trio.so
Templates

Data Classification Policy Template (Word + Excel Matrix)

Free data classification policy template in Word and PDF: four levels with examples, a handling matrix, labeling rules, forms and an Excel register.

Mountain landscape representing leadership perspective and vision
Written by
Atena Ladchartabi

Content writer, trio.so

Published on
Modified on

This template gives you the rules for sorting company information by how much harm its exposure would cause, and for handling each level the same way every time. It names the levels, says who decides a dataset's level, how documents and email are labeled, and what people may do with each level: where it can be stored, how it can be sent, whether it can go on a USB drive or a personal phone, and how it is deleted.

This guide walks through each part for an IT manager, security lead or data protection officer at a company of 20 to 1,000 people. The download has the policy in Word and PDF, with a one-page handling matrix, the classification matrix, an inventory worksheet, request and acknowledgment forms, a labeling guide and a filled-in example, plus an Excel register that checks your data inventory for you.

In short

  • Use four levels (Public, Internal, Confidential, Restricted), or three if four is too many for your staff.
  • Decide a level by the impact of disclosure, change or loss. Personal data is never Public.
  • Write handling rules for each level: storage, sharing, email, printing, devices, cloud apps, removable media, disposal.
  • Give every dataset an owner, label what is Confidential or above, and review levels and access on a schedule.

What Is a Data Classification Policy?

A data classification policy (some companies call it an information classification policy) is the document that sets your sensitivity levels and the rules for each. It answers three questions for every employee: how sensitive is this file, how do I mark it, and what am I allowed to do with it. ISO/IEC 27001:2022 asks for exactly this in Annex A control 5.12 (classification of information) and 5.13 (labelling of information), and CIS Controls v8 Safeguard 3.7 asks for a data classification scheme that you review every year.

The policy sits between two others. The data encryption policy says how encryption is done; the classification policy says which data must be encrypted. The data retention policy says how long each record is kept; the classification policy says how it is disposed of once that time is up.

How to Write a Data Classification Policy in 7 Steps

  1. List your datasets. Start with the places personal and financial data lives: CRM, HR, finance, file shares, email.
  2. Name an owner for each. A business lead, not IT.
  3. Set the levels. Four (or three), each with a plain definition and examples your staff recognize.
  4. Classify by impact. Use the matrix; personal data is never Public.
  5. Write the handling rules. Storage, sharing, email, printing, devices, cloud apps, removable media, disposal.
  6. Label and train. Headers, footers, subject tags, file names, then a short session with real examples.
  7. Review. Check levels and access on a schedule and the policy once a year.

Start with the datasets that would hurt most if they leaked: payroll, HR records, customer exports, contracts, source code. You don't need a full inventory before the policy takes effect. A short, correct list beats a long one nobody maintains.

The Four Classification Levels, with Examples

Section 5 of the template defines four levels. Most templates and most companies use these four names; the definitions and examples are what make them usable.

LevelWhat it coversExamples
PublicInformation approved for release. Disclosure causes no harm.Website pages, press releases, published price lists, job ads
InternalDay-to-day business information for staff. Disclosure would be awkward but cause little harm.Org charts, internal guides, meeting notes, most email
ConfidentialInformation whose disclosure could harm the company, customers or staff, or break a contract.Customer lists, contracts, source code, unpublished financials, most personal data
RestrictedThe most sensitive information. Disclosure could cause serious harm, regulator action or lawsuits.Payroll and bank details, health data and other special category data, passwords and keys, deal plans

Two default rules prevent most arguments: anything without a level is Internal until its owner classifies it, and personal data is never Public. Special category data under GDPR Article 9 (health, biometric and genetic data, racial or ethnic origin, political opinions, religious beliefs, trade union membership, sex life or sexual orientation) goes to Restricted unless the owner and the privacy lead write down why a lower level is enough. Payment card data and health records under HIPAA are other common Restricted examples.

If four levels are too many, use three: merge Restricted into Confidential and keep the Restricted handling rules for a named list of datasets such as payroll, health data, and passwords and keys. Microsoft's guidance for its sensitivity labels makes the same point from the other side: labeling works less well when people face more than five main labels.

