DORA Metrics Benchmark Statistics (2026)

DORA's benchmarks are the most cited delivery-performance numbers in the industry, and they are also the most misquoted. This roundup collects the sourced figures from DORA's 2024 and 2025 reports, the earlier historical bands, and the secondary studies that test them, with the year and population behind each number. Where sources disagree, we say so.

Key takeaways

What DORA measures now

DORA's current model uses five software delivery metrics, split into two groups. Throughput covers change lead time, deployment frequency and failed deployment recovery time. Instability covers change fail rate and deployment rework rate. DORA's metrics guide describes this split, and DORA's history page explains how the metrics changed over time, including the 2023 rename of "time to restore service" to failed deployment recovery time.

• Five metrics in the current model: change lead time, deployment frequency, failed deployment recovery time, change fail rate and deployment rework rate (CIOPages on DORA's current guidance, 2026).
• DORA's guidance is to apply the five metrics to one application or service at a time and track it against its own history, not to rank teams against each other (CIOPages, 2026).
• The 2014 definition relied on three metrics, and the program added availability in 2018 (DORA, metrics history).

Performance benchmarks: the 2024 bands

The 2024 Accelerate State of DevOps report was published October 22, 2024, based on a survey of more than 39,000 professionals (InfoQ, 2024). It was the last edition to publish the elite, high, medium and low tiers. Infrazen's reproduction of the 2024 table lists the bands:

• Elite: change lead time under one day, on-demand deploys, change fail rate of about 5%, and recovery in under one hour.
• High: lead time between one day and one week, deploys between once per day and once per week, change fail rate of 20%.
• Medium: lead time between one week and one month, change fail rate of 10%, recovery under one day.
• Low: lead time between one and six months, deploys between once per month and once every six months, change fail rate of 40%, recovery between one week and one month.

Source: Libertify summary of DORA 2024, 2024. Share of respondents in each cluster, 2024 survey. Scale: percent of 100.

The distribution moved between 2023 and 2024. GetDX's summary of the 2024 report reports the high cluster falling from 31% to 22% and the low cluster rising from 17% to 25%. The elite cluster stayed roughly flat.

Source: GetDX, summary of DORA 2024, 2024. Scale: percent of 100.

• Medium-cluster teams showed a lower change fail rate (10%) than high-cluster teams (20%). DORA explained this as a trade of throughput for stability (re-cinq, DORA metrics, 2026).
• The 2024 report compared elite and low performers as 127x faster on lead time, 182x more deployments, 8x lower change failure rate and 2,293x faster failed deployment recovery (Libertify, 2024).

Conflicting and historical numbers

Several secondary articles circulate figures that do not match the 2024 report, and you should check any multiple before it goes into a steering deck.

• Historical: one 2026 guide cites elite teams deploying 973x more often and recovering 6,570x faster than low performers. These multiples come from earlier DORA research and do not match the 2024 ratios above, so treat them as historical (Gitmore, 2026).
• Conflict on recovery: one 2024 summary says elite teams recover "8 times quicker," which matches the change fail rate ratio, not the recovery ratio of 2,293x reported elsewhere (Hatica, 2025).
• Conflict on tier bands: one vendor guide lists low-tier change failure at 64% and high-tier at 10%, while the 2024 table shows 40% and 20% (Octopus, DORA metrics guide). Use the DORA-derived table, not the vendor version.

AI adoption and delivery outcomes

The 2025 report is the first DORA edition to organize findings around AI, and it drops the four-tier model in favor of team archetypes. Google Cloud reports the findings from nearly 5,000 technology professionals (Google Cloud, 2025). The report was published September 23, 2025 (Eltexsoft, 2026).

Source: Google Cloud, 2025 DORA report

• More than 80% of respondents believe AI has increased their productivity (Google Cloud, 2025).
• Of the 2025 respondents, 65% say they rely heavily on AI for software development, split into 37% moderate, 20% a lot and 8% a great deal (Google Cloud blog via BigDATAwire, 2025).
• The 2024 trust figure was 39.2%, so the 2025 share of 30% represents a modest improvement, not a turnaround (GetDX, 2024; Google Cloud, 2025).

Source: re-cinq, citing DORA 2024, 2026; Google Cloud, 2025. Scale: percent of 100.

Source: Jellyfish interview with DORA lead Nathen Harvey, 2025. Self-reported perception. Scale: percent of 100.

