Modern data science training results in a paint by numbers approach not suited for the uniqueness and complexity of human medical data. Methodologies developed for maximising ad-clicks on large scale datasets don’t translate well to imperfect, or sparse data common in healthcare contexts. Seek assurance of methodological rigour and accuracy of a body of work or report. We have deep expertise in seeing through mistakes, bias, or misrepresentative data analyses. We can reanalyse or make recommendations to protect against spurious inference.