Multi-Factor Risk Priority Scoring for Complex Systems

16 concepts12 papers

When engineering or managing complex systems—whether they are medical devices, deep learning pipelines, or global supply chains—achieving absolute zero risk is an impossible goal. Instead, as Per Ola Kristensson et al. argue in Human-Computer Interaction for AI Systems Design: Reflections on an Online Course on Human-AI Interaction for Professionals, the objective must shift toward establishing an acceptable level of risk through rigorous, systematic failure analysis.

But how do we quantify risk objectively? If we only measure how catastrophic a failure is, we might over-allocate resources to highly dramatic but virtually impossible scenarios. Conversely, if we only look at how often something happens, we might ignore rare but fatal black swan events.

To build a reliable safety framework, we must decompose risk into orthogonal dimensions. This is the core intuition behind Failure Mode and Effects Analysis (FMEA). By breaking down risk into distinct, measurable factors, FMEA allows operators to map a complex, multi-dimensional vulnerability space into a single, actionable metric: the Risk Priority Number (RPN).

Decomposing Risk: The RPN Formula

The classic FMEA framework calculates the RPN as a multiplicative product of three distinct factors:

$$\text{RPN} = \text{Severity} \times \text{Occurrence} \times \text{Detectability}$$

As detailed by Ilija Đjekić et al. in Scientific Challenges in Performing Life-Cycle Assessment in the Food Supply Chain, each factor captures a fundamentally different aspect of a failure mode:

  1. Severity ($S$): The magnitude of the negative consequence if the failure occurs.
  2. Occurrence ($O$): The likelihood or frequency of the failure happening.
  3. Detectability ($D$): The difficulty of catching the failure before its effects cause harm.

Why multiply these factors instead of adding them? Multiplication acts as a non-linear filter that penalizes balanced vulnerabilities. Consider two failure modes scored on a scale of 1 to 10:

  • Mode A (Extreme but obvious): Severity = 10, Occurrence = 1, Detectability = 1. $\text{RPN} = 10 \times 1 \times 1 = 10$.
  • Mode B (Moderate but stealthy): Severity = 4, Occurrence = 4, Detectability = 4. $\text{RPN} = 4 \times 4 \times 4 = 64$.

Even though Mode A features a maximum severity score, its overall risk profile is low because it is rare and easily intercepted. Mode B is far more dangerous because it represents a systemic blind spot: a moderately severe issue that happens regularly and is consistently missed. The multiplicative RPN naturally surfaces these stealthy risks.

For instance, Liangying Yi et al. demonstrated in Failure Mode and Effects Analysis in Reducing the Incidence of Sterilization Cycle Interruptions in Low-Temperature Hydrogen Peroxide Plasma Sterilizers that isolating these hard-to-detect process vulnerabilities allowed a hospital to slash equipment cycle interruptions from 5.48% to 0.87%.

Turning Scores into Actionable Thresholds

Once RPNs are calculated for every conceivable failure mode, they serve as an objective framework for resource allocation. Rather than debating which safety measure to implement based on intuition, teams can establish clear mathematical thresholds for intervention.

In a study on maternal healthcare protocols, Ying Liu et al. utilized a simplified two-factor RPN ($\text{Severity} \times \text{Occurrence}$) in Using healthcare failure mode and effect analysis as a method of vaginal birth after caesarean section management. By setting a strict threshold ($\text{RPN} \ge 8$), the authors classified specific clinical scenarios as high-risk, triggering mandatory, targeted remedial actions. This systematic thresholding moves accountability away from individual blame and toward proactive, systemic mitigation.

Similarly, Chaitanya Kalavagunta et al. used RPN comparisons in Is a weekly qualitative picket fence test sufficient? A proposed alternate EPID‐based weekly MLC QA program to evaluate entire quality assurance (QA) protocols in radiation therapy. By calculating the aggregate risk reduction across different QA schedules, the authors provided an objective, quantitative justification for replacing standard qualitative tests with more rigorous, automated imaging checks. This demonstrates that FMEA is not just for identifying single flaws; it is a tool for comparing the systemic efficacy of different operational modes, a benefit also highlighted by Long Han et al. in Research on the Effect of Business Management Training on Improving the Level of Enterprise Management.

