Sensitivity Analysis with Root Cause Analysis
Sensitivity Analysis Integrated with Root Cause Analysis · Also known as: SA-RCA, sensitivity-driven root cause analysis, parameter sensitivity with failure analysis, sensitivity-informed RCA
Sensitivity Analysis with Root Cause Analysis (SA-RCA) is an integrated engineering method that first quantifies how much each input parameter or process variable drives variability in a system output, then applies structured root cause analysis to the most influential factors to identify and eliminate the underlying failure mechanisms. The combination transforms numerical rankings of influence into actionable diagnoses, making it particularly effective in quality engineering, reliability analysis, and process improvement contexts.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use SA-RCA when a system or process exhibits unacceptable variability or failure rates and the number of candidate explanatory variables is large enough that intuition alone cannot reliably identify the dominant drivers. It is well-suited to manufacturing quality control, reliability engineering, chemical process optimization, and risk management in complex systems. The method requires quantitative output data and a model or dataset that links inputs to outputs. Do not use it when the system is too poorly understood to even enumerate candidate variables (pure discovery phase), when only a single variable is suspected (simpler hypothesis testing suffices), or when organizational or social factors are the primary causes — in those cases qualitative methods such as process audits or interview-based RCA should lead.
Strengths & limitations
- Focuses diagnostic effort on variables that mathematically matter most, preventing resource waste on low-influence factors.
- Combines quantitative rigor (sensitivity indices) with structured qualitative reasoning (RCA), making findings both credible and actionable.
- Global sensitivity methods (e.g., Sobol) handle nonlinear interactions and correlated inputs that one-at-a-time approaches miss.
- Produces a ranked, evidence-based priority list that supports transparent, defensible engineering decisions.
- Applicable across domains — manufacturing, reliability, simulation modeling, environmental engineering, and financial risk.
- Sensitivity analysis requires a functioning quantitative model or a sufficiently large dataset; it cannot rank variables when data are scarce or the system is poorly characterized.
- Global sensitivity methods can be computationally expensive, especially for high-dimensional systems with many input parameters.
- The method identifies which variables drive variance but does not by itself explain why — the RCA step requires domain expertise and time.
- Results are only as reliable as the model or data used; if the sensitivity analysis is run on a poorly specified model, the variable rankings will mislead rather than guide.
Frequently asked
What is the difference between local and global sensitivity analysis?
Local sensitivity analysis (one-at-a-time, OAT) varies one input at a time while holding all others fixed; it is fast but misses interaction effects. Global sensitivity analysis (Sobol indices, Morris method) simultaneously varies all inputs across their full ranges and correctly apportions output variance even when inputs interact. For realistic engineering systems with correlated or interacting variables, global methods are strongly preferred.
Does sensitivity analysis replace root cause analysis or complement it?
It complements rather than replaces. Sensitivity analysis tells you which variables matter most quantitatively, narrowing the search space for the RCA team. RCA then provides the causal explanation for why those variables are problematic. Without SA, RCA teams risk investigating the wrong variables; without RCA, SA results remain a ranking with no actionable fix.
How many input variables can SA-RCA handle?
There is no fixed limit. For OAT screening, even dozens of variables are feasible. For global methods, computational cost grows with the number of inputs — Sobol analysis may require thousands of model evaluations for high-dimensional problems, making surrogate models (e.g., polynomial chaos expansion, Gaussian process emulators) practical necessities for very large variable sets.
When should I use fault tree analysis instead of a fishbone diagram in the RCA step?
Use fault tree analysis (FTA) when you need to quantify failure probabilities and trace logic-gate combinations of failures — it integrates naturally with the quantitative output of sensitivity analysis. Use the fishbone (Ishikawa) diagram when you want a structured brainstorm of all possible cause categories (machine, method, material, man, environment, measurement) and a visual communication tool. FTA is more formal and data-intensive; fishbone is faster and more collaborative.
What software tools support this combined workflow?
For sensitivity analysis: SALib (Python), Sensitivity Analysis Toolbox (MATLAB), and commercial tools such as Simlab or Crystal Ball. For root cause analysis: dedicated RCA software (Causelink, TapRooT) or general diagramming tools. Many reliability engineering platforms (Reliasoft, Isograph) integrate both capabilities within a single environment.
Sources
- Saltelli, A., Ratto, M., Andres, T., Campolongo, F., Cariboni, J., Gatelli, D., Saisana, M., & Tarantola, S. (2008). Global Sensitivity Analysis: The Primer. John Wiley & Sons. ISBN: 978-0470059975
- Andersen, B., & Fagerhaug, T. (2006). Root Cause Analysis: Simplified Tools and Techniques (2nd ed.). ASQ Quality Press. ISBN: 978-0873896924
How to cite this page
ScholarGate. (2026, June 3). Sensitivity Analysis Integrated with Root Cause Analysis. ScholarGate. https://scholargate.app/en/experimental-design/sensitivity-analysis-with-root-cause-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Design of experimentsExperimental design↔ compare
- Failure Mode and Effects AnalysisExperimental design↔ compare
- Fault Tree AnalysisReliability↔ compare
- MONTE-CARLO-SIMULATIONDecision-making↔ compare