Talk to a specialist
Home / Resources / SOC 2 guide / SOC 2 for SaaS and cloud service providers

SOC 2 · Practical business guide

SOC 2 for SaaS and cloud service providers

Plan SOC 2 around the SaaS service customers use. Connect tenant boundaries, privileged support, releases and cloud responsibilities to practical controls and evidence.

In this guide
Build assurance around the service customers actually consumeResolve the service boundary before collecting evidenceMake tenant isolation a lifecycle decisionControl privileged support without making service unworkableConnect frequent releases to complete change recordsUnderstand cloud and software dependencies at the responsibility boundaryHypothetical example: an enterprise feature changes the assurance boundaryHypothetical example: an acquisition introduces a second operating modelUse management review to find changes in the servicePrepare a customer assurance package that stays accurateReference basisFrequently asked questionsPlan the next decisionRelated SOC 2 guides

Build assurance around the service customers actually consume

For a SaaS business, SOC 2 preparation should begin with a precise account of the service, its commitments and the activities that could undermine them. A platform may share infrastructure across thousands of tenants while offering different identity options, support privileges and recovery commitments. Those differences influence scope and control design. A generic collection of company policies cannot explain them adequately.

SOC 2 is an independent CPA examination relating to a service organisation’s system and controls. Its subject matter can address security, availability, processing integrity, confidentiality or privacy according to the engagement. Management remains responsible for the service description, commitments, controls and evidence; software and consulting can support that work.

This guide helps SaaS leaders and technical teams examine tenant boundaries, support privileges, release records and supplier responsibilities before they commit to a customer assurance deadline. Confirm formal criteria mapping and engagement decisions with the independent firm using authorised current AICPA materials. The examples are hypothetical and illustrate decisions rather than actual customer engagements.

Resolve the service boundary before collecting evidence

Name the product, production environments, management interfaces, relevant integrations and service functions. Identify whether onboarding, customer support, billing operations and professional services interact with customer information or influence the service commitments. A function does not become irrelevant simply because it sits outside engineering or uses a separately purchased application.

Describe what varies by product tier or deployment model. A managed private environment may use different release approvals, recovery methods or administrator access from the shared platform. Decide whether those variations can be represented accurately within one scope. An enterprise feature advertised to customers should not disappear from the description merely because few tenants currently use it.

Follow actual customer information through ingestion, processing, storage, export and deletion. Include queues, diagnostic tools, support uploads and backups when relevant. An architecture diagram is useful only if it reflects the system being operated. Pair it with an owner-maintained inventory and a process for reviewing material changes.

Document excluded products and the rationale clearly. Check that marketing, procurement responses and report distribution do not imply coverage broader than the examination. The system description has its own criteria, distinct from the Trust Services Criteria used to evaluate controls. Treat description accuracy as an active management responsibility.

Make tenant isolation a lifecycle decision

Tenant isolation depends on more than the main application query. Review how tenant context is established, propagated and checked across background jobs, storage paths, search indexes, exports and support operations. Identify what could happen if a request contains an unexpected identifier or if a job retries with incomplete context. The recommended design review should address the service’s actual architecture.

Separate design confidence from operating evidence. A threat model or test plan may explain why the architecture should protect boundaries. Test execution, defect resolution and relevant release records show what the organisation did. Neither a diagram nor a single successful test establishes that every later change preserved the boundary.

Define how new features enter this review. A bulk export, analytics dashboard or administrative search can create an access path different from ordinary customer interaction. Give the responsible engineering owner a clear trigger for involving security and the relevant data owner. Retain the decision, identified risks and validation appropriate to the change.

Account for internal tools. Employees may access tenants through a management console with permissions broader than customer roles. Decide which tasks require that access, how they are authorised and how misuse or error can be detected. A strong customer authentication feature does not compensate automatically for an uncontrolled internal support path.

Control privileged support without making service unworkable

Map the situations in which staff need customer access: troubleshooting, restoring configuration, investigating abuse or helping with a failed integration. For each situation, define the permitted action, approval route, access duration and record. Where the platform supports constrained or temporary access, evaluate how it can reduce standing privilege while allowing the service team to resolve incidents.

Clarify the customer’s role. Some services require explicit tenant approval before support entry; others rely on an agreed contractual model. Make the operational procedure match the actual commitment. If urgent access is allowed under defined conditions, explain who can authorise it and what later review is required. Do not leave staff to infer exceptions from sales conversations.