How to Classify Data by Impact

The data owner sets the level, using the classification matrix (Appendix B of the template). For each dataset, ask what would happen if it were disclosed, changed or lost, in five areas: money, legal and regulatory, reputation, harm to people, and operations. The highest impact in any area sets the level.

The classification matrix (Appendix B): the highest impact in any row sets the level
The classification matrix (Appendix B): the highest impact in any row sets the level

This follows the US federal method without copying it. FIPS 199 rates the potential impact on confidentiality, integrity and availability as low, moderate or high, and NIST SP 800-60 maps common information types to those ratings. NIST SP 800-53 control RA-2 asks organizations to categorize their information, document the rationale and have the decision approved. The template uses its own four levels and treats FIPS 199 as a rough guide: Internal is about low, Confidential about moderate, Restricted about high.

Watch for combinations. A list of employee names is Internal; the same list joined with salaries and bank details is Restricted. When in doubt, pick the higher level and ask the security lead.

Data Classification Matrix: Handling Rules by Level

The handling matrix is the part employees use. It turns each level into concrete do's and don'ts. Section 8 of the template has nine rules for each level, and Appendix A repeats them on one printable page:

RulePublicInternalConfidentialRestricted
AccessAnyoneAll staff and approved contractorsPeople with a business need, granted by roleNamed people approved by the data owner
StorageAny approved systemCompany systems and managed devicesApproved company systems and managed, encrypted devicesSystems the owner names; encrypted; access logged
Sharing outside the companyAllowedWith a manager's OK, to partners under contractUnder an NDA or contract, encrypted, owner informedOnly with the owner's approval, encrypted, to a named recipient
Email and file transferNo limitsCompany email and approved file sharesCompany email with encryption, or an expiring link to an approved share; never personal emailEncrypted attachment or expiring encrypted link only; never personal email or chat apps
PrintingNo limitsCollect prints right awayPrint only when needed; never leave copies on desks or printersAvoid printing; numbered copies, locked away, shredded after use
Mobile devices and BYODNo limitsManaged devices, or personal devices allowed under the BYOD policyManaged devices; personal devices only in approved apps, no local copiesCompany-owned, managed devices only
Cloud apps and AI toolsAnyApproved company appsApproved apps with company sign-in and MFA; no personal accounts or public AI toolsApproved apps named in the inventory, with MFA and access logs; no AI tools unless approved for this data
Removable mediaAllowedCompany-issued drivesEncrypted company drives only, loggedNot allowed without an approved exception
Retention and disposalDelete when no longer neededFollow the retention scheduleFollow the retention schedule; delete securely; shred paperFollow the retention schedule; delete securely with a record; wipe or destroy media

Edit the cells to match your tools: name your approved file share, your secure email option and your AI tool policy. The rule for cloud apps and AI tools matters more each year, because pasting a customer list into a public AI chat is sharing outside the company.

Encryption Expectations by Level

Section 9 sets what each level needs at rest, in transit, on removable media and on mobile devices:

LevelAt restIn transitRemovable mediaMobile devices
PublicNot neededTLS on public sitesAllowed unencryptedDevice encryption applies anyway
InternalFull-disk encryption on every company laptop and phoneTLS for email and web appsCompany-issued drivesPasscode and device encryption on
ConfidentialEncrypted storage and full-disk encryption on devicesTLS 1.2 or later; encrypted links or attachments outside the companyEncrypted company drives onlyManaged device with encryption and a passcode
RestrictedEncrypted storage with access to keys limited to IT; recovery keys held centrallyEncrypted end to end where possible; never plain emailNot allowed without an exceptionCompany-owned, managed, encrypted devices only

Full-disk encryption on every company laptop and phone covers Internal data at rest and makes a lost device far less serious. CIS Controls v8 has four encryption safeguards: data on end-user devices (3.6), on removable media (3.9), in transit (3.10) and at rest (3.11). The table covers all four. The data encryption policy template sets algorithms, key management and recovery keys in detail.