Source: Google Cloud blog via BigDATAwire, 2025. Scale: percent of 100.

• Speed and stability are moving in different directions. The 2025 report found AI adoption now associated with higher throughput, while the stability relationship stayed negative (LCMH analysis of DORA 2025, 2026; Splunk review of DORA 2025, 2025).
• In 2024, DORA estimated that every 25% rise in AI adoption was associated with a 1.5% drop in throughput and a 7.2% drop in stability (The New Stack on DORA 2024, 2024).
• The same 2024 model associated the same 25% adoption rise with 3.4% better code quality, 3.1% faster code review and 7.5% better documentation quality (cusy on DORA 2024, 2024).

Source: cusy on DORA 2024, 2024; The New Stack on DORA 2024, 2024. Estimated change per 25% rise in AI adoption, 2024 model. Bar length scaled to the largest absolute value (7.5%).

Vendor and independent studies on AI productivity

Developer self-reports and telemetry do not always agree, and the DORA team has said so. These studies are useful checks, not DORA benchmarks.

• An independent 2025 study cited in one analysis found developers believed AI made them about 20% more efficient, while measured task time was about 19% longer (Aviator on DORA 2025 and METR, 2025).
• Faros AI, a vendor, analyzed telemetry from more than 10,000 developers in July 2025. It reported 21% more tasks completed and 98% more pull requests merged, with organizational delivery metrics flat (Faros AI, vendor data, 2025).
• A 2025 Stack Overflow developer survey, as cited by one 2026 analysis, found 84% of developers use or plan to use AI tools, and 66% describe AI output as "almost right" (Unblocked citing Stack Overflow, 2025).

Team archetypes and the 2025 model

Instead of tiers, the 2025 report used cluster analysis to define seven team profiles built on performance, stability, well-being and friction (Google Cloud, 2025). The two strongest profiles are the most useful for benchmarking.

Source: LCMH analysis of DORA 2025, 2026; Efficiently Connected on DORA 2025, 2025; Splunk, 2025. Shares are approximate and reported by secondary sources.

Platform engineering as a precondition

The 2025 report links platform quality to whether AI produces organizational gains (Google Cloud AI Capabilities Model report, 2025). Adoption figures differ widely depending on how "platform" is defined.

Source: DORA platform engineering capability guidance, 2025

• Gartner predicted in October 2022 that 80% of software engineering organizations would have platform teams by 2026. DORA's 2025 figure differs in population and definition, so the two are not directly comparable (Eltexsoft, 2026).
• A separate survey, reported by Platform Engineering, found 55% of global organizations had adopted platform engineering, while a Futurum Group survey put operational adoption at 32% (Platform Engineering, 2026). The gap reflects different definitions of adoption.

What this means for your sign-off

DORA's numbers are useful for a release decision only when they are read against the right population and the right metric. These implications apply the findings above to SAP and ERP release programs.

• Measure one system at a time. DORA's own guidance is to track a single application or service against its own history. Apply that to each S/4HANA landscape rather than comparing your SAP program with an industry tier.
• Treat change fail rate as a regression signal. In the 2024 bands, a change fail rate near 5% marks elite performance and 40% marks low performance. Your regression escape rate per release is the closest equivalent, and it should be tracked with the same discipline. See what to automate first in S/4HANA regression.
• Use recovery time as a rollback test. DORA's elite band recovers in under an hour. For an ERP cutover, the equivalent question is whether your rollback and hypercare runbook can restore service within the window your business accepts. Check it against the template library's Go/No-Go and cutover runbooks.
• Put guardrails on AI-generated test assets. With 90% AI use and 30% of respondents expressing little or no trust in AI output, require review of AI-generated test scripts before they enter a sign-off pack. The AI-generated test review checklist covers this step.
• Check platform maturity before you add AI tools. The 2025 findings tie AI gains to a quality internal platform. For an SAP shop, that means a stable test environment, automated pipelines and clear ownership, which is the foundation covered in the SAP Testing CoE operating model.
• Be cautious with headline multiples. Several 2026 articles repeat ratios that come from older DORA research or misstate the 2024 data. Cite the year, cluster and metric for every number in a steering pack.

Frequently asked questions

What share of teams were elite DORA performers in 2024?