Keep support evidence useful and proportionate. A ticket can record why access was needed, the authorisation, session reference, actions taken and closure. It should not become a repository of unnecessary customer content or secrets. Link protected technical records where possible, with access limited to people who need to review them.

Test the difficult cases: a departed employee with a separate console account, a contractor working outside normal hours, or an escalation that changes from observation to data modification. The purpose of the exercise is to discover where identity, approval and logging processes diverge before those gaps become routine.

Our SOC 2 access-control guide provides a practical account-lifecycle approach for privileged support, emergency access and reviews that lead to decisions.

Connect frequent releases to complete change records

A SaaS team can deploy often and still maintain clear accountability. Start with the normal release path: proposal, review, validation, authorisation, deployment and verification. Identify which steps are enforced by the platform and which depend on a person. The evidence should allow a reviewer to connect a production change to its decision and outcome without relying on one engineer’s memory.

Establish the population of changes from a reliable production source. A ticket list may omit direct configuration changes, infrastructure updates or emergency actions. Reconcile the sources relevant to the platform so the organisation understands what changed and where evidence resides. Complete population management is a practical preparation discipline; the auditor determines the procedures and samples for the engagement.

Design emergency handling deliberately. Define who may act, what minimum record must exist and how review follows when service is stabilised. Preserve the actual sequence. Retrospective approval should be described as such, not entered in a way that appears to precede the deployment.

Review the pipeline itself as an important part of the service. Changes to build permissions, signing keys, deployment identities or branch protections can affect many subsequent releases. Give those changes ownership and evidence appropriate to their impact, rather than treating the delivery platform as an invisible background utility.

For the approval and verification chain behind frequent releases, read our SOC 2 change-management guide.

Understand cloud and software dependencies at the responsibility boundary

Identify the services whose failure or misuse could affect the scoped platform. Consider infrastructure, identity, monitoring, customer communications, payment workflows and specialist processing services where relevant. Assign an owner who understands both the contract and the technical integration. Purchasing approval alone does not establish that the service is configured or monitored appropriately.

A supplier report can inform assessment, but read its scope, period, relevant control information and customer responsibilities in context. The AICPA’s public report walkthrough identifies topics such as complementary user entity controls, subservice organisations and carve-out or inclusive treatment. Detailed interpretation should be addressed with qualified reviewers and the independent CPA.

Translate applicable responsibilities into your own control register. If the SaaS organisation must configure access, retain certain records or manage encryption settings, name the owner and evidence. A supplier’s assurance does not show that your team performed those tasks. Revisit the mapping when the service configuration or supplier offering changes materially.

Plan for dependency disruption. Decide what the platform can continue to do, how customers will be informed and how restoration will be verified. A recovery plan that assumes every external service remains available may look complete while failing to address the interruption that matters most to customers.

Hypothetical example: an enterprise feature changes the assurance boundary

Consider a SaaS workflow provider preparing for a larger customer procurement review. Its existing platform has a shared application and database, with tenant-specific authorisation enforced in normal requests. An enterprise customer asks for a support-assisted bulk export that combines workflow records and uploaded files. The sales team views it as a small feature because it uses information already stored.

The engineering review identifies a separate worker that fetches files using a service identity. The worker accepts a list generated by the application, but its retry path does not preserve the original tenant context. The team pauses the feature release and designs explicit tenant checks for both the initial request and retries. It tests mismatched identifiers, interrupted jobs and access revocation during processing.

Management also defines who can initiate the export, how customer approval is recorded and how long the generated package remains available. Support receives a procedure that links the request to the export job without copying customer files into the ticket. The platform owner adds the new flow and storage location to the service inventory and assesses the description implications.

For readiness review, the team can show the design decision, defect record, tests, release approval and a controlled demonstration of the request workflow. The review establishes the feature’s tenant boundary, support responsibilities and information lifecycle, with evidence of the decisions and tests behind them. Any examination conclusion remains the independent firm’s responsibility.

Hypothetical example: an acquisition introduces a second operating model

Imagine a scheduling SaaS company that acquires an analytics product. Both brands will be sold together, but the acquired team uses a different cloud account, deployment process and support provider. Leadership initially wants one report that can address customer assurance requests for both products. The programme owner first compares the actual systems and commitments rather than assuming a common corporate owner creates a common control environment.

