Failure-Margin Analysis (FMA): A protocol for complexity-informed synthesis that prioritizes the failure margin over central tendency in the study of complex adaptive systems

A

BSTRACT

Existing synthesis methodologies like systematic review, meta-analysis, realist synthesis, and critical interpretive synthesis share a structural tendency to orient analysis toward central tendency; the pooled effect, the typical outcome, the average programmed performance. In complex adaptive systems (CAS), this orientation is epistemologically problematic. The mean corresponds to no real system state as the system is always moving within its distribution. The failure margin, the conditions under that interventions reliably do not work is where causal system structure is most legible. No named, standardized methodology currently exists that takes this position as its explicit epistemological foundation. The paper proposes and operationalizes failure-margin analysis (FMA), a replicable protocol for complexity-informed synthesis that deliberately prioritizes the failure margin over central tendency. FMA comprises six mandatory decision nodes. System boundary definition, failure operationalization, failure-margin evidence assembly, complexity framework selection, feedback loop mapping, and leverage point identification and tier assignment. The paper describes the epistemological basis of each node, specify quality criteria, and identify common errors. Leverage point tier assignment uses the Meadows hierarchy of system intervention. FMA is distinguished from existing synthesis methods by its constitutive use of complexity frameworks as analytical instruments, its explicit exclusion of aggregated central-tendency data as primary evidence, and its output form, a failure typology coupled with a leverage point map rather than a pooled estimate or programmed theory. The method is demonstrated through a worked example in tuberculosis control, where six interventions are analyzed using the FMA protocol. FMA offers a standardized, replicable methodology for researchers who need to understand why complex interventions fail under real-world conditions, without generating new primary datasets or conducting field studies. It is applicable wherever evidence of intervention failure exists, but the structural reasons for that failure remain analytically invisible under conventional synthesis approaches.