← Back to blog

Social Engineering Assessments: Scope, Method and Reporting

November 22, 2026
Social Engineering Assessments: Scope, Method and Reporting

A social engineering assessment is an authorized test of whether your people and procedures hold up when someone tries to talk their way past your technical controls. Testers pose as a vendor, a colleague or a help desk caller, and send phishing emails, make phone calls or walk into the building. You get a documented answer to one question: what would a real attacker be able to get from your staff? The assessment pays off when it is scoped tightly, run under written rules and tied to fixes, and when the same provider that tests also helps you close the gaps. That is why we treat it as one part of a wider security program, not a one-off stunt.

AccountNext-Nexus
Bring Security Findings Into Focus
Nexus brings cybersecurity, IT management, and compliance together to help organizations address security gaps within a wider program.
Visit

TL;DR

  • A social engineering assessment simulates attacks on people (phishing, phone pretexting, text messages, physical entry) with written authorization, and reports what worked and what to fix.
  • Human behaviour is involved in roughly 60% of breaches, according to the 2025 Verizon Data Breach Investigations Report, so the test targets a real route in.
  • The rules of engagement (the signed document that defines what testers may and may not do) matter more than the clever pretext.
  • Report rate and time to report tell you more than click rate. Fixes and a retest turn the findings into lower risk.
  • Healthcare, payment-card and federal-style programs expect testing to be documented, so one provider covering test, fix and compliance evidence saves effort.

Table of Contents

What a social engineering assessment covers

Social engineering means manipulating a person into handing over information or taking an action that helps an attacker. An assessment does that deliberately, with permission, and measures the result.

The US government's testing guide, NIST SP 800-115, lists social engineering among the techniques used to validate vulnerabilities, with the example of posing as a help desk agent to ask for passwords. It also states that no single technique gives a complete picture, so organizations should combine techniques. For you, that means a social engineering test is one lens, not a full security verdict.

The usual channels are:

  • Phishing: emails that try to get a click, a login or an attachment opened.
  • Vishing: phone calls, often to the help desk, finance team or reception.
  • Smishing: text messages with links or requests.
  • Physical pretexting: someone entering a site by claiming to be a contractor, courier or new hire.

The help desk call deserves special attention. A confident caller who knows an employee's manager and start date can often talk an agent into a password or MFA reset, which is why practitioners point to callback-based identity verification as a fast-payback control, as the Strobes testing guide notes.

Assessment, testing or penetration test: is there a difference?

The three terms are used almost interchangeably in this field. "Social engineering penetration testing" usually stresses that testers go beyond measuring clicks and try to turn a success into access, such as a captured login used to reach a mailbox. Combined with technical exploitation and activity after the break-in, it becomes a full red team exercise. On its own it produces a phishing or pretexting assessment. The PCI Security Standards Council's penetration testing guidance treats social engineering as a component of penetration testing, not a separate exercise.

If you are choosing between test types, our guide to matching network, web and cloud penetration tests to risk shows where the human layer fits beside the technical ones.

Why the human layer is worth testing

The 2025 Verizon report found the human element involved in about 60% of breaches, roughly unchanged from the year before. It also found third-party involvement in breaches doubled from 15% to 30%. Both figures matter for a small business: you may have few staff, but each one holds keys to email, payments or patient records, and your vendors' staff hold some too.

Training alone does not tell you whether a Friday-afternoon request from a fake supplier would succeed in your finance team. A test does. Our own articles on common employee cybersecurity vulnerabilities and the employee cybersecurity training guide for HR leaders cover the behaviours that the test results will point to.

How the engagement runs, step by step

A well-run assessment follows phases. The FedRAMP penetration test guidance is a clear public model, and its structure is a good yardstick for any provider.

1. Authorization and scoping

Before anything happens, you sign rules of engagement. In the US this document is the legal authorization for the testers' activity, which would otherwise risk falling under the federal computer crime law, 18 U.S.C. §1030. FedRAMP asks that the acceptable social engineering pretexts be fully worked out before signing.

Agree on:

  • which channels are in scope (email only, or phone and physical too);
  • which groups are targeted (all staff, finance, help desk, executives);
  • who in your organization knows about the test, and who is the emergency contact;
  • how captured credentials and personal data are stored and destroyed.

For physical work, add a safe word, live emergency contacts and a letter the tester can show if challenged by security or police.

Where it goes wrong: vague scope. If nobody decided whether executives' personal accounts are off limits, you find out mid-test. Write the exclusions down.

2. Reconnaissance

Testers gather what an attacker could find: staff names and roles from public profiles, email formats, vendor relationships, office locations. This is what makes a pretext believable. It is also why the test result is more than a generic phishing score: it shows what your public footprint gives away.

3. Execution

Testers run the approved pretexts only. Hard exclusions belong in the rules: no threats, no real harm, and no exploiting personal crises such as layoffs or bereavement.

For phishing, FedRAMP's guidance requires that test emails be allow-listed on your security systems and delivered unmodified. The campaign includes a personalized email, a landing page, open tracking and capture of submitted credentials. Any data submitted counts as a failure, whether or not it was real. Allow-listing is a deliberate choice: you are testing people, not whether your filter happens to catch the message. Separately, understanding SPF, DKIM and DMARC setup helps you see why spoofed sender domains are blocked in real attacks, which is a control worth confirming alongside the human test.

4. Reporting

A useful report states scope, attack vectors, timeline, tests performed, findings with risk ratings and access paths, with sensitive data sanitized and passwords left out, as FedRAMP specifies. The US General Services Administration's penetration test exercise guidance likewise expects a summary of tests, findings, a severity rating and mitigation recommendations. If a report only says "22% clicked", it is not finished.

