This ebook explains how SecurityScorecard used machine learning to retune the weights of its 200+ measurement types so overall scores align more strongly with the relative likelihood of a publicly disclosed data breach. The study backtested 4.5 years of data (2019–mid 2023) using 16,583 breach cases and a size-matched cohort of 16,583 non-breached organizations, estimating breach timing with a 90-day offset and averaging measurements around the reference date. Results show breach likelihood rises consistently as grades worsen, with an “F” grade 13.8x more likely to be breached than an “A,” representing a 79% improvement over the prior methodology, supported by footprint-size normalization using z-scores.

Join for free to read