Control frameworks tell you which controls to have, but not how much risk each one removes, so security budgets are argued on coverage percentages and maturity scores rather than on loss reduction. The FAIR Controls Analytics Model (FAIR-CAM), created by Jack Jones, the author of FAIR, and published by the FAIR Institute, describes how controls actually affect the frequency and magnitude of loss, individually and as a system. This guide explains its three control domains and their functions, how efficacy, reliability and dependencies are treated, how to value a single control against a scenario, and where FAIR-CAM fits alongside NIST CSF 2.0 and SP 800-53. It is written for risk analysts, security architects and CISOs who already use or are adopting FAIR and want to answer "which controls buy the most risk reduction?". For FAIR itself see the Cyber Risk Quantification Guide.
Key takeaways
- Frameworks are anatomy, FAIR-CAM is physiology. The FAIR Institute's overview uses this comparison: frameworks list controls, FAIR-CAM describes how they function and what their risk reduction is worth.
- Three domains, not one. Loss event controls act on loss directly; variance management controls keep those controls working; decision support controls improve the decisions that create and fund them.
- Operational efficacy is what counts. A control's value depends on how well it works when it is working, how well it works when it is degraded, and how often and for how long it is degraded.
- Dependencies change the arithmetic. Prevention functions combine like a Boolean OR; detection sub-functions and containment combine like an AND, so one weak link defeats the chain.
- Each function has its own unit. Percentages for avoidance or visibility, time for monitoring and containment; adding them together into one score hides the very thing you are trying to measure.
What FAIR-CAM is and what problem it solves
FAIR, standardised by The Open Group as the Risk Taxonomy (O-RT, version 3.0.1) and Risk Analysis (O-RA, version 2.0.1) standards, defines risk as the probable frequency and probable magnitude of future loss. Controls appear in FAIR mainly through resistance strength and, implicitly, through loss magnitude. That is enough to analyse a scenario but not to explain why a control matters, how its failure modes affect risk or how indirect controls such as patch management or security metrics earn their keep.
FAIR-CAM fills that gap. The FAIR Institute's overview treats a control as anything that directly or indirectly affects the frequency or magnitude of loss, and separates a control (the thing) from its control functions (the ways it affects loss). One control can serve several functions: awareness training, for example, can both improve personnel capability and reduce human error. The FAIR Institute's FAIR-CAM page describes the overview white paper as available to contributing members and lists mappings to NIST SP 800-53, CIS Controls, ISO 27000 and HITRUST, some marked as in progress. FAIR-CAM is presented as an extension of FAIR, not as a separate Open Group standard.
The FAIR Model and Open FAIR Standard Guide covers the underlying taxonomy; this page assumes you know loss event frequency, threat event frequency, vulnerability and loss magnitude.
The three control domains and their functions
| Domain | Function group | Functions | What it changes |
|---|---|---|---|
| Loss event control | Prevention | Avoidance, deterrence, resistance | Contact frequency, probability of action, probability that an action succeeds |
| Detection | Visibility, monitoring, recognition | Whether and how quickly an event is noticed | |
| Response | Containment, resilience, loss reduction | How long the threat keeps access, how fast operations recover, how much realised loss is offset | |
| Variance management | Prevention | Reducing change frequency and the probability that change creates variance | How often loss event controls fall out of their intended state |
| Identification | Threat intelligence, controls monitoring | How quickly shifts in threat capability or variant controls are found | |
| Correction | Treatment selection and prioritisation, implementation (for example patching) | How quickly variance is fixed | |
| Decision support | Misaligned decision prevention | Defining and communicating expectations, situational awareness, capability, incentives | How often decisions create unacceptable risk |
| Misaligned decision identification | Root cause analysis, post-mortems, auditing | How quickly poor decisions are found | |
| Situational awareness | Asset, threat and control data; analysis; reporting | Quality of the information decisions rely on |
Loss event controls
- Avoidance: reduces how often a threat comes into contact with the asset, for example removing internet exposure of a management interface.
- Deterrence: reduces the probability of illicit action after contact, for example visible monitoring and sanctions for insiders.
- Resistance: reduces the probability that an action succeeds, for example phishing-resistant authentication or a web application firewall.
- Visibility, monitoring, recognition: evidence must be captured, someone or something must review it, and normal must be told apart from abnormal. Logging without review delivers no detection.
- Containment, resilience, loss reduction: ending threat access, restoring operations and offsetting realised loss, for example through insurance.
Variance management and decision support
These domains act indirectly. Configuration management, vulnerability scanning and patching do not stop an attacker themselves; they keep resistance and detection controls in their intended state. Metrics, policy and governance do not touch an attack at all; they influence the decisions that determine which controls exist. The overview says their value is measured by observing downstream effects on loss event controls over time, which is why maturity scores alone cannot express it.
