On this page
Many organisations buy a "penetration test" and receive a PDF export from an automated scanner. Scanners are useful, but they are not a penetration test. A real test is performed by a person who thinks like an attacker, chains weaknesses together and proves what an attacker could actually achieve.
Here is what you should expect, and what to ask for before you sign.
1. A scope you agreed on, in writing
A good engagement starts with a rules-of-engagement document that answers:
- What is in scope: specific domains, IP ranges, applications, APIs, mobile apps or cloud accounts.
- What is out of scope: production databases you cannot risk, third-party services you do not own.
- Test type: black-box (no information), grey-box (user accounts and some documentation) or white-box (source code and architecture access). Grey-box usually gives the best value for time spent.
- Testing windows and contacts: who to call if something critical is found or a system becomes unstable.
If a provider cannot explain the scope back to you in plain language, the report will not be useful either.
2. A recognised methodology
Ask which methodology the team follows. For web applications that normally means the OWASP Web Security Testing Guide and the OWASP Top 10; for APIs, the OWASP API Security Top 10; for infrastructure, frameworks such as PTES or NIST SP 800-115.
Methodology matters because it makes coverage repeatable. It is how you know that authentication, session management, access control, input handling and business logic were all examined, not just whatever the scanner happened to flag.
3. Manual testing, especially for access control and business logic
The most damaging findings we see are rarely exotic. They are things like:
- A user changing an ID in a URL and viewing another customer's invoice (IDOR / broken object-level authorisation).
- A "read-only" role that can still call an admin API endpoint directly.
- A discount or refund flow that can be replayed.
Here is the first one as a tester sees it. Logged in as customer 1042, we ask for someone else's invoice:
# Our own invoice: expected$ curl -s -H "Authorization: Bearer $TOKEN_1042" https://api.example.com/v1/invoices/88231 | jq '.customer_id'1042# Change one digit: another customer's invoice$ curl -s -H "Authorization: Bearer $TOKEN_1042" https://api.example.com/v1/invoices/88232 | jq '{customer_id, name, total}'{ "customer_id": 1187, "name": "A. Khan", "total": "PKR 482,500"}[!] HIGH: server returns any invoice to any logged-in user (no ownership check)4. Findings with evidence and real-world risk
Every finding in the report should include:
- A clear title and severity, ideally scored with CVSS and adjusted for your business context.
- Affected assets: exact URLs, parameters, hosts or endpoints.
- Reproduction steps and evidence: requests, responses and screenshots, so your developers can verify it themselves.
- Impact in business terms: "an unauthenticated attacker can download all customer records" is far more actionable than "Medium: insecure direct object reference".
- Specific remediation guidance, not just "validate input".
There should also be an executive summary a non-technical stakeholder can read in five minutes: overall risk, the top issues and what to fix first.
5. A retest
A penetration test is only valuable if the issues get fixed. A good provider includes a retest within an agreed window to confirm your fixes work and to update the report's status. That closed-out report is also what auditors and enterprise customers usually want to see.
6. Between tests: continuous visibility
A penetration test is a snapshot. The week after it ends, someone may deploy a new subdomain or open a port. That is why we pair periodic testing with continuous external attack surface management, so new exposure is spotted between engagements rather than at the next annual test. You can read more about that in our introduction to EASM.
Questions to ask any provider
Use this as a checklist when comparing proposals:
- Can I see a sanitised sample report?
- Who exactly will be testing, and what are their qualifications?
- How much of the test is manual?
- Is a retest included, and within what timeframe?
- How do you handle critical findings discovered mid-test?
If you would like answers to those questions from us, get in touch and we will scope a test around what matters most to your business.
Written by
Muaaz Ahmad
Penetration Tester, FutaqSec