About 19% of respondents in the 2024 Accelerate State of DevOps survey fell into the elite cluster, with 22% high, 35% medium and 25% low (Libertify summary of DORA 2024, 2024).

Does AI improve DORA metrics?

It depends on the metric and the year. The 2024 model estimated that a 25% rise in AI adoption was associated with a 1.5% drop in throughput and a 7.2% drop in stability (The New Stack, 2024). By 2025, AI was associated with higher throughput, while the stability relationship stayed negative (LCMH, 2026).

How many professionals use AI at work?

DORA's 2025 survey of nearly 5,000 technology professionals found 90% use AI at work, and more than 80% believe it has increased their productivity (Google Cloud, 2025).

Did DORA replace the elite, high, medium and low tiers?

Yes. The 2025 report swapped the four tiers for seven team archetypes built through cluster analysis (RedMonk on DORA 2025, 2025). The 2024 tier table remains the last published version of the benchmarks.

How much of the industry has an internal developer platform?

DORA's 2025 data shows 90% of organizations use at least one internal developer platform, and 76% have a dedicated platform team (DORA, 2025). Other surveys report much lower adoption because they define platform engineering differently (Platform Engineering, 2026).

How large is the gap between elite and low performers?

The 2024 report put elite teams at 127x faster change lead time, 182x more deployments, 8x lower change failure rate and 2,293x faster failed deployment recovery than low performers (Libertify, 2024). Older multiples in circulation are historical and should not be mixed with these figures.

Sources

1. Announcing the 2024 DORA Report (Highlights from the 10th DORA report), Google Cloud, October 22, 2024
2. Announcing the 2025 DORA Report: State of AI-assisted Software Development, Google Cloud, September 2025
3. 2025 DORA State of AI-assisted Software Development report (PDF), Google Cloud, September 2025
4. 2025 DORA Report: AI Capabilities Model Report, Google Cloud, 2025
5. How are developers using AI? Inside Google's 2025 DORA report, Google, 2025
6. DORA's software delivery performance metrics, DORA, updated 2026
7. DORA metrics history, DORA, 2026
8. Platform engineering capability guidance, DORA, 2025
9. State of DevOps 2024: DORA Report Findings (interactive summary), Libertify, 2024
10. DORA Metrics Explained: The Five Measures and the 2024 Benchmarks, InfraZen, 2026
11. 2024 DORA report summary, GetDX, 2024
12. 2024 Accelerate State of DevOps Report Shows Pros and Cons of AI, InfoQ, November 2024
13. DORA 2024: AI and Platform Engineering Fall Short, The New Stack, 2024
14. DORA Report 2024: How AI is changing software development, cusy, 2024
15. Google Cloud's 2025 DORA Report Finds 90% of Developers Now Use AI in Daily Workflows, BigDATAwire, 2025
16. AI as Amplifier: Taking a Closer Look at the 2025 DORA Report, Jellyfish, 2025
17. DORA 2025: AI Amplifies Your Strengths (and Your Weaknesses), LCMH, 2026
18. DORA 2025: AI-Assisted Development and Team Performance, Efficiently Connected, 2025
19. State of DevOps 2025: Review of the DORA Report on AI Assisted Software Development, Splunk, 2025
20. DORA 2025: Measuring Software Delivery After AI, RedMonk, December 2025
21. DORA Metrics: All 5, With 2025 Benchmarks and AI, re-cinq, 2026
22. DORA has five software delivery metrics now, CIOPages, 2026
23. DORA Metrics Explained: 4 Key Metrics, Benchmarks and How to Improve (2026), Gitmore, 2026
24. The 2024 Accelerate State of DevOps Report: Hatica's Take, Hatica, 2025
25. Understanding the 4 DORA metrics and top findings from 2024/25 DORA report, Octopus Deploy, 2025
26. AI Won't Fix Broken Systems: Lessons from the 2025 DORA Report, Aviator, 2025
27. DORA Report 2025 Key Takeaways: AI Impact on Dev Metrics, Faros AI (vendor), 2025
28. DORA Metrics in the AI Era: Still Enough?, Unblocked, 2026
29. DevOps Statistics 2026: DORA Benchmarks and Adoption, Eltexsoft, 2026
30. What Is Platform Engineering? Inside the Discipline Reshaping Modern Software Delivery, Platform Engineering, 2026

Last updated: October 25, 2026. Statistics are refreshed as new research is published.