Efficacy, reliability and dependencies
FAIR-CAM distinguishes three forms of efficacy:
- Intended efficacy: how well the control performs when it operates as designed.
- Degraded efficacy: how well it performs when variant, for example a rule set left in monitor mode or an endpoint agent that stopped reporting.
- Operational efficacy: the blend of the two, weighted by variance frequency (how often the control becomes variant) and variance duration (how long it stays that way). Operational efficacy drives risk reduction value.
For a binary control (it either works or it does not), the overview states operational efficacy as intended efficacy multiplied by reliability. That has a practical consequence: improving reliability through variance management can deliver more risk reduction than buying a stronger control.
| Dependency type | Where it applies | Practical effect | Example |
|---|---|---|---|
| Boolean OR | Prevention functions | If any one operates as intended, the loss event is unlikely; layers add value | MFA plus conditional access plus device compliance on a remote access path |
| Boolean AND | Detection sub-functions and containment | If any one fails, detection or containment does not happen; the weakest element sets the result | Logs captured, but nobody monitors the queue at weekends |
| Control set | Controls that jointly serve one function | Value only exists when all members are present | Backups, tested recovery procedures and recovery infrastructure |
On coverage, the overview cautions that a control present on half of systems does not simply halve susceptibility, because attackers can probe several systems until they find an unprotected one. Model coverage gaps as their own exposure rather than scaling efficacy linearly.
Units of measure
The overview assigns each function its own unit and says units should not be combined across functions. A working set for analysis:
| Function | Unit | Typical data source |
|---|---|---|
| Avoidance | Percentage reduction in contact | External attack surface scans, exposure inventory |
| Resistance | Percentage of threat attempts resisted | Red team and breach and attack simulation results |
| Visibility | Percentage of relevant events captured | Log source coverage against the asset inventory |
| Monitoring | Time until evidence is reviewed | SOC queue metrics |
| Recognition | Percentage of malicious activity correctly identified | Detection testing, purple team exercises |
| Containment | Time to end threat access | Incident records |
| Resilience | Time to restore operations | Recovery tests |
| Variance (any control) | Variance frequency and duration | Configuration drift reports, agent health, patch latency |
Testing data for resistance and recognition comes most credibly from adversary emulation. See the Breach and Attack Simulation (BAS) Guide, and for variance data the Audit Evidence and Continuous Controls Monitoring Guide.
Valuing a control step by step
The overview sets out a sequence for measuring a control's risk reduction value. Applied in practice:
- Identify the relevant scenarios. A control has value only against named loss scenarios. Pull them from your library; see the Cyber Risk Scenario Library Guide.
- Measure current risk for each scenario with the control as it operates today.
- Determine how the control affects risk: which function or functions it serves, and therefore which FAIR factor it changes.
- Estimate intended efficacy from test data, vendor claims checked by testing, or calibrated estimates.
- Estimate degraded efficacy for its common variant states.
- Estimate variance frequency and duration from drift, health and change records.
- Account for dependencies: is it one of several OR-combined preventive layers, or an essential link in an AND chain?
- Account for other controls serving the same function, so value is not double counted.
- Re-run the analysis with and without the control, or with the proposed change, and report the difference in annualised loss and in the tail of the loss exceedance curve.
The overview's worked example values a web application firewall against a USD 10 million loss scenario, with probability falling from 10 percent to about 2 percent, an illustrative USD 800,000 risk reduction. It stresses that a full analysis needs software built for the model and that accuracy matters more than precision: represent uncertainty as ranges rather than pretending to exactness.
Decision guidance: when to use FAIR-CAM and how far to go
| Situation | Recommended depth | Why |
|---|---|---|
| No FAIR analysis yet | Do not start with FAIR-CAM; scope and quantify top scenarios first | Control value is relative to scenario risk |
| Quantified top scenarios, budget debate | Value the 5 to 10 controls with the largest spend or proposed change | Directly answers "what does this buy?" |
| Recurring control failures | Focus on variance frequency and duration | Reliability gains are often cheaper than new controls |
| Detection investment | Model visibility, monitoring and recognition as an AND chain | Shows where the weakest element limits the whole |
| Small organisation, no analyst | Use the function taxonomy qualitatively to find gaps | The taxonomy alone exposes "logs collected, never reviewed" style gaps |
Large organisations typically embed FAIR-CAM reasoning in a quantification platform; the Cyber Risk Quantification Platform Buyers Guide lists what to ask. Smaller teams can still map each control in their catalogue to one or more FAIR-CAM functions, which immediately shows functions with no control and controls with no clear function.
