Root cause analysis (RCA) is a critical problem in many real-world scenarios. RCA enables the identification of faulty or failing mechanisms in a system by comparing anomalous observations with corresponding reference (i.e., regular) observations. However, existing approaches rely either on heuristic methods or on conditional independence tests with a strong unconfoundedness assumption, and thus fail to exploit other complicated distributional constraints in the presence of latent variables. To relax these assumptions, we model the underlying system as a causal model and the anomalous system as a change in the structural functions of the same causal model. Specifically, to handle unobserved confounders, we establish an implicit connection between distributional constraint testing and root cause analysis. To adapt our approach to data generated from arbitrary causal models, we employ the deep causal model (DCM) framework, in which we design the causal model using neural networks. Finally, we illustrate how our method, RCA-DCM, can utilize different levels of partial graphical knowledge to perform RCA. We evaluate RCA-DCM against state-of-the-art baselines on simulated datasets, a physics-based causal chamber and two micro-service applications. RCA-DCM improves top-1 accuracy over the strongest baseline on both Sock Shop (0.880 vs. 0.752) and Online Boutique (0.776 vs. 0.712), and when the true root cause in the causal chamber is unobserved and acts as a latent confounder, it recovers the exact root-cause set more often than any competing method (perfect recovery rate (PRR) 0.846 vs. 0.731).
Figures & tables
Figure 1: Baselines fail when latent confounders are present and non-root-cause variables experience larger distributional shifts than the root causes; Given F→T , deciding whether Y is a root cause ( F→Y ) is equivalent to deciding whether F is a valid instrument for the effect of T on Y .
Figure 2: (a) No confounding. (b)–(c) Bow graphs with an unobserved confounder between X and Y (dashed ↔ ).
Figure 3: Workflow example. In the true augmented graph the fault shifts X and Z , so R∗={X,Z} , and latent confounders act on both X↔Z and Z↔Y . The induced pair P∗=(Pn∗,Pa∗) falls outside the shaded Valid-IV region of the input space: it violates the instrumental inequalities for the effect of X on Z , which rules out F being excluded from Z (exclusion condition) and hence forces the edge F→Z . Each candidate V is tested by freezing fV while the mechanisms on V∖{V} shift freely. The violation makes V∖{Z} infeasible, so ℓ∗(V∖{Z})>0 certifies that Z lies in every R∈SolRC(G∗) and hence in R . For V=Y the score attains zero, which is inconclusive: it is consistent with V∖{Y}∈SolRC(G∗) , and we draw no conclusion from it.
Figure 4
Figure 5: Perfect recovery rate (PRR) with 95% confidence intervals. (a) increasing latent confounding, (b) heterogeneous anomalies where a non-cause can out-shift a true cause, (c) real causal-chamber data with the root cause observed vs. hidden. RCA-DCM remains strong across all three settings; comparisons use a paired test, so overlapping intervals do not imply the absence of a difference.
Sock Shop
Online Boutique
Method
T1
T3
T5
T1
T3
T5
DCM
0.88
0.99
0.99
0.78
0.90
0.98
NSigma
0.75
0.97
0.99
0.70
0.91
0.96
BARO
0.74
0.98
0.99
0.62
0.90
0.94
CIRCA
0.65
0.96
0.99
0.52
0.89
0.98
RCD
0.38
0.54
0.58
0.49
0.59
0.62
Table 1: Average Top- k accuracy over the five fault types (best in bold).
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Ns
Na
DCM
BARO
RCD
#Instances
1000
1000
0.63
0.39
0.32
78
1000
500
0.58
0.48
0.48
95
500
500
0.47
0.47
0.31
51
1000
100
0.27
0.35
0.24
131
100
100
0.23
0.24
0.02
62
Appendix
Table 2: Perfect recovery rate (PRR) under varying normal ( Ns ) and anomalous ( Na ) sample sizes.
CPU
MEM
DISK
DELAY
LOSS
Average
Dataset
Method
T1
T3
T5
T1
T3
T5
T1
T3
T5
T1
T3
T5
T1
T3
T5
T1
T3
T5
Sock Shop
DCM
1.00
1.00
1.00
0.96
0.96
0.96
0.76
1.00
1.00
0.92
1.00
1.00
0.76
1.00
1.00
0.88
0.99
0.99
NSigma
1.00
1.00
1.00
0.96
0.96
0.96
0.64
0.96
1.00
0.60
0.92
1.00
0.56
1.00
1.00
0.75
0.97
0.99
BARO
1.00
1.00
1.00
0.48
0.96
0.96
0.68
0.96
1.00
0.92
1.00
1.00
0.64
1.00
1.00
0.74
0.98
0.99
CIRCA
0.96
1.00
1.00
0.56
0.96
0.96
0.68
0.84
1.00
0.48
1.00
1.00
0.56
1.00
1.00
0.65
0.96
0.99
RCD
0.76
0.92
0.96
0.28
0.48
0.48
0.32
0.52
0.52
0.28
0.48
0.52
0.28
0.32
0.44
0.38
0.54
0.58
Appendix
Table 3: Per-fault Top-1/Top-3/Top-5 accuracy. Best value per column within each dataset is in bold.