[Identify Failure Modes] ──> [Score S, O, and D] ──> [Calculate RPN (S × O × D)]
                                                               │
    ┌───────────────────────── < ──────────────────────────────┤
    ▼                                                          ▼
[RPN < Threshold]                                      [RPN ≥ Threshold]
    │                                                          │
[Accept Risk / Monitor]                                [Mandatory Mitigation]

Analyzing Interconnections and Lifecycles

Modern systems are rarely static or isolated; their risks often emerge from the interfaces between components or across different stages of a lifecycle.

Consider dynamically interconnected medical devices in an operating room. Martin Leucker highlights in OR.NET - Approaches for Risk Analysis and Measures of Dynamically Interconnected Medical Devices that evaluating individual devices as standalone units is insufficient. When devices dynamically transfer or take over functions from one another, new interconnection-specific risks arise. The author addresses this by proposing a modular FMEA framework: individual devices are analyzed independently, followed by a separate FMEA focused entirely on the interconnections, using a gap analysis to catch emergent risks.

This lifecycle-wide application is also vital in software and artificial intelligence. In the context of algorithmic auditing, Ben Hutchinson et al. integrated FMEA directly into the early "mapping" stage of an end-to-end AI development framework in Closing the AI accountability gap: defining an end-to-end framework for internal algorithmic auditing. By forcing developers to brainstorm failure modes and their downstream consequences before deployment, the framework closes the accountability gap in AI systems design. This mirrors the qualitative Hazard Identification (HazID) methodologies used in broader digital health interventions, such as those explored by Stuart Harrison et al. in Improving safety claims in digital health interventions using the digital health assessment method.

Hybrid Models: Overcoming Subjectivity and Linear Limits

While the standard FMEA framework is powerful, it faces two major criticisms: the scoring of factors can be highly subjective, and a simple product formula cannot capture complex, cascading root causes. To address these limitations, researchers pair multi-factor scoring with complementary engineering tools.

To combat the inherent vagueness of human risk perception, Cai-qiao Huo et al. combined FMEA with Fuzzy Set theory in Research on the University Network Information Security Risk Management Model Based on the Fuzzy Sets. By mapping qualitative expert opinions onto trapezoidal fuzzy numbers, the authors quantified the uncertainty of risk factors in university network security, providing a more mathematically robust ranking of physical and communication vulnerabilities. Similarly, in supply chain management, Chih-Hung Hsu et al. paired the qualitative KJ method (for initial factor grouping) with FMEA in Enhancing Supply Chain Agility with Industry 4.0 Enablers to Mitigate Ripple Effects Based on Integrated QFD-MCDM: An Empirical Study of New Energy Materials Manufacturers to derive precise risk weights, helping manufacturers mitigate disruptive ripple effects.

Furthermore, while FMEA is excellent at identifying what might fail and prioritizing it, it does not inherently map how multiple small component failures combine to cause a system-wide disaster. For this, researchers combine FMEA with Fault Tree Analysis (FTA).

As shown by Byongug Jeong et al. in Quantitative risk assessment of fuel preparation room having high-pressure fuel gas supply system for LNG fuelled ship, FTA models the logical relationships (using AND/OR gates) between human errors and mechanical failures. By combining the top-down logical rigor of FTA with the bottom-up prioritization of FMEA, engineers can calculate exact system failure probabilities while maintaining a clear, multi-factor view of where to deploy preventative measures.

Key Foundational Work

OR.NET - Approaches for Risk Analysis and Measures of Dynamically Interconnected Medical Devices (2014) — contributed 4 concept elements to this direction.
Research on the University Network Information Security Risk Management Model Based on the Fuzzy Sets (2015) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_risk_priority_number_calculation_5ef6059a,
  title = {risk priority number calculation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/risk-priority-number-calculation/rdr_5ef6059a2a244515a95222e5fa846c11},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_risk_priority_number_calculation_5ef6059a_v1,
  title = {risk priority number calculation},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/risk-priority-number-calculation/rdr_5ef6059a2a244515a95222e5fa846c11/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}

See something wrong? Report a mistake