Labeling Documents, Email and Files

A level helps only if people can see it. Section 7 says where the label goes, and Appendix F gives the exact text:

WhereHow to labelRequired for
DocumentsLevel in the header or footer of every page, and in the document propertiesConfidential and Restricted; Internal optional
EmailLevel in the subject line, for example [Confidential], or the email client's sensitivity labelConfidential and Restricted
Files and foldersLevel in the file name or metadata; shared folders and sites carry the highest level they holdConfidential and Restricted
Systems and repositoriesLevel recorded in the data inventory and in the system's description or settingsEvery system
Removable media and paperA sticker or cover sheet with the level; Restricted copies numberedConfidential and Restricted
Spoken informationSay the level before sharing it in a meeting or callRestricted

Keep the labels short and the same everywhere, for example "Confidential" in the footer, "[Confidential]" in an email subject and "_CONF" at the end of a file name. If you use Microsoft 365, sensitivity labels in Microsoft Purview can add headers, footers and watermarks and encrypt the file, and they can be applied automatically or recommended to the user. A label stored in a file's metadata can also be read by other tools.

Roles: Data Owner, Custodian, User and Security Lead

Classification fails when IT is expected to decide what the finance team's spreadsheets are worth. Section 4 splits the work:

RoleResponsibilities
Data ownerThe business lead who is accountable for a dataset. Sets its level, approves access, approves reclassification and reviews both on schedule.
Data custodian (IT)Runs the systems that hold the data and applies the technical rules for each level: encryption, access, backups, logging, disposal.
Data users (all staff and contractors)Handle data by its level, keep labels in place, and report data that looks wrongly classified or exposed.
Security lead (policy owner)Owns this policy, keeps the data inventory, approves exceptions and runs the yearly review.
Privacy lead or DPOAdvises on personal and special category data and on retention, and is told when personal data is reclassified.
ManagersMake sure their teams are trained and that new datasets get an owner and a level.

Access Reviews, Reclassification and Exceptions

Levels and access drift over time. Section 12 asks data owners to review who has access to Confidential and Restricted datasets on a schedule (the example uses every 180 days for Confidential and every 90 days for Restricted) and to remove access the day someone leaves. Section 13 covers reclassification: only the owner changes a level, levels go down when the reason is gone (a launch makes a deck Public) and up when data is combined. Section 14 handles exceptions, for example a payroll export that can't land in an encrypted folder until a software upgrade, with a compensating control and an expiry date no more than 12 months out. Section 15 asks for training when people join and a refresher every year, with examples from their own work, and everyone signs the acknowledgment form in Appendix E. The access control policy template covers how access is granted in the first place.

Data Classification Policy Example: Kestrel Bay Software

The download includes the policy filled in for Kestrel Bay Software (fictional), a made-up 140-person company that sells scheduling software. Its security lead classified four everyday datasets with the matrix:

DatasetWhat decides itLevelHandling
Customer listNames, work emails and contract values exported from the CRM. Disclosure would hurt customers' trust and could be a reportable personal data breach.ConfidentialFooter and file name carry the level; it stays in the CRM and the sales share, never on a USB drive or personal cloud; the event agency gets it only under the data processing agreement, through an expiring encrypted link.
Payroll fileNames, salaries, bank details and tax IDs; sick-pay lines can reveal health information, a GDPR Art. 9 special category.RestrictedOpen to the two payroll staff and the outside accountant; encrypted share; never sent as an email attachment; deleted from the export folder after upload.
Marketing deckSlides for a product launch in January. Early disclosure would tip off competitors, but no one is harmed.InternalShared on the company drive; the marketing lead reclassifies it as Public on launch day and removes the Internal footer.
Source codeThe product itself. No personal data, but a leak would let others copy features or find security flaws.ConfidentialKept in the company's Git hosting with MFA; repositories marked Confidential in their settings; signing keys and secrets live in a vault and are Restricted.

Then the register checked the first inventory on 16 November. 9 datasets, 4 with no open issue:

DatasetLevelResult
Customer list (CRM export)ConfidentialOK
Payroll file (monthly)RestrictedNot encrypted, Review overdue
Q1 launch deckInternalOK
Product source codeConfidentialNot labeled
Employee sick notesConfidentialSpecial category below top level
Newsletter sign-ups exportPublicPersonal data marked Public, No review date
Board minutesRestrictedReview due soon
Support tickets backupConfidentialNo owner, Not encrypted, Not labeled, No review date
Public price listPublicOK

The checks found what a first inventory usually finds. The payroll export sits in an unencrypted folder under an exception that expired at the end of October. A support desk backup on a USB drive has no owner, no label and no encryption. A newsletter export holding personal data is marked Public, and employee sick notes, which are health data, sit one level too low.

