This paper examines what is actually known about failure rates in improvement and transformation initiatives, and why the headline '~70% failure claim' travels across domains. An integrative evidence synthesis maps failure-rate claims in academic and managerial literatures. Each claim is coded by failure type, unit of analysis, denominator, time horizon and rate form, and appraised for evidential robustness. Reported rates vary widely and are not comparable unless definitions, denominators and horizons are specified. Much academic repetition of '~70%' relies on secondary citation or underspecified claims rather than recalculated rates. Managerial benchmarks frequently define 'failure' as the complement of a stringent top-box success threshold and/or as non-sustainment over multi-year horizons, yielding higher apparent failure prevalence than narrower endpoints. Overall, the evidence does not support treating '70% failure' as a universal planning parameter; failure prevalence is best understood as a distribution conditional on what is counted and for how long. This study reframes the '70% debate' as a measurement and reporting problem, integrates academic and managerial evidence using a failure-type taxonomy and robustness appraisal, and proposes a 'Failure-Rate Reporting Minimum' to make future claims more falsifiable and practically useful.
Waal de (Thu,) studied this question.