Yes, DORA metrics can help compare software delivery performance nationally, but they are not a stand-alone measure of software quality. A credible public benchmark would pair them with reliability, product quality, user outcomes and public value—and publish enough context for readers to judge whether comparisons are fair.
What DORA metrics measure—and what they do not
DORA’s Core Model separates software delivery performance from reliability. Its four delivery measures are change lead time, deployment frequency, change fail percentage and failed deployment recovery time. The model represents reliability separately through service level objectives (SLOs), and connects organizational capabilities and performance to wider outcomes, including well-being.
As an Amazon Associate I earn from qualifying purchases.
The four delivery measures
- Change lead time: how long a change takes to move through delivery.
- Deployment frequency: how often changes are deployed.
- Change fail percentage: how often changes result in a failure.
- Failed deployment recovery time: how long recovery takes after a failed deployment.
These measures describe aspects of delivery flow and change outcomes. They should not be collapsed into a single score and presented as comprehensive product quality. DORA describes its Core Model as a practitioner guide for continuous improvement, derived from its research program and updated more conservatively than individual research publications.
Recommended Free Tools
Reliability and broader quality need their own measures
SLOs provide a separate way to express reliability expectations. A national quality framework would also need measures appropriate to the products and services being assessed—for example, product quality, accessibility or security where relevant, user outcomes and public value. ISO/IEC 25020:2019 is one reference for the breadth involved: ISO says it can be applied to measurement models for system and software product quality, quality in use, data quality and IT service quality. Its scope is broader than delivery flow; it does not prescribe a national DORA implementation.
#1 Best Overall
Why a national league table would be easy to misread
DORA’s studies are not a national census
DORA’s 2025 report record describes more than 100 hours of qualitative data and survey responses from nearly 5,000 technology professionals around the world. Those study characteristics provide useful evidence about practitioners, but they do not make the respondents a nationally representative census of software producers.
DORA’s FAQ also cautions that annual reports can draw on different, potentially overlapping groups of respondents and may use new or updated questions and analyses. Changes between years can reflect changing real-world conditions, sampling or analytical methods; DORA says it cannot determine the cause with certainty. It does not release raw study data, citing privacy and confidentiality. A government building its own benchmark should therefore document its population, sampling frame, definitions, missing-data treatment, uncertainty and revisions rather than treating DORA report results as a ready-made national time series.
“Deployment” does not mean the same thing everywhere
For an organization operating a production service, deployment can be a meaningful event to count. For an open-source project that releases code to third parties but may not operate the resulting production service, the same wording may not fit. A 2023 study by Sánchez Ruiz and co-authors proposed adapting the measures for open-source projects: release frequency, lead time for released changes, time to repair code and bug issue rate. The authors applied their approach to Angular, Kubernetes, TensorFlow and VS Code. That work illustrates the need to adapt definitions to the population; it does not validate a nationwide benchmark across sectors.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsOther boundaries matter too: what counts as a deployment or a failure, who attributes a failure to a change, when recovery ends, what observation period is used, and whether results describe a team, service, organization or jurisdiction. Without shared answers, apparent differences may reflect measurement choices as much as performance.
Rank #3
What public-sector guidance demonstrates
The UK Health Security Agency Collaboration’s development standards page, published on 17 July 2025 and last updated on 3 September 2025, shows how DORA-style indicators can be operationalized for teams. It calls for measuring and reviewing at least every two weeks and gives the following traffic-light bands:
| Indicator | Green | Amber | Red |
|---|---|---|---|
| Deployment frequency | Daily or more | Weekly | Less than weekly |
| Lead time | Minutes or hours | One to three days | More than three days |
| Change failure rate | Less than 10% | 10–20% | More than 20% |
| Time to restore service | Less than one hour | One to four hours | More than four hours |
This is a practical public-sector example, not evidence that these bands are national norms or suitable for every service. Application criticality, delivery model and other context can affect what a useful target looks like. Teams should not turn the bands into cross-jurisdiction rankings without evidence that the underlying services and measurements are comparable.
How to design a useful national benchmark
- Set the purpose first. Internal improvement, public accountability, procurement oversight and national statistical comparison create different data needs and incentives. Decide which purpose the benchmark serves before choosing indicators.
- Define the population and unit of measurement. Specify whether the benchmark covers production services, software releases or another unit, as well as service criticality, deployment events, failure attribution, recovery endpoints, aggregation level and observation period. Adapt measures when the activity being measured does not involve operating a production service.
- Pair delivery measures with quality and outcomes. Report the DORA delivery indicators alongside reliability measures such as SLOs and relevant indicators of product quality, accessibility, security, user outcomes and public value. Select measures that fit the service rather than implying that one universal bundle captures every kind of quality.
- Pilot and document the method. Publish indicator definitions, data sources, missingness, uncertainty and revision history. Test whether teams can apply the definitions consistently before making public comparisons.
- Show distributions and context. Stratify results by meaningful characteristics and explain the limits of comparison. OECD guidance on sub-national benchmarking highlights comparability conditions and warns that composite indicators can mislead when misinterpreted. A dashboard with context is more defensible than an unexplained national score or league table.
- Use results to support learning. Treat indicators as prompts to investigate bottlenecks and outcomes, not as self-explanatory verdicts about a team or service. Avoid attaching high-stakes rewards or sanctions to raw ranks until comparability and resistance to gaming are established.
Interpret movement carefully, not as “higher is always better”
A benchmark needs to show trade-offs, not only direction of change. DORA’s 2024 report describes cases in which AI adoption can increase individual productivity, flow and job satisfaction while negatively affecting delivery stability and throughput. It also reports that internal developer platforms can improve productivity while reducing stability or throughput in some cases. A single headline score can conceal those differences and make a mixed result look like an unambiguous improvement or decline.
For policymakers, the useful question is not simply whether a metric rose or fell. It is whether the measurement is comparable, what changed in the delivery system, and whether reliability, product quality and user outcomes moved in a desirable direction as well. DORA measures can contribute to that assessment; they cannot answer all of it on their own.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
