← Other Blogs

How to measure phishing behavior

To measure phishing behavior, combine the reporting rate, the click rate and the time to report across comparable simulations. Define for each indicator which people or messages count and how long you observe the response. An improvement in a simulation shows a change in that test, not real incidents avoided. The formulas in this guide are a measurement proposal for security teams, not a description of how any platform calculates its figures. The data in the examples is hypothetical.
Human risk management
Phishing

Delivery sets the denominator for every rate

Before calculating results, separate planned recipients from recipients with confirmed delivery. A blocked or bounced message does not give the same chance to respond as a delivered one. Confirmed delivery does not mean the person read the message.

In this guide, each campaign sends at most one simulated message to each recipient, and the denominator is the number of unique recipients with confirmed delivery. If a campaign sends several messages per person, decide first whether you measure people or messages and use the same unit in the numerator and the denominator. Do not add repeated interactions as if they were different people.

Delivery rate (%) = unique recipients with confirmed delivery ÷ unique planned recipients × 100

If you plan 500 recipients and delivery is confirmed for 400, the delivery rate is 80%. The click and reporting rates in this example are calculated on those 400 recipients, not on the 500 planned.

Record how delivery is confirmed, what happens with unknown deliveries and which events you exclude. Do not treat an unconfirmed delivery as a failed delivery or as an observed opportunity.

The reporting rate measures use of the reporting channel

Reporting rate (%) = unique recipients who reported the simulation within the observation window ÷ unique recipients with confirmed delivery × 100

If 80 of the 400 recipients report the message within the defined period, the rate is 20%. Count each person once even if they send several reports. The numerator must match the same population as the denominator.

Specify which channels count, how a report is linked to the campaign and when the window ends. If the log only captures the report button, a report by phone or chat may be left out. A low rate can reflect difficulty recognizing the simulation, a little-known channel or incomplete logging.

The UK NCSC recommends counting how many people report, not only how many click, so that reporting is treated as a success (NCSC, revised February 2024). As we explain in our analysis of phishing awareness training and reporting, the reporting rate and the time to report are operational measures. We read them together with delivery and the other campaign results.

The click rate does not capture the whole response

Click rate (%) = unique recipients with at least one recorded click within the observation window ÷ unique recipients with confirmed delivery × 100

If 60 of the 400 recipients click, the rate is 15%, and a person who clicks twice counts once. Agree on what counts as a click and how to exclude automatic clicks by security tools where possible. Technical logs do not always let you tell them apart with certainty.

Reporting and clicking are not mutually exclusive. A person can click and then report, so do not add both rates as if they described separate groups. If you need to know the order of actions, define it and verify that the log lets you reconstruct it.

Time to report needs a starting point

For each recipient who reports within the window, calculate the time between the confirmed delivery of their message and their first recorded report. Then summarize the times with the median, the middle value when the durations are sorted.

For example, if five people report 2, 5, 8, 30 and 60 minutes after delivery, the median is 8 minutes. Add the number of valid reports and the length of the window. If there are no reports, write "no data" instead of zero.

The median describes only those who reported and whose times could be linked to delivery. A shorter median does not imply a better overall response if the reporting rate or the log coverage changed. If you only have the send time, do not present it as the delivery time.

Compare campaigns under known conditions

Before interpreting a change, record the channel, the recipients, the exclusions, the delivery rate, the scenario, its difficulty, the run period and the observation window. Also check whether the reporting process, the security tools or the way actions are logged changed.

In the hypothetical example below, campaign A reaches a reporting rate of 20% and a click rate of 15%. Campaign B reaches 24% and 10%.

Hypothetical example Campaign A Campaign B
Planned recipients 500 500
Confirmed deliveries 400 450
Delivery rate 80% 90%
People who reported 80 108
Reporting rate on delivered 20% 24%
People who clicked 60 45
Click rate on delivered 15% 10%

In this example, reporting rises by 4 percentage points, clicks fall by 5 and delivery moves from 80% to 90%. These differences describe what was recorded, but they do not show that an intervention caused the improvement. The people, the message difficulty, the timing or the logging system may have changed.

