
Data classification: from label to control
Most people hear data classification and think of labels on documents. That is the last step. What sits underneath decides whether those labels mean anything: knowing which types of information you hold and where, who decides for each type, and what each level actually requires for storage, access, backup and retention.
Classify types of information, not individual files. Twenty to thirty types, each with a location and an owner, cover almost everything a small or mid-sized organisation holds.
Confidentiality is only one of three questions. Availability and integrity drive your backup and recovery time, and they are the ones most often forgotten.
A level without a table of controls is decoration. Each level should fix where the data may live, who may see it, how it may leave and when it must go.
What data classification is, and why the label comes last
Data classification is the agreement on how strongly you protect different types of information. The label on a document is the visible result, but the agreement itself sits a layer deeper: what information you have, how bad it is if it leaks, changes or disappears, and what you therefore put in place at each level.
Most projects do it the other way round. The labels come first and the question of what they mean comes later. The result is an organisation where “confidential” sits on half the documents and nobody can say what happens differently from “internal”.
So work in this order: a list of information types first, then an owner and a level for each, then a table that sets the controls for each level, and only then the technology that enforces it. ISO/IEC 27001 follows the same line: Annex A asks for an inventory of information and associated assets (5.9) before it gets to classification (5.12).
Start with what you have and where it lives
Classify types, not files. An organisation has millions of files and at most a few dozen types of information: customer records, payroll, quotes and contracts, designs or drawings, the accounts, personnel files, the password vault. In a small or mid-sized business twenty to thirty types cover almost everything.
For each type, record in a simple table:
- Where it lives: the system or location, such as the finance package, a SharePoint site, the directors’ mailbox or a folder on the file server.
- Where else it lives: the export to Excel, the email attachment, the copy in a test environment, the backup. That is where the surprises are, and a copy deserves the same protection as the original.
- Who sees it outside the organisation: the accountant, the payroll bureau, a supplier that runs the system.
The same table helps your UK GDPR record of processing: which types contain personal data is right there.
Three questions per type: confidentiality, integrity, availability
Look only at confidentiality and you have classified half of it. The usual triad is CIA: confidentiality (how bad is it if someone else sees it), integrity (how bad is it if it changes without anyone noticing) and availability (how long can you do without it). Score each type low, medium or high on all three.
- The staff rota is hardly confidential, but without it the business stops in the morning: availability high.
- Supplier bank details are no secret, but a changed account number is exactly how invoice fraud works: integrity high.
- Personnel files can be out of reach for a day but must never be public: confidentiality high.
The value is in what follows. Availability drives your backup and recovery time, integrity decides where you log changes and ask for a second pair of eyes, and only confidentiality is about labels and sharing. The UK government’s own scheme (OFFICIAL, SECRET, TOP SECRET) is built for its own information; a business does not need to copy it, but it does need all three questions.
What a level means: the table of controls
This is the document most organisations are missing, and without it a data classification policy is a colour code. For each level, set out:
- Where it may live: only in managed systems, or also on a laptop, in a personal cloud, on a USB stick.
- Who may access it: everyone in the organisation, a named group, or only on request with the owner’s approval.
- How it may leave: freely, only to known parties, only encrypted, or not at all.
- How it is protected: multi-factor authentication, encryption, logging of who opens it.
- Backup and recovery: how often, how far back, and how quickly it can be restored. That follows from the availability score.
- Retention and disposal: how long, and how it goes, including from retired devices.
Keep every cell short enough to check. “Appropriate security” is not a control; “only in the HR site, with multi-factor authentication, external sharing switched off” is.
Who decides: an owner for each information type
Every information type has an owner, and that is someone from the business, not from IT. The head of HR for personnel files, the finance lead for the accounts, the sales director for quotes. That person sets the level, approves access and reviews the list once a year.
IT is the custodian, not the owner. The custodian makes sure the controls in the table are actually configured, and flags it when practice drifts from the table. Let IT set the level and you get a classification that is technically sound and commercially wrong: the administrator does not know which quote is sensitive.
Agree as well what happens when an owner leaves. A type without an owner will, within a year, have access rights nobody can explain.
Retention, reclassification and disposal
A level is not for ever. A press release is highly confidential until the day it goes out and public after that. A quote is sensitive while the negotiation runs and mostly archive afterwards. Record for each type when the level changes, and who changes it.
And record when it has to go. Some minimum retention periods are set by law, such as those for tax records; for personal data the UK GDPR says the opposite, that you keep it no longer than necessary. Information that no longer exists cannot leak, and that is the cheapest control in the whole table.
Do not forget the places outside the main system: backups with long retention, old mailboxes of people who have left, and devices being retired. How that last part is done provably is on our page about data destruction and wiping.
Where our work stops
What we do. The part that is technology: mapping which systems and copies hold each information type, translating the table of controls into settings (permissions, multi-factor authentication, encryption, sharing settings, backup and recovery) and showing periodically where practice drifts from the table. We are ISO/IEC 27001 certified and work this way ourselves.
What we do not do. We do not decide how sensitive your information is; the owner in your organisation decides that, and we help turn it into a table and into settings. We are not lawyers and not a data protection officer: which retention periods apply to you by law, and whether something is special category data, is for someone qualified to judge. And we certify no one: whether your classification meets ISO/IEC 27001 is for a certification body to determine.
Questions we get about this
The five that come up most often.
What is data classification?
Data classification is the agreement on how strongly you protect different types of information. You group information types by how bad it is if they leak, change or disappear, and set out for each level what has to happen technically and organisationally. Labels on documents are the visible result, not the starting point.
What are the usual data classification levels?
Most organisations use three or four, such as public, internal, confidential and highly confidential. The number matters less than what each level means: without a table that sets storage, access, sharing, backup and retention per level, the names are decoration. Score availability and integrity as well, not just confidentiality.
Who is responsible for data classification?
An owner from the business for each information type, such as the head of HR for personnel files or the finance lead for the accounts. They set the level and approve access. IT is the custodian: it makes sure the agreed controls are actually in place.
Is data classification a legal requirement?
The UK GDPR does not use the term, but it does require security appropriate to the risk, and you cannot show that without knowing which data is sensitive. If you are pursuing ISO/IEC 27001, classifying information is required. Beyond that it is not an obligation but the way to decide where your security spending does the most good.
Will you decide which of our information is confidential?
No. How sensitive an information type is, is for its owner in your organisation to decide, and which retention periods apply by law is a legal question. We map where the information lives, translate the agreements into settings and show where practice drifts. We do not certify anyone either; that is for a certification body.
Further reading
What data classification connects to.
Shall we lay your information types next to your systems?
We map where each type of information lives, copies included, and which settings protect it today. You decide which level it belongs at.
Practical IT knowledge in your inbox
New guides on management, security and the workplace, written by the people doing the work. No sales talk, and you can unsubscribe in one click.