Running a control rationalisation review
A common first use is a rationalisation review before a budget cycle. Take the controls with the largest annual cost, map each to its functions and scenarios, and ask three questions. Does another control already deliver the same OR-combined preventive function on the same path, so that removing one changes little? Is the control an essential member of an AND chain or control set, so that cutting it removes the value of its partners as well? Is its current weakness efficacy or reliability? The answers usually sort spend into three groups: controls to keep and harden, controls whose reliability should be fixed before anything new is bought, and controls whose contribution is marginal and can be retired or consolidated. Record the reasoning against scenario IDs so the next review can test whether the predicted change in risk actually followed.
Where identity fits
Identity controls map cleanly onto FAIR-CAM functions, which makes them good early candidates. Phishing-resistant MFA and least privilege are resistance controls; removing unused service accounts and standing privileges is avoidance, because there is nothing left to contact; secrets scanning and identity threat detection are visibility and recognition; credential revocation and session termination are containment. Machine credentials show the variance problem clearly: a rotation policy is only as effective as the share of keys and tokens actually rotated on schedule, so variance frequency is measurable from the secrets inventory. The NHI Risk Scoring Guide provides the population data.
Framework mapping
| FAIR-CAM element | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 |
|---|---|---|
| Loss event prevention | PR (Protect) function | Control families such as AC, IA, SC |
| Detection functions | DE.CM-01 and the DE function | SI-4 System Monitoring, AU family |
| Response and resilience | RS.MI-01, PR.DS-11 | IR family, CP family |
| Variance management | ID.IM-02, GV.OV-03 | CA-7 Continuous Monitoring, CA-2 Control Assessments |
| Decision support | GV.RM-02, GV.OV-03 | PM-6 Measures of Performance, PM-9 Risk Management Strategy |
| Scenario risk measurement | ID.RA-05 | RA-3 Risk Assessment |
The overview itself warns that mapping is not straightforward: control descriptions are often ambiguous, broad controls map to several functions, ordinal scoring scales do not carry efficacy information and CSF functions are too high level to measure. Treat the table as orientation, not as a substitute for function-level analysis.
Detection and measurement
- Function coverage: percentage of top scenarios where every FAIR-CAM function in the relevant chain has at least one assigned control.
- Variance rate: variance frequency and mean duration for the 10 highest value controls, tracked monthly.
- Evidence age: share of efficacy estimates backed by a test less than 12 months old.
- Decision link: share of security investment decisions supported by a with-and-without control analysis.
- Forecast versus outcome: compare predicted control performance (for example containment time) with incident records each year.
Common failure patterns
- Treating FAIR-CAM as a scoring scheme: converting functions back into 1 to 5 maturity ratings throws away the units that make it useful.
- Adding detection percentages: visibility of 90 percent and recognition of 50 percent do not average to 70; in an AND chain the combined effect is lower than either.
- Ignoring variance: valuing controls at intended efficacy. A strong control that is variant a quarter of the time can be worth less than a modest one that rarely fails.
- Counting one control twice: crediting the same MFA deployment under several scenarios and functions without checking for overlap.
- Linear coverage assumptions: assuming 50 percent deployment equals 50 percent of the benefit, which the overview explicitly cautions against.
- Precision theatre: presenting a single-point dollar figure when the inputs are wide ranges. Report ranges and the assumptions behind them.
Practitioner checklist
- Top loss scenarios are scoped and quantified before any control is valued.
- Every catalogue control is mapped to one or more FAIR-CAM functions.
- Functions with no control, and controls with no function, are listed and reviewed.
- Intended and degraded efficacy estimates cite test evidence or calibrated estimates.
- Variance frequency and duration are measured for the highest value controls.
- Detection and containment are analysed as AND chains; prevention layers as OR.
- Coverage gaps are modelled as exposure, not as a linear reduction in efficacy.
- Each function keeps its own unit; no composite score is reported.
- Investment proposals show risk with and without the control, as ranges.
Standards and references
- FAIR Institute: FAIR Controls Analytics Model (FAIR-CAM)
- FAIR Institute: An Overview of FAIR-CAM
- The Open Group: Risk Taxonomy (O-RT) Standard
- The Open Group: Risk Analysis (O-RA) Standard
- NIST: IR 8286A Rev. 1, Identifying and Estimating Cybersecurity Risk for ERM
- NIST: Cybersecurity Framework 2.0
- NIST: SP 800-53 Rev. 5
Related NHI Mgmt Group resources: Cyber Risk Quantification Guide · FAIR Model and Open FAIR Standard Guide · Cyber Risk Scenario Library Guide · Key Risk Indicators for Cyber Risk Guide · Open FAIR Certification Guide