5. Remediation and retest

Findings only reduce risk when someone fixes them and the test is repeated. PCI DSS v4.0 requirement 11.4.4 requires retesting until exploitable vulnerabilities are resolved, according to this compliance summary. Even where no standard demands it, a retest is how you prove a fix worked.

Preparing without spoiling the result

NIST warns that testing is more intrusive than simply examining systems and can cause unexpected disruption, so you should decide up front how intrusive the techniques may be. For social engineering, the disruption is usually human: stress, embarrassment, a rushed call to the police.

A short preparation list:

  1. Name one senior sponsor who authorizes the test in writing.
  2. Limit the circle that knows about it, but make sure HR, legal and the help desk lead are inside it.
  3. Make sure staff already know how to report a suspicious message, and that the report button or mailbox works.
  4. Decide in advance how you will debrief everyone afterward.
  5. Agree that nobody is disciplined for failing. A test that punishes people teaches them to hide mistakes.

The same principles apply to ongoing simulations: avoid collecting real passwords, minimize data collected, skip malware and emotionally manipulative bait, and finish with a debrief and improvement plan, as practitioners such as KevinSec advise. Our piece on a phishing simulation program that shows results in 6 to 12 months turns that one-off test into a routine.

Reading the results: metrics that matter

Four phishing numbers cover most needs:

MetricHow it is calculatedWhat it tells you
Click rateUnique clicks divided by messages deliveredBasic susceptibility
Credential rateCredential submissions divided by messages deliveredReal exposure
Report rateReports to security divided by messages deliveredWhether people act as sensors
Mean time to reportMedian minutes from delivery to first reportHow fast your response can start

The Strobes guide argues report rate and time to report matter more than click rate, because resilience is detection and response, not a perfect score. Track them across repeated tests and by department so training goes where it is needed.

A high click rate also does not automatically mean careless staff. It can point to a missing reporting channel, confusing MFA prompts or a culture without verification habits. Fix those causes first. Phishing-resistant sign-in removes much of the prize from a stolen password, which is the focus of our 90-day phishing-resistant MFA plan, and our email security playbook covers the mail-side controls.

Compliance: what US organizations should know

Testing people is also evidence for auditors.

  • Healthcare: the HIPAA Security Rule, at §164.308(a)(8), requires periodic technical and non-technical evaluations of security controls. Penetration testing is a recognized way to meet that, and healthcare organizations commonly align it with NIST SP 800-115, scope it to systems that store or process electronic protected health information (ePHI), test at least annually and retest high-risk findings, according to an HIPAA testing overview. Your compliance officer or counsel should confirm what applies to your entity.
  • Payment cards: PCI guidance puts social engineering inside penetration testing, and findings are ranked by severity with the targets affected.
  • Federal-style programs: FedRAMP and GSA documents show the level of documentation that mature programs expect.

Keep the report, the rules of engagement and the retest record together. That file is what you hand an auditor.

What this means for different organizations

Small businesses have the most to gain from a narrow, well-defined test: email phishing plus a help desk call covers the most common routes in without a large bill.

Healthcare providers should scope to ePHI systems and the people who touch them, such as front desk, billing and the help desk, and keep evidence for HIPAA.

What this means for different organizations

Startups often have strong cloud controls and informal verification habits. A pretext that asks a founder or engineer for access is a realistic test. See our guide to penetration testing in startups.

Larger organizations benefit from comparing departments and sites, and from combining social engineering with technical testing so the chain from one phished login to a sensitive system is visible.

Our view: test the people, then fix the system around them

A social engineering result is only as useful as what happens next. We see the common failure as fragmentation: one vendor runs the test, another owns email security, a third manages the help desk, and a fourth handles compliance evidence. The finding lands in a gap between them.

Our approach is to keep these pieces under one umbrella. Penetration testing, vulnerability management, 24/7 threat detection, managed IT and compliance assessments for SOC 2, HIPAA, PCI-DSS and ISO 27001 sit with one provider, under one SLA and one point of accountability. When a test shows the help desk can be talked into a reset, the people who run your IT and your detection can change the procedure and watch for the follow-on behaviour.

Pricing is flat-rate and monthly, so budget conversations do not turn into surprise invoices. For a small team, that predictability is part of the security.

Nick - Sr. Executive

Getting an assessment that leads to fixes

If you want the test, the remediation and the compliance evidence in one place, our cybersecurity and compliance services cover penetration testing, incident response, 24/7 monitoring, managed IT and assessments for SOC 2, HIPAA, PCI-DSS and ISO 27001 under a single contract and SLA.

AccountNext-Nexus

To discuss scope for your organization, start at Nexus.

FAQ

How often should we run a social engineering assessment?

At least annually is the common pattern for compliance-driven programs, with lighter phishing simulations in between. Test again after major changes such as a new help desk process, a merger or a new MFA rollout. Repeating the same measures lets you see a trend.

Can employees be fired for failing a test?

That is a policy and legal question for your HR and counsel, but from a security standpoint it backfires. People who fear punishment stop reporting real incidents. Set a no-blame rule in the rules of engagement and use results to improve training and processes.

Do we need to tell staff a test is coming?

Not the details. A small circle, including HR, legal and the help desk lead, should know and sign off. Everyone else should already know how to report suspicious messages and calls, and be told in the debrief afterwards.

What is the difference between a phishing simulation and a social engineering penetration test?

A simulation mostly measures clicks and reports across many employees. A penetration test uses several channels, such as phone and physical access, and tries to turn a success into actual access to show impact. The rules of engagement and reporting are stricter for the second.

Sources