A data classification policy is useful when people can turn a label into clear decisions about access, storage and sharing. To create one, define the classification levels and specify the handling controls that apply to each.
Labels such as public, internal, confidential and restricted are easy to understand in principle. The challenge is applying them consistently across teams, systems and information types, so sensitive data receives appropriate protection without creating unnecessary friction.
This guide explains what a data classification policy should cover and how to set meaningful criteria for each level. It outlines practical controls for access, encryption, sharing, retention and review. The aim is a workable framework that protects information while supporting business operations.
A data classification policy groups information according to factors such as sensitivity, business impact and the potential consequences of exposure, alteration or loss. It connects those categories to handling rules so staff can make consistent decisions about access, storage and sharing. For example, customer records may need more protection than marketing material approved for publication.
For personal data, classification can help an organisation apply appropriate handling and security measures. UK GDPR does not prescribe a universal classification scheme, so internal labels should support the organisation’s obligations rather than be presented as proof of compliance. ICO guidance on special category data explains that special category data is personal data requiring additional protection because of its sensitivity, with specific requirements governing its processing.
Map applicable legal, regulatory and contractual requirements to the information and processes they affect. The requirements depend on the organisation’s activities and sector. Record the rationale for the controls selected and review it when relevant obligations or business processes change. The ICO’s guidance on data protection by design reinforces this risk-based approach, requiring appropriate technical and organisational measures throughout the information lifecycle.
Assign responsibility for applying and governing the policy. Data owners can classify information and oversee its use, while governance leaders approve the policy and security teams support its implementation and review. Staff need clear instructions for handling information and a route for asking questions when a case is unclear.
Use realistic examples in training, such as contrasting a public product page with confidential customer records. Explain the definitions and handling rules, then reinforce them with scenarios relevant to different roles. This helps staff understand how to classify information and what to do when a case does not fit neatly into one category.
Set a review process that accounts for changes to data, systems and business activities. Name a policy owner and record decisions, approvals and revisions. Common problems include vague labels, inconsistent classification and controls that do not reflect how information moves through cloud services, shared drives, paper records and suppliers.
Turn each classification into specific handling rules that staff can apply consistently. The policy should connect labels to practical controls, monitoring and evidence.
Define each level using observable criteria, such as the likely impact of unauthorised access or disclosure. Specify the types of information covered and give examples drawn from actual business records. Avoid relying on terms such as ‘sensitive’ without explaining what they mean in practice.
Clear criteria help teams make consistent decisions across different systems. They should also explain who is responsible for resolving uncertain cases and how classification decisions are recorded.
For each level, document how information is created, stored, shared, archived and deleted. A confidential file, for example, may require approved storage and controlled sharing. Encryption may also be appropriate, depending on the information, its location and the organisation’s risk assessment.
Set rules for external sharing, retention and secure disposal. A classification should remain useful throughout the information lifecycle, not just when a document is created. Where tools such as Microsoft Purview are used, specify whether a label only identifies sensitivity or also applies configured protection, such as encryption or access restrictions. Microsoft Learn confirms that Microsoft Purview sensitivity labels can be used without protection settings or can be configured to enforce controls including encryption, access restrictions and content markings.
Specify who may access each classification and the business reason for that access. Need-to-know access limits information to people with a legitimate requirement, while role-based permissions can help apply that restriction consistently. For higher-risk information, consider stronger measures such as multi-factor authentication and access logging where they are appropriate to the systems and risks involved.
State who grants, reviews and removes permissions so access remains aligned with job responsibilities. Where access governance forms part of the classification model, CyberOne’s Identity & Access Management consulting covers identity governance, privileged access management and access control.
Logging and periodic access reviews can help identify permissions that no longer fit a person’s role. Where technology supports automation, apply labels or policy actions against clear rules and review exceptions.
A four-level model can give teams a practical way to distinguish information by sensitivity. These labels are a starting framework, not a universal standard. Adapt their meanings to your organisation’s information, risks and existing controls.
Public information is approved for release outside the organisation, such as a published annual report. Internal information supports business activity but is intended for staff or authorised partners, such as an internal project plan. Confidential information could cause commercial or personal harm if disclosed, such as an unreleased product roadmap. Restricted information requires the narrowest access, such as authentication secrets or highly sensitive investigation material.
Define classifications by consequences and permitted use rather than relying on labels alone. Staff should be able to distinguish an internal draft from a document approved for publication without interpreting the policy differently from their colleagues.
Assess the information, its context and the impact of exposure, alteration or loss. A dataset may contain fields with different levels of sensitivity. Decide whether to classify the whole file according to its most sensitive content or separate it into controlled components. Record the reasoning for unusual cases, especially when teams share data across functions or with external parties.
Keep the model proportionate. If staff struggle to distinguish confidential from restricted information, clarify the criteria before adding another level. Additional labels are useful only when they lead to distinct handling decisions.
Specify how each label appears in the formats people use, including document headers, email subject lines and shared repositories. Define who applies the label, when it should change and how it remains visible on printed or exported copies. These details help recipients understand handling expectations when information moves between teams.
Use consistent terminology across digital and physical environments. A label identifies a classification, but the protection depends on the rules and controls applied to it. With Microsoft Purview, this distinction is especially important because a sensitivity label can simply identify sensitivity or can enforce configured protection settings. Make sure staff can see what action the label requires.
A template offers a structured starting point for documenting classifications, responsibilities and handling procedures. It cannot determine which information is critical to your organisation or how specific obligations apply. Adapt the framework to your data, operating environment and existing controls rather than implementing it unchanged.
Before adopting a template, check that it covers the operational detail staff and reviewers need:
Avoid running multiple classification schemes without a clear mapping. If a template uses a numerical scale alongside named levels, document how each value corresponds to a label, which controls it triggers and how staff should apply it across systems. Without that mapping, the same information could receive conflicting classifications in different environments.
Test each clause against actual practices before approval. Use examples such as a public report, an internal procedure and restricted records. Confirm that responsibilities, controls and exceptions work in daily procedures, then record approval and assign an owner for future updates. A useful policy is clear enough to apply, not merely comprehensive on paper.
CyberOne’s Data Security as a Service, powered by Microsoft Purview, supports data discovery, classification and the management of data security policies. This may be relevant when policy definitions are in place but the organisation needs support connecting them to data protection controls and business processes.
For organisations implementing Microsoft Purview, CyberOne also offers Microsoft Purview Deployment. The service includes data discovery, classification strategy, Microsoft Information Protection configuration and phased deployment, making it a particularly direct next step when an organisation wants to translate a classification policy into technical controls.
Where access governance is also a priority, CyberOne’s Identity & Access Management consulting can support alignment between identities, permissions and organisational responsibilities. Organisations that first need a broader picture of their security posture can instead use a Cyber Security Gap Analysis & Maturity Assessment to identify gaps and establish improvement priorities.
A policy creates value when its decisions become part of everyday operations. Connect ownership, approval and review with existing governance so classification can keep pace with changes to data and business priorities. Clear accountability helps teams sustain the policy after publication.
For organisations using the Microsoft security ecosystem, classification can form part of a broader data protection model rather than operating as an isolated policy. CyberOne’s Data Security as a Service provides the closest fit where ongoing Microsoft Purview-based data security management is required.
If data security needs to operate alongside broader detection and response, CyberOne separately offers MXDR as a Service for continuous monitoring, investigation and response within Microsoft environments.
For organisations specifically developing or implementing classification controls, Microsoft Purview Deployment provides the most direct next step for connecting policy decisions with data discovery, classification and technical protection.