During the Beyond DORA: Metrics, AI and the future of platform engineering webinar, James mentioned research finding (something like) lowering MTTR does not lead to better reliability outcomes. Is it possible to get a link to that?

MTTR measures how fast a system can recover. Lowering it doesn’t change the failure rate. MTTR is important for uptime, but it’s not a measure of reliability.

James mentioned research findings that indicate meaningfully moving your Mean Time To Recovery (MTTR) does not actually improve your reliability outcomes (3449 3502). He emphasized that there is no correlation between MTTR and most reliability outcomes, suggesting that companies might be investing in AI tools to improve MTTR without seeing a corresponding improvement in overall reliability. This highlights the importance of having correct measures and understanding what success looks like before investing in solutions.

clipped video

It's worth noting that reliability is a separate set of metrics in the DORA research. It's always worth going deeper than the 4 keys as they have established other good measures, like reliability and well-being, that all modify the effect of software delivery performance on organizational outcomes.