Accelerate: State of DevOps 2019
https://dora.dev/research/2019/dora-report/2019-dora-accelerate-state-of-devops-report.pdf
Method
"This study employs a cross-sectional, theory-based design." ... "We use cluster analysis to identify our software delivery performance profiles... A solution using Ward's method was selected." ... "The structural equation models (SEM)... were tested using Partial Least Squares (PLS) analysis."
Population
"we used snowball sampling to obtain respondents... Our sample is likely limited to organizations and teams that are familiar with DevOps."
What it does not show
Every metric and every predictor is self-reported on a Likert scale about a single application; nothing is measured from production telemetry, so there is no independent check that reported deploy frequencies or failure rates match system behaviour. Cross-sectional, so the path coefficients are correlational despite being framed as predictive. Snowball sampling that the authors say is likely limited to organisations already familiar with DevOps.
Nicole Forsgren, Dustin Smith, Jez Humble, Jessie Frazelle
Cluster analysis separates self-described delivery performance into distinct profiles, and continuous delivery capabilities — trunk-based development, automated testing, deployment automation, loosely coupled architecture — predict cluster membership in a PLS-SEM model.
Tier II: Large-N cross-sectional survey with cluster analysis and structural equation modelling. Graded II for scale and analysis, but note the data are self-reported rather than production telemetry — see what it does not show.