tickets, reviews, and records that demonstrate the controls are working over time.
The 90-day timeline was aggressive but achievable because the company was small (30 employees) and the CTO was committed. A larger company with more legacy systems would need more time.
The single most impactful change was removing admin access from everyone. Not because admin access is inherently wrong — in a small startup, everyone having admin is convenient. But it's the first thing an auditor checks, and it's the hardest to remediate quickly because it requires political will, not just technical work.
If you're preparing for SOC 2, start with access controls. Everything else is easier.