Google Cloud Expert on Strategic Use of DORA Metrics
In a recent interview, Google Cloud’s Nathen Harvey explained that DORA metrics should be used as strategic alignment tools, not just operational KPIs. He advised leaders to focus on their own trendlines rather than industry benchmarks to show progress and to pair quantitative DORA data with qualitative developer experience (DX) surveys. Harvey also predicted that the next generation of engineering leaders will supplement DORA with custom, domain-specific metrics tied to business outcomes.
- The DORA (DevOps Research and Assessment) initiative was co-founded in 2014 by researchers Nicole Forsgren, Jez Humble, and Gene Kim to scientifically measure software delivery effectiveness; Google acquired the program in 2018. Their research found that high performance in software delivery is a predictor of organizational performance, with top-performing teams being twice as likely to meet or exceed their business goals. - The four original DORA metrics are grouped into two categories: velocity and stability. Velocity is measured by Deployment Frequency (how often code is deployed) and Lead Time for Changes (time from commit to production), while stability is measured by Change Failure Rate (percentage of deployments causing failures) and Time to Restore Service (how quickly service recovers). - DORA research classifies teams into performance tiers from "Low" to "Elite." Elite performers deploy multiple times per day, have a change failure rate between 0-15%, and can restore service in under an hour. - Complementing DORA's quantitative systems data, Developer Experience (DX) metrics focus on qualitative feedback from developers. DX frameworks measure factors like cognitive load, flow state, and feedback loops, often through surveys, to understand the friction developers face in their daily work. - The DORA framework itself is evolving with the industry. In 2024, a new metric called "Rework Rate" was introduced, and "Mean Time to Recovery" was reclassified as "Failed Deployment Recovery Time" and moved from a stability metric to a throughput metric. - The annual research report, now called "The State of AI-Assisted Software Development," reflects the growing impact of AI. The 2025 report found that while AI adoption correlates with faster code shipment, it can also lead to higher instability and increased rework, amplifying a team's existing strengths or weaknesses. - In response to AI's impact, the DORA team introduced the AI Capabilities Model, which outlines seven foundational practices for effective AI adoption. This model emphasizes that success with AI depends on maturing existing DevOps principles rather than creating a separate AI-specific strategy.
