ESSENTIAL GUIDANCE · BUYER GUIDE · 7 min read

Part of Validate insights →

How to scope a penetration test: a practical guide

This guide explains how to scope a penetration test, from defining your goals and approved targets to agreeing access, rules of engagement and retesting.

Business review lead: Mathew Colley · Sales Director

Start with the business services and risks that need assurance. Map the systems, identities, APIs, cloud resources and third-party dependencies that support them. Then note what is in scope, what is out, what you are assuming, how testing may run and how newly found assets will be handled.

FREE EDITABLE RESOURCE

Download the penetration test scoping workbook

The free penetration test scoping workbook helps buyers record what the test needs to prove, which targets are approved, and how the work will run. It is available as an editable Word file and a PDF.

  • Scope and access: Record the business objective, target systems, exclusions, user roles and test access.
  • Safety and ownership: Agree testing windows, rules of engagement, escalation contacts and stop conditions.
  • Evidence and sign-off: Specify the report, remediation support, retest expectations and who approves the scope.

Download editable Word workbook ↓ Download PDF version

No form is needed. Filling in the workbook does not allow testing; active testing still needs explicit written authorisation from the relevant system owner.

Define the assurance question

State what the organisation needs to learn or protect: whether an unauthenticated user can reach customer data, whether a normal user can cross a tenant or role boundary, whether a major release introduced a new attack path, or whether agreed fixes close the previously demonstrated finding. The answer tells you which assets, identities, techniques and proof the engagement needs.

Map the business service and its dependencies

Map the domains, IP ranges, applications, APIs, cloud resources, identity providers, data stores, remote access services and third-party dependencies that support each key user flow. Compare asset records with DNS, certificate, cloud, identity, supplier and recent-change records. If the inventory is incomplete, say so and agree whether discovery is part of the work.

Include identities, roles and business flows

List the roles and trust boundaries that need testing, including unauthenticated users, normal users, admins, service accounts, separate customer tenants and third-party dependencies. Identify high-impact flows such as account recovery, payment, data export, approval and role change. Provide sample test accounts where they can be used safely.

Define the approved scope precisely

Record the penetration testing scope: applications, APIs, domains, IP ranges, cloud environments, user roles and test environments, plus exclusions, limits and assumptions. Every active target needs written permission from its owner; a supplier link, passive external visibility or technical access does not grant permission to test a third party. For each major exclusion, note the untested risk it leaves.

Agree rules of engagement

Set the approved targets and scope version, testing windows, source addresses, allowed and banned methods, data-handling rules, escalation contacts, stop conditions and change-control route. Explain how major findings and possible service impact will be handled. These controls are the rules of engagement for penetration testing and make the work safer without turning it into a simple checklist.

Independent guidance: Read the NCSC penetration-testing guidance →

Allocate responsibilities and prerequisites

State who will provide test accounts, architecture information, allow-listing, test data, access and internal approvals; who will watch the environment; and who can pause or extend testing. Record dependencies, lead times and what happens if prerequisites are late or missing. Clear roles protect the schedule, business assumptions and safety of the work.

Prepare access, context and test data

Provide dedicated accounts for the roles and boundaries in scope, agree how credentials will be sent and removed, and confirm VPN routes, source-address allow-listing, client certificates and other access needs before the test window. Share useful design, data-flow, API and permission details. Use representative non-production data where possible and keep any sensitive proof to the minimum needed to show a finding.

Prepare monitoring and recovery

Confirm who will watch service health and security alerts, what logs should be kept, who can isolate an account or service and how an unexpected incident will be handled. Check backup and recovery plans, but do not treat a backup as permission to use unsafe methods. The rules of engagement remain the main control.

Define critical escalation and stop conditions

Agree what needs immediate contact rather than waiting for the final report, such as confirmed access to highly sensitive data, an active compromise, unexpected disruption or contact with a third party that was not approved. Name primary and backup contacts and who can pause testing. Unless incident response has been agreed in advance, the provider only reports and escalates as agreed; the customer's incident team handles containment, review and notification decisions.

Control late discoveries and scope changes

If testing finds an unlisted host, API, identity route or supplier link, pause work on it, confirm ownership and permission, assess risk and record whether it is added, delayed or excluded. Being tied to an in-scope domain does not by itself permit testing a newly found asset.

Describe the approach and depth

Explain how automated coverage and human review will be mixed, which recognised methodologies guide the work and how business logic, access control and chained weaknesses will be checked. Note any assumptions about accounts, architecture information or test data. The method should support consistency and safety while still leaving room for expert judgement.

