Skip to content
All posts

GRC

The Security Policies a Small Company Actually Needs

The seven short policies that cover most of the risk in a small company, what each must decide, what a usable one looks like, and how to keep them true.

HSHaris SiddiqueGRC Consultant

· 10 min read

On this page

Most small companies meet security policies the same way. A customer questionnaire asks "Do you have an access control policy?", someone downloads a pack of thirty templates, and the folder is filled by Friday. Six months later nobody can say what the policies require, and the first time anyone checks, the company is not doing what its own documents say.

This post is for founders, operations leads and the person who has been handed "sort out our policies". It sets out the small set of policies that cover most of the risk in a company of five to a hundred people, what each one must decide, what a usable policy looks like on the page, and how to keep the set alive without a compliance team.

Why most small-company policies fail

The templates are not the problem. A template is a reasonable starting point. The problem is that a policy is a promise, and a downloaded policy is a promise someone else wrote.

Three failures come up again and again when we review policy sets.

The policy describes a company that does not exist. It says access is reviewed every quarter, laptops are encrypted and backups are tested monthly. None of that is happening. The document is now worse than no document: it tells a customer, an auditor or an insurer something untrue, and it is in writing.

Nobody owns it. The policy says "IT is responsible". In a twenty-person company "IT" is one engineer who also ships features. When a rule has no named owner, it has no one to notice when it stops working.

It cannot be followed. Forty pages in the language of the standard, in a folder nobody opens. People learn the rules from whoever sat next to them in their first week.

Good policy work fixes all three by doing less, more honestly.

The seven policies that cover most of the risk

The NIST Cybersecurity Framework 2.0 puts policy under its Govern function, and its wording is a good test for any small company. GV.PO-01 asks that policy for managing cybersecurity risk is "established based on organizational context, cybersecurity strategy, and priorities and is communicated and enforced". Established, communicated and enforced: a template alone meets only the first.

Based on that test, this is the set we start small companies with. The Annex A references are to ISO/IEC 27001:2022, for anyone who will later pursue certification.

PolicyWhat it must decideEvidence it should produceISO 27001 Annex A
Information security policyWhy security matters to this business, who leads it, how other policies are approvedSigned by leadership, review dateClause 5.2, 5.1
Acceptable useWhat staff may do with company devices, accounts, data and AI tools; rules for remote workSigned acknowledgements at onboarding5.10, 6.7, 8.1
Access controlHow access is requested, approved and removed; MFA; who holds admin rightsJoiner and leaver records, access review results5.15, 5.18
Backup and recoveryWhat is backed up, how often, where copies live, how restores are testedBackup job reports, dated restore tests8.13
Incident responseWhat counts as an incident, who to tell, who decides, who talks to customersIncident log, exercise notes5.24 to 5.26
Supplier securityWhich suppliers need checks before use, and what the checks areSupplier list with risk level and review date5.19 to 5.21
Data handling and retentionHow data is classified, where each class may be stored, how long it is kept, how it is deletedData inventory, deletion records5.12, 5.13, 8.10

Two more become necessary depending on what you do. If you build software, add a secure development policy covering code review, secrets and dependency updates (Annex A 8.25 to 8.28). If you run your own servers or a large estate of laptops, add vulnerability and patch management with clear fix deadlines; our post on prioritising vulnerabilities with EPSS and KEV shows how to write that rule so it can be measured.

Why these seven

Each one answers a question that comes up in a real bad day. A laptop is stolen: acceptable use and data handling say whether it held customer data and whether it was encrypted. A leaver still has access to the billing system: access control says whose job it was to remove it. Ransomware hits the file server: backup and recovery says whether you can restore, and incident response says who makes the call. Our runbook for the first 24 hours of a ransomware incident shows how much depends on those decisions being made in advance.

What a usable policy looks like

A policy a small team will follow fits on one or two pages and has the same parts every time. Here is a shortened access control policy written the way we recommend.

ACCESS CONTROL POLICY                       v1.2 · Owner: Head of Engineering
Approved by: CEO, 2026-11-01                Next review: 2027-05-01

Purpose   Only the right people can reach company systems and data, and access
          is removed as soon as it is no longer needed.
Scope     All staff and contractors. All systems listed in the system inventory.

