# DevOps and TestOps in 2026: DORA data and challenges

URL: https://sdclabs.com/devops-testops/

> Only 16.2% of DORA 2025 respondents deploy on demand. Change failures, rework and flaky tests now limit delivery more than tooling does.

# Why do only 16% of teams deploy on demand? DevOps and TestOps in 2026

[Mykhailo Poliarush](https://sdclabs.com/author/poliarush/) 2026-03-04Updated 2026-11-175 min read

In 2026, almost every team has a pipeline, but few have fast, safe delivery. In the [DORA 2025 report](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) , only 16.2% of respondents deploy on demand, multiple times a day, while most ship weekly or less often and many changes still break production. The gap is no longer tooling. It is test reliability, platform quality and the way work flows from commit to customer.

## How often do teams deploy, and how fast do they recover?

DORA measures delivery with five metrics: lead time for changes, deployment frequency and failed deployment recovery time for throughput, and change fail rate and rework rate for instability. The 2025 survey ran from June 13 to July 21, 2025 and drew on nearly 5,000 technology professionals.

Deployment frequency is spread wide. The largest group in the [DORA 2025 data](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) deploys between once a week and once a month (31.5%), and 55.4% deploy less often than once a week, while only 22.7% deploy at least once a day.

How often organizations deploy to production

**On demand (multiple per day)**

16.2%

**Once per hour to once per day**

6.5%

**Once per day to once per week**

21.9%

**Once per week to once per month**

31.5%

**Once per month to once per six months**

20.3%

**Fewer than once per six months**

3.6%

Source: [DORA, State of AI-assisted Software Development 2025, Fig. 12 (nearly 5,000 respondents, June-July 2025)](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf)

Lead time follows the same shape. In the [same DORA survey](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) , 24.4% get a commit running in production in less than a day, while 31.9% need one day to one week and 28.3% need one week to one month.

Most teams recover within a day, but a long tail does not. In [DORA 2025](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) , 21.3% restore service in less than an hour after a failed change and 56.5% within a day, but about 15% take more than a week.

## How widespread are CI/CD and platform engineering?

Practice is close to universal. In the [CD Foundation's 2024 State of CI/CD report](https://cd.foundation/announcement/2024/04/16/state-of-ci-cd-report-press-release/) , 83% of developers said they were involved in DevOps-related activities, and CI/CD tool use was associated with better performance across all DORA metrics. That is the latest edition we could find.

Usage keeps growing. GitHub's [Octoverse 2025](https://github.blog/news-insights/octoverse/octoverse-a-new-developer-joins-github-every-second-as-ai-leads-typescript-to-1/) counted 11.5 billion GitHub Actions minutes in public projects in 2025, up 35% from 8.5 billion in 2024 (GitHub's own data, public projects only).

Maturity still separates teams. In the [CNCF Annual Cloud Native Survey](https://www.linuxfoundation.org/hubfs/Research%20Reports/CNCF_Annual_Survey_Report_1.15.26.pdf) (628 respondents, September 2025), 42% of organizations at the start of their cloud native journey have adopted CI/CD, against 91% of the most advanced "innovators". GitOps shows a sharper split: 0% of beginners and 58% of innovators run GitOps-compliant deployments, according to the [same CNCF report](https://www.linuxfoundation.org/hubfs/Research%20Reports/CNCF_Annual_Survey_Report_1.15.26.pdf) .

Platform engineering is now the default way to deliver this. [DORA 2025](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) found that 90% of organizations have adopted at least one internal platform, 76% have at least one dedicated platform team, and 29% already run several platforms.

## What are the main challenges for DevOps and TestOps teams?

**1. Changes still fail too often.** In [DORA 2025](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) , only 16.7% of respondents keep their change fail rate at 4% or lower, and 37.8% report that more than 16% of changes degrade service and need a hotfix, rollback or patch.

Change fail rate: share of changes that degrade service

**0-2% of changes**

8.5%

**2-4%**

8.1%

**4-8%**

19.6%

**8-16%**

26%

**16-32%**

19.5%

**32-64%**

12.5%

**Over 64%**

5.9%

Source: [DORA, State of AI-assisted Software Development 2025, Fig. 14 (nearly 5,000 respondents, June-July 2025)](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf)

**2. Unplanned work eats capacity.** Rework rate counts deployments made only to fix a user-facing bug. In the [DORA 2025 data](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) , 47.4% of respondents say more than 16% of their deployments in the last six months were this kind of unplanned fix.

**3. Flaky tests erode trust in the pipeline.** A red build that is not caused by the code teaches people to ignore red builds. The next section shows how big that problem gets.

**4. Tool sprawl.** The [CD Foundation report](https://cd.foundation/announcement/2024/04/16/state-of-ci-cd-report-press-release/) found that deployment performance is worse when teams use multiple CI/CD tools of the same form, likely because of interoperability problems. More platforms also mean more governance: DORA describes the leader's job as moving from "having a platform" to "governing a complex platform of platforms".