Specify evidence and outputs

Define what the customer will receive: live notice of major findings, reproducible technical evidence, severity rationale, practical remediation guidance, an executive view, a technical report and any portal access. Make sure limits and untested areas are stated clearly so the final report is not read as assurance for a wider environment.

Include remediation support and retesting

The proposal should explain how questions will be handled after delivery, which fixes are eligible for retest, the retest window and what closure proof will be given. A report records the findings; retesting checks whether the agreed fixes address the previously demonstrated finding within the agreed retest scope and records any remaining limits.

Hold a final go or no-go review

Confirm that written permission is complete, the scope and rules use the same version, accounts and access routes work, monitoring and escalation contacts are ready, needed supplier approvals exist and no recent change or incident makes the window unsafe. If a key prerequisite is missing, delay, reduce or revise the scope rather than letting an unrecorded trade-off become the plan.

Understand the limits

A penetration test is a time-boxed check of an agreed scope. It cannot prove that no flaw exists, that an incomplete inventory is complete or that a service will stay secure after it changes. Combine testing with secure design, asset management, monitoring, vulnerability management and sensible retesting.

Q&A

Why should scoping begin with business services rather than a list of IP addresses or applications?

Starting with business services helps the team test what matters most: the systems, identities, data flows and dependencies that support key user journeys and risks. A target list alone can miss identity routes, APIs, cloud resources, suppliers or business logic paths that are key to the service.

Does finding an asset during testing mean it can be tested right away?

No. If an unlisted host, API, identity route or supplier link is found, pause work until ownership, written permission and risk are checked. Being tied to an in-scope domain or having network access is not enough to permit testing.

What should be in the rules of engagement?

The rules of engagement should define the approved targets, scope version, testing windows, source addresses, allowed and banned methods, data-handling rules, escalation contacts, stop conditions and change-control route. They should also explain how major findings and possible service impact will be handled.

How should sensitive data and proof be handled during a penetration test?

The test should use representative non-production data where possible and keep sensitive proof to the minimum needed to show a finding. Data-handling rules should be agreed before testing starts, and the final outputs should provide reproducible technical evidence without extra exposure.

What are the main limits of a penetration test?

A penetration test is time-boxed and limited to an agreed scope. It cannot prove that no flaws exist, that an incomplete inventory is complete or that a service will stay secure after later changes. It should be paired with secure design, asset management, monitoring, vulnerability management and sensible retesting.

Ready for the Pentesys point of view?

This guide covers the essentials. Continue into our technical analysis for a firmer position, practical implications and recommended action.

Read “A testing methodology should protect rigour without constraining curiosity” →
TALK TO PENTESYS

Apply this guidance to your environment.

Tell us what you need to understand, test or prove. We'll help you choose a proportionate next step—without forcing the conversation into a predefined package.

Discuss your penetration-testing scope →Explore Validate

Keep reading

Pentesys trust and assurance

Recognised credentials. Responsible AI. Experienced people.

Clear evidence of our company standards, testing services and access to AI programmes.

Independent company assurance

CREST member company for Penetration Testing in EMEA

CREST member company

Penetration Testing in EMEA. Company standards independently assessed by CREST.

Valid until 1st November 2027View company certificate
CREST Security Testing accreditation for penetration testing

CREST-accredited human-led penetration testing

Independent assurance for Pentesys’s human-led penetration-testing service.

Human-led testingView CREST listing

AI-enabled testing assurance

CREST accreditation for AI-enabled penetration testing

CREST-accredited AI-enabled penetration testing

Pentesys is one of the first ten companies worldwide accredited by CREST for AI-enabled penetration testing. The accreditation assesses the responsible, secure and appropriate use of AI in penetration-testing services.

AI-enabled penetration testingView CREST announcement

Responsible-AI commitment

CREST AI Charter signatory

CREST AI Charter signatory

A separate commitment to responsible, transparent and accountable use of AI in cybersecurity.

Responsible-AI commitmentAbout the CREST AI Charter

Secure organisational baseline

Cyber Essentials certified

Cyber Essentials certified

Pentesys is certified across the whole organisation against the Cyber Essentials baseline controls.

Recertification due 12 August 2027View Cyber Essentials certificate

Controlled AI programme access

Pentesys participates in Anthropic’s Cyber Verification Program (CVP) for Claude and has approved access to OpenAI Daybreak Blue.

Anthropic CVP for Claude

Participation in Anthropic’s Cyber Verification Program.

Programme participationAbout Anthropic CVP

OpenAI Daybreak Blue

Approved access for defensive cybersecurity work.

Approved programme accessAbout Daybreak Blue

CREST accreditation applies to Pentesys’s service delivery, not to an AI model, the Portal or an appliance.

Meet the expertise behind our testing