Common Data Classification Mistakes

  • Too many levels. Six levels with fine distinctions make people guess; three or four with clear examples work.
  • Marking everything Confidential. When every file has the same label, the label tells nobody anything and the handling rules get ignored.
  • No owners. IT can't judge the business value of a sales export. Every dataset needs a named business owner.
  • A policy with no handling rules. Levels without rules for email, USB drives, printing and cloud apps change nothing.
  • Forgetting paper, spoken information and data held by vendors.
  • Classifying once and never again. Data changes, gets combined and gets copied; levels and access need a review date.

How DLP Tools Use Classification Labels

Data loss prevention tools enforce handling rules at the point where data leaves: an email to an outside address, an upload to a personal cloud account, a copy to a USB drive. Many of them can act on classification labels. Microsoft Purview stores a sensitivity label in clear text in the file's metadata, so other apps and services can read it and apply their own protection. That is why the labeling section matters even if you don't run a DLP tool today: labels you apply now are what a DLP rule can use later. Our data loss prevention policy template sets the DLP rules, and the guide to data loss prevention with MDM covers the device side.

Data Classification in ISO 27001, NIST, CIS Controls and the GDPR

No framework prescribes your level names. They ask that you have a scheme, apply it and protect data by its level. The template's mapping (Appendix H) points each section to these references:

FrameworkWhere classification appears
ISO/IEC 27001:2022A.5.12 classification of information and A.5.13 labelling of information, with A.5.9 inventory, A.5.10 acceptable use, A.5.14 information transfer, A.7.10 storage media, A.8.10 information deletion, A.8.12 data leakage prevention and A.8.24 use of cryptography.
NISTSP 800-53 Rev. 5 RA-2 security categorization and MP-3 media marking; FIPS 199 and SP 800-60 for impact levels (background).
CIS Controls v83.7 data classification scheme, reviewed yearly; 3.1 data management process; 3.2 data inventory; 3.3 access control lists; 3.4 retention; 3.5 disposal; 3.6, 3.9, 3.10 and 3.11 encryption.
GDPRArticle 9 special categories of personal data and Article 32 security of processing, which asks for measures that fit the risk, including encryption.
HIPAA and PCI DSSExamples of regulated data (health records, payment card data) that usually sits at the top level.

The mapping is a reading aid; your auditor or counsel decides what counts for your scope. For the wider frameworks, see the ISO 27001 checklist and the GDPR compliance checklist.

Data Classification Register (Excel)

The workbook turns the policy into a list of things to fix:

  • Data Inventory flags datasets with no level or no owner, Confidential or Restricted data that isn't encrypted or labeled, personal data marked Public, special category data below the top level, and level reviews that are overdue or missing.
  • Handling Matrix holds the rules per level; the inventory reads whether each level must be labeled and encrypted from it.
  • Access Reviews flags reviews that are overdue, due soon or never done, with the interval set per level.
  • Exceptions shows which approvals are missing, expired or about to expire.
  • Settings, Example, Dashboard and Sources hold the levels and review intervals (switch to three levels by clearing one row), the filled-in example, counts per level and the references. Formulas only, no macros.
The register's Example tab: a made-up company's inventory, with the checks it raises
The register's Example tab: a made-up company's inventory, with the checks it raises

What's in the Download

The policy PDF: cover, handling rules by level and the classification matrix
The policy PDF: cover, handling rules by level and the classification matrix
  • Word policy (Data-Classification-Policy-trio.so.docx): 17 sections, the quick reference, the classification matrix, the inventory worksheet, the reclassification and exception form, the acknowledgment form, the labeling guide, the filled-in example and the framework mapping, with fillable fields.
  • PDF (Data-Classification-Policy-trio.so.pdf): the same policy, forms and example, ready to print.
  • Excel register (Data-Classification-Register-trio.so.xlsx): the tabs described above.