Method
Typical time / dataset
BARO
∼0.01 s
RCD
∼0.4 s (median)
DCM (ours)
∼41 s (median)
Appendix
Table 4: Per-dataset wall-clock time in the synthetic setting ( 12 variables, 4 latents, Ns=Na=1000 ).
Figure 6: Non-rc Y has higher shift in median than RC X .
Root cause analysis (RCA) in complex systems is challenging due to error propagation across multiple variables, the need for structural causal knowledge, and the computational cost of inference at test time. We introduce PRIM (Prior-fitted Root cause Identification with Meta-learning), a causal meta-learning approach that frames RCA as a Bayesian inference task over a synthetic prior of causal models. By marginalising out structural uncertainty, PRIM implicitly identifies changes in the data-generating mechanism between baseline and anomalous periods. In doing so, PRIM infers distributional differences without explicit statistical testing, and implicitly learns causal structure without model fitting at test time. Following the simulation-based meta-learning paradigm of prior-fitted networks, PRIM uses a Model-Averaged Causal Estimation (MACE) transformer neural process that jointly attends over observational and anomalous samples and the causal structure of nodes, enabling zero-shot inference in 17,ms for systems with up to 100 variables. Across synthetic benchmarks and two realistic benchmark datasets, PetShop and CausRCA, PRIM is competitive with methods that are aware of the system's causal graphical structure a priori while outperforming graph-unaware methods on several tasks. Lightweight fine-tuning to specific domains and data dynamics improves performance further.
Christopher Lohse, Anish Dhir, Amadou Ba +3
University of Dublin, Trinity College, Dublin, Ireland · IBM · Gatsby Computational Neuroscience Unit, University College London +2
Root-Cause Analysis (RCA) seeks to identify the variables responsible for abnormal system behavior in complex domains such as manufacturing, cloud computing, and healthcare. Existing approaches face a critical bottleneck: graph-based causal methods can identify intervention targets but typically require a known or accurately estimated causal graph, while graph-free statistical methods either localize marginal anomalies rather than structural causes, or rely on restrictive assumptions about graph structure or functional form. We propose StableRCA, a local mechanism-level RCA framework that avoids global graph discovery by estimating local Markov boundaries and detecting conditional distribution shifts within them. Leveraging the Independent Causal Mechanism principle, we show that intervention targets can be identified with probability converging exponentially in sample size under faithful Markov boundary recovery and non-degenerate mechanism shifts. Experiments on synthetic benchmarks and five real-world datasets demonstrate that StableRCA is robust to graph misspecification, effective under multiple intervention targets, scalable to large systems, and reliable across diverse application domains. Code is available at: https://anonymous.4open.science/r/StableRCA-E362
Finding the source of failures, known as Root Cause Analysis (RCA), is essential for identifying the root causes of anomalies and maintaining the reliability of complex systems. While causal theory has advanced data-driven RCA, existing frameworks assume causal sufficiency, failing to account for the unobserved latent variables prevalent in real-world environments. To address this gap, we propose PAG-RCA. This framework models system failures as parametric interventions over Partial Ancestral Graphs (PAGs) to perform RCA in the presence of latent variables. We use standard causal identification algorithms to find the source of failures by quantifying causal effects over the PAG. When an effect is identifiable, candidate root causes are ranked based on their exact intervention effects. When effects are structurally unidentifiable, our framework (for the first time in the RCA literature) integrates partial identification to evaluate and score candidates using analytical causal bounds. By integrating latent variables and partial identification at once our framework ensures robust RCA even under data scarcity and latent-variable scenarios where traditional methods degrade. Evaluations on synthetic data, microservice anomaly benchmarks and power-grid cascading failures demonstrate that PAG-RCA consistently outperforms state-of-the-art data-driven baselines. By improving data-driven RCA performance under data scarcity, this methodology advances reliable automated diagnostics in partially observable complex networks.
Henrique O. Caetano, Rafael Arone, Carlos Dias Maciel
Signal Processing Laboratory, São Carlos School of Engineering, University of São Paulo, São Paulo, Brazil · Guaratinguetá School of Engineering, São Paulo State University, Guaratinguetá, Brazil