Rules
 1. Every account is named to one person. Shared logins must not be used,
    except the break-glass admin account, which is stored in the password vault.
 2. Multi-factor authentication must be on for email, the identity provider,
    the cloud console, source control and the finance system.
 3. Access is granted on request by the person's manager and approved by the
    system owner. The request is recorded in the IT ticket queue.
 4. Admin rights are held by no more than three named people per system.
 5. Leavers lose all access on their last working day. HR raises the ticket.
 6. Each system owner reviews who has access every six months and records
    the result, including what was removed.

Exceptions  Any exception is approved by the CEO, recorded in the exceptions
            log with an end date, and reviewed at that date.
Evidence    Ticket queue (joiners, leavers), access review records, MFA report
            from the identity provider.

Notice what is not there: no history, no definitions copied from a standard, no rule the team cannot meet. Every rule uses "must" or a number, so anyone can tell whether it is followed, and the evidence line says where to look.

Every policy needs the same six parts: a named owner (a person's role, not a department), approval and review dates, a one-line scope, numbered testable rules, an exceptions route with end dates, and an evidence line. Without an exceptions route, people break rules quietly instead.

What you can skip or merge for now

Policy packs often contain separate documents for clear desk, password, email, mobile device, remote working, encryption, cryptographic key management, logging, change management and more. In a small company most of these are two or three rules each, and they belong inside the seven above.

  • Password, mobile device, remote working, clear desk and email fit into acceptable use and access control.
  • Encryption becomes one rule in data handling ("laptops must use full-disk encryption; customer data at rest must be encrypted") and one in acceptable use.
  • Logging and change management can wait until you have systems that need them, then sit in a short operations standard.

Merging is not cutting corners: each rule lives in one place, has one owner and gets checked. Split them out when you grow.

How to make people actually follow them

GV.PO-01 says policy must be communicated and enforced. In a small company that means three things.

Put the policy where the work happens. People read a policy once, if then. They follow processes. So the leaver rule belongs in the HR offboarding checklist and the access request rule in the ticket form. The document records the decision; the process carries it out.

Enforce with settings, not reminders. Wherever a rule can be a setting, make it one. MFA can be required in the identity provider. Disk encryption can be enforced by device management. Personal file-sharing links can be blocked for the company drive. A rule a system enforces does not depend on anyone remembering it.

Ask for acknowledgement, then teach the few rules that matter. Have staff sign acceptable use at onboarding. Then spend ten minutes on the handful of rules that prevent real harm: how to report something suspicious, why they must never approve an MFA prompt they did not start, where customer data may and may not go. Our security awareness work starts from the same short list.

Keeping the set alive

GV.PO-02 asks that policy is "reviewed, updated, communicated, and enforced to reflect changes in requirements, threats, technology, and organizational mission". For a small company, two kinds of review are enough.

Calendar review. Once a year, each owner reads their policy, checks the evidence line, and either confirms it or changes it. Put the dates in a shared calendar, not in the policy folder.

Trigger review. Some events should reopen a policy straight away: a new major system, an incident or near miss, a new contract requirement, a big change in headcount, and any technology people start using before anyone has decided the rules. Generative AI tools are the current example. If staff are pasting customer data into a chatbot, your acceptable use and data handling policies need a rule this month, not at the annual review.

Keep a one-page register: policy, owner, version, last review, next review, and the open exceptions. That page is also the first thing a customer, an insurer or an auditor will want to see.

Questions to ask before you sign off a policy

  • Does every rule describe what we actually do today, or will do by a date we have agreed?
  • Is there a named owner who knows they own it?
  • Could someone outside the team test each rule and get a yes or no?
  • Is every rule that can be enforced by a setting actually enforced that way?
  • Is there a written way to request an exception, with an end date?
  • For each rule, do we know which record proves it is followed?
  • Is the review date in someone's calendar?

In our GRC engagements with small companies, we do the work in that order. We start by asking what the company actually does, not what the template says. We then write the seven policies against that, mark the gaps as dated actions rather than hiding them in the text, and link each rule to the record that proves it. If certification is the goal, the same set becomes the base of the ISMS; our post on what ISO 27001:2022 certification takes for small companies covers what comes next.

  • Policies
  • ISMS
  • Governance
  • SME
HS

Written by

Haris Siddique

GRC Consultant, FutaqSec

Keep reading

Not sure where to start?

A 30-minute scoping call is enough to understand your environment and recommend the right first step. No obligation.

Book a free consultation