Where trio.so Fits

trio.so does not classify, label or scan your data, and it has no DLP of its own. It covers the device side of the handling rules: the laptops and phones that store and carry classified data stay encrypted, USB storage can be blocked where the policy says so, lost company devices can be locked or wiped, and the console shows who holds admin rights. In practice that means:

  • A device inventory with each device's owner, serial number, installed apps and encryption status.
  • BitLocker policies on Windows, FileVault policies on Macs, passcode policies for iPhone, iPad and Android, and required storage encryption on Android.
  • BitLocker and LUKS recovery keys held in the console, where revealing one takes the right role and a fresh sign-in.
  • An alert when a device reports an unencrypted disk.
  • A block on USB storage for company-owned Windows, Linux and Android devices and supervised iPhones and iPads (Macs have no storage block).
  • Policy status for each device: applied, failed or pending.
  • Console roles, so each admin gets only the permissions their job needs, and a CSV export of console members with their roles, status and last-active dates.
  • User sync from Microsoft Entra ID and Google Workspace and SCIM 2.0 provisioning; MFA and single sign-on for your users are in the Zero Trust plan.
  • Remote lock on Mac, iPhone, iPad and Android, and remote wipe of company-owned devices.
  • An integration with Forcepoint DLP that brings its incidents into the console and links them to managed devices.
  • Compliance evidence exports you can add to the audit file.

It manages Windows, macOS, Linux, iOS/iPadOS and Android from one console. To block USB storage on Windows step by step, see how to disable USB ports. Plans start at $5 per device.

Start free trial Book a demo

14-day free trial. No credit card required.

Related Templates and Guides

Sources. Last reviewed November 2026. Framework references are short pointers to ISO/IEC 27001:2022, NIST SP 800-53 Rev. 5 and FIPS 199, the CIS Controls v8 and the GDPR; the wording of every section is ours. trio.so is not affiliated with or endorsed by ISO, NIST or the Center for Internet Security.

Ready-to-use Templates

Must-have Template Toolkit for IT Admins

Explore All
Template Toolkit

Keep classified data on encrypted, managed devices

trio.so sets BitLocker, FileVault and phone encryption, shows each device's encryption status, blocks USB storage on company-owned Windows, Linux and Android devices, and wipes lost company devices.

Start free trial

Manage Every Device From One Console

trio.so brings device management, patching and security policies for Windows, macOS, Linux, iOS/iPadOS and Android into one console. Enroll a device and see it in the console within minutes.

14-day free trial. No credit card required.

trio.so brings device management, patching and security policies for Windows, macOS, Linux, iOS/iPadOS and Android into one console. Enroll a device and see it in the console within minutes.

Smiling womanAbstract geometric patternAbstract geometric patternSmiling womanSmiling woman

Frequently Asked Questions (FAQ)

Have questions? We've got answers. This section covers some of the most commonly asked questions related to this topic.

It is a written set of rules that sorts company information into sensitivity levels, says who decides each dataset's level and how it is labeled, and sets what people may do with each level: storage, sharing, email, printing, devices, cloud apps, removable media and disposal.

Most companies use Public, Internal, Confidential and Restricted. Public is cleared for anyone, Internal is for staff, Confidential is for people with a business need, and Restricted is for named people only, such as payroll, health data and passwords.

It is a grid with the levels across the top. One version lists impact criteria (money, legal, reputation, people, operations) to decide a level; the other lists handling rules for each level. The template includes both, as Appendix B and Appendix A.

The data owner, a business lead accountable for the dataset, sets its level and approves access. IT, as custodian, applies the technical rules. Every employee handles data by its level, and the security lead owns the policy and the inventory.

Review the policy and the scheme at least every 12 months, as CIS Controls v8 Safeguard 3.7 asks, and after major changes. The template also sets review dates per dataset, for example every 365 days for Confidential and every 180 days for Restricted.

ISO/IEC 27001:2022 Annex A includes 5.12 classification of information and 5.13 labelling of information. If those controls are in your Statement of Applicability, auditors expect a documented scheme, labels and handling rules, which this template provides.

Not under this template. Personal data is Internal at the lowest, and usually Confidential. Special category data under GDPR Article 9, such as health data, is Restricted unless the owner and the privacy lead record why a lower level is enough.

Related

From the blog

The related industry news, interviews, technologies, and resources.