The assessment finds that the analytics service receives exports from the scheduling platform and retains them under a separate deletion process. Its support provider can access customer datasets through a standalone account store. Central identity controls therefore do not cover every relevant user, and the main platform’s departure evidence cannot demonstrate timely removal from the acquired service.

Management chooses a staged integration. It assigns accountable owners to the remaining local controls, documents the data transfer and retention decisions, and implements a coordinated departure check while identity integration is planned. The team confirms the proposed scope and timing with the independent CPA and updates customer-facing statements to describe the actual coverage.

The evidence plan reflects both operating models until integration is complete. It records which controls are shared, which remain local and when changes become effective. This avoids treating an acquisition announcement as proof of operational consolidation. It also gives leadership a decision record for whether to accelerate integration, narrow near-term scope or accept a later examination milestone.

Use management review to find changes in the service

Create a review rhythm that brings product, engineering, security, support and commercial owners together around material changes. Useful inputs include new integrations, privileged access trends, significant incidents, unresolved control issues and commitments made to customers. The review should identify decisions and accountable actions, not merely repeat a dashboard score.

Check whether evidence reflects the whole relevant population. Compare identity inventories with actual consoles, production changes with release records, and suppliers with observed data flows. Investigate unexplained differences. A small number of sampled records may be reassuring, but management still needs to understand the completeness of the process it claims to operate.

Treat evidence access as a security concern. Store customer-sensitive material in approved locations, define reviewer permissions and remove unnecessary copies. Where a record must be redacted, preserve its meaning and explain the redaction. Do not edit substantive facts to make the record appear consistent with the expected outcome.

Use findings to improve control design, execution and the records needed to evaluate them. Distinguish a poorly designed process, a failure to follow an adequate process and an evidence problem. Those causes require different corrective actions. Keep the issue visible until the chosen action has been implemented and its effectiveness checked.

Prepare a customer assurance package that stays accurate

Decide who can receive the report, through which approved channel and with what accompanying explanation. SOC 2 reports are intended for knowledgeable, restricted users rather than unrestricted public marketing. Follow the report’s use provisions and obtain advice where distribution questions arise. AICPA’s current quality discussion also warns against treating a fast, superficial exercise as credible assurance.

Help customers understand the service and period covered, significant exclusions and responsibilities they must perform. Avoid a badge or sales statement that implies every product, location and future period is covered. When a buyer asks about a later change, answer with a factual account of that change and available evidence rather than assuming an older report resolves the question.

Maintain a controlled owner for procurement responses. Engineering and support should be able to correct statements about features or practices before they become commitments. Feed recurring customer questions into product and risk decisions so assurance work improves the service’s clarity as well as its documentation.

Reference basis

AICPA SOC suite of services; AICPA SOC 2 reporting guide public description; 2018 SOC 2 Description Criteria, revised implementation guidance 2022; AICPA SOC 2 report walkthrough public outline; Journal of Accountancy: promises of fast and easy threaten SOC credibility.

Frequently asked questions

Does SOC 2 apply only to large SaaS providers?

Company size alone does not determine whether the work is useful. Customer assurance needs, service commitments, operational maturity and procurement expectations drive the decision. Scope and control design should reflect the actual service.

Does a cloud provider report cover our SaaS application?

It can support assessment of the provider within its stated scope. Your application, configurations, people, integrations and applicable customer responsibilities still require management attention and evidence.

Can we keep deploying during preparation?

Yes, provided the organisation manages changes through the intended process and keeps the description and evidence current. Discuss material changes and their engagement implications with the independent CPA.

Must every SaaS provider include all five categories?

Category selection should follow the service commitments, relevant risks and engagement scope. Confirm the formal selection and applicable criteria with the independent firm rather than using a universal checklist.

What should the technical team prepare first?

Prepare a reliable service boundary, data-flow view and ownership map. Then test representative access, support and release workflows to identify design and evidence gaps before scaling collection.

Plan the next decision

Use a SaaS readiness discussion to resolve the platform boundary, tenant administration model and evidence gaps that could delay a customer review. Share a high-level architecture summary and the service commitments under consideration; keep customer records, credentials and proprietary configuration details out of the initial enquiry.

Discuss a customer SOC 2 requirement for your SaaS platform with CyberVantage360. Our SOC 2 readiness and implementation service can be scoped around the product and tenant boundary, privileged support design, release evidence and a delivery plan grounded in the required report scope.