To interpret the trend better:

  1. Keep definitions and windows constant. If they change, document the change and avoid presenting the rates as directly equivalent.
  2. Describe the population. The same cohort makes comparison easier, but absences, new joiners and different deliveries still matter.
  3. Record the difficulty. Note language, theme, suspicious cues, urgency and channel. Classify scenarios with criteria agreed before interpreting results.
  4. Show counts as well as percentages. With small samples, a few events move the rate a lot. Avoid firm conclusions from a single campaign.
  5. Use comparison groups and random assignment when feasible. Keep conditions as similar as possible and review the design before attributing a difference to training. A before-and-after comparison alone does not establish causality.

Use results to improve processes, not to label people

Aggregated results can guide a review of the reporting channel, the training content or the scenarios. Before breaking them down by team or function, check that the groups are large enough to reduce the risk of identification and that the intended use is authorized. Aggregation alone does not guarantee anonymity.

Avoid turning an isolated click into a permanent label or using the simulation as punishment, and define who can access identifiable data, for how long and for what purpose. Involve privacy, legal counsel and HR before introducing individual breakdowns.

NIST published its guide for cybersecurity and privacy learning programs in September 2024 (NIST SP 800-50 Rev. 1, September 2024). It proposes evaluating and improving the program with metrics, within an approach that encourages behavior change and a security and privacy culture. The operational definitions on this page are a practical proposal, not formulas prescribed by NIST.

What simulations can and cannot tell you

Simulations show responses observed in simulated scenarios and under the conditions of the campaign. A drop in clicks or a rise in reports helps you investigate what changed and decide what to test next. It does not show by itself that real attacks were avoided, that breaches fell or that a specific intervention caused an improvement.

Economic measurement raises another question, how to value the investment in a program, which our guide to security awareness ROI covers. To decide what support to offer based on these measures, use the signals matrix for prioritizing training. To combine them with real detection data, see the guide to the detection and training cycle.

How to evaluate a measurement platform

When you evaluate a platform, ask for a demo with a test campaign. Check how it defines delivery and reporting, which denominator it uses and who can access identifiable data.

At Zepo we connect threat detection, simulations and adaptive training. The definitions in this guide are a general proposal, and the metrics and denominators in dashboards may vary by product and configuration.

Frequently asked questions

How do you calculate the phishing reporting rate?In this guide, the rate is calculated by dividing the unique recipients who reported within the observation window by the unique recipients with confirmed delivery, and multiplying by 100. Each person counts once.

Which denominator should you use?To compare responses per person, we propose using unique recipients with confirmed delivery and counting in the numerator only people from that same group. If you measure messages instead of people, both parts of the formula must use messages. Document the unconfirmed deliveries.

What does time to report indicate?It indicates how much time passed between confirmed delivery and a person's first recorded report. The median summarizes the times of those who reported, not of the whole population.

Do simulations show that real attacks were avoided?No. Simulations show responses to simulated scenarios. Assessing real incidents or attributing changes to an intervention takes other evidence and an appropriate design.

Written By:
Zepo Intelligence
Subscribe to our newsletter
Blog content:
Act now before attackers do
Unify deepfake simulations, personalized training, and risk analytics into a single platform that builds measurable defense.
Talk to an expert

How Zepo helps companies

When everything connects, results follow

Paula Pereira

Digital Information Security Manager

“I would recommend Zepo to colleagues at other companies because I believe it has met all our needs. It has allowed us to run three types of campaigns that other tools we have tried simply cannot do. And beyond the product itself, the support from the whole team has helped us get far more out of it.”

+9K

Employees Protected

–10%

Click Rate on Attacks

+18%

Training Completion Rate

Ramon Fernandez Blanco

Cybersecurity & Digital Product Manager

“Since implementing Zepo, employee awareness has increased significantly. Employees now actively discuss cybersecurity and phishing campaigns, and suspicious emails are quickly reported instead of ignored.”

+600

Employees Protected

–15%

Credentials Submitted

+26%

Training Completion Rate

Jonathan Nelson

Director of Risk Intelligence

“Zepo’s vision for a real-time, hyper-personalised, multi-platform cybersecurity solution is truly unique and stands head and shoulders above the competition.”

+100

Employees Protected

Get Smarter Before Attackers Strike.