**5. Culture, not technology.** The top challenge in deploying containers in the [CNCF 2025 survey](https://www.linuxfoundation.org/hubfs/Research%20Reports/CNCF_Annual_Survey_Report_1.15.26.pdf) was "cultural changes with the development team" at 47%, ahead of lack of training and security at 36% each.

Speed without stability is also a risk with AI-assisted coding: [DORA 2025](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) found that AI adoption now improves delivery throughput but still increases delivery instability. More code arriving faster puts more weight on tests and pipelines.

## How much do flaky tests really cost?

A flaky test passes and fails on the same code. The best data comes from large engineering organizations that measured their own pipelines.

- **Google, 2016.** In a [Google Testing Blog post](https://testing.googleblog.com/2016/05/flaky-tests-at-google-and-how-we.html) , about 1.5% of all test runs reported a flaky result, almost 16% of tests had some level of flakiness, and about 84% of observed pass-to-fail transitions involved a flaky test. Google also wrote that new flakiness appeared about as fast as it was fixed.
- **GitHub, 2020.** In its monolith, [GitHub reported](https://github.blog/engineering/infrastructure/reducing-flaky-builds-by-18x/) that 1 in 11 commits (about 9%) had at least one red build caused by a flaky test. A system to detect and manage flaky tests cut that to 1 in 200 commits, an 18x improvement.
- **Atlassian, 2025.** The [Atlassian engineering team](https://www.atlassian.com/blog/atlassian-engineering/taming-test-flakiness-how-we-built-a-scalable-tool-to-detect-and-manage-flaky-tests) wrote that flakiness caused as much as 21% of master build failures in the Jira Frontend repository and about 15% of Jira backend failures, and estimated that the reruns waste over 150,000 developer hours a year.

These are single-company reports from vendors' engineering blogs, not industry averages. They agree on one point: flakiness does not go away by itself. It needs detection, ownership and a quarantine policy, which is what TestOps adds to DevOps.

## What does good DevOps and TestOps look like?

Good teams treat delivery and testing as one system:

- **Measure the five DORA metrics per service** , including rework rate, and review the trend every month. DORA's advice is continuous improvement for the same application over time, not a league table.
- **Make the pipeline trustworthy.** Track flakiness per test, quarantine flaky tests with an owner and a deadline, and keep the main branch green. GitHub's 18x drop came from managing flaky tests as data, not from retries alone.
- **Shift tests left and keep them fast.** Run unit and API tests on every change, and keep slow end-to-end suites focused on critical user journeys.
- **Treat the platform as a product.** [DORA 2025](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) calls high-quality platforms a force multiplier for organizational performance, productivity and team well-being, and reports that 76% of organizations already run a dedicated platform team. Standardize on one CI/CD path per kind of workload instead of several overlapping tools.
- **Map the value stream.** DORA calls value stream management, the practice of improving the flow of work from idea to customer, a force multiplier.

To start, pick one service. Baseline its deployment frequency, change fail rate and flaky-test rate, fix the worst bottleneck, then expand.

## How can SDCLabs help with DevOps and TestOps?

SDCLabs was founded in Krakow in 2011, and we are the founders of the open-source frameworks [Codeception](https://sdclabs.com/test-automation/codeception/) and [CodeceptJS](https://sdclabs.com/test-automation/codeceptjs/) . Our senior engineers set up CI/CD pipelines, test automation and the TestOps practices that keep them reliable, using AI-assisted workflows where AI agents draft and engineers review and decide what ships. Engagements start in 1 to 2 weeks.

If your team deploys less often than it wants, or no longer trusts its red builds, see our [DevOps and TestOps services](https://sdclabs.com/services/devops-and-testops/) or [contact us](https://sdclabs.com/contact/) . For the testing side, see [QA and test automation](https://sdclabs.com/services/qa-and-test-automation/) , and for AI across the lifecycle, read [AI in the SDLC: 2026 trends](https://sdclabs.com/ai-sdlc-trends-2026/) .

Sources checked on 2026-11-17.

## Frequently asked questions

### How often do most teams deploy to production in 2026?

In the [DORA 2025 report](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) , the largest group (31.5%) deploys between once a week and once a month, 55.4% deploy less often than once a week, and only 16.2% deploy on demand.

### What share of changes fail in production?

In [DORA 2025](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) , only 16.7% of respondents keep their change fail rate at 4% or lower, and 37.8% say more than 16% of changes degrade service and need remediation.

### How common are flaky tests?

At Google in 2016, about 1.5% of test runs were flaky and almost 16% of tests showed some flakiness ([Google Testing Blog](https://testing.googleblog.com/2016/05/flaky-tests-at-google-and-how-we.html) ). In 2025, [Atlassian](https://www.atlassian.com/blog/atlassian-engineering/taming-test-flakiness-how-we-built-a-scalable-tool-to-detect-and-manage-flaky-tests) estimated that flaky-test reruns waste over 150,000 developer hours a year.

### How widespread is platform engineering?

[DORA 2025](https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf) found that 90% of organizations have adopted at least one internal platform and 76% have at least one dedicated platform team.

[#devops](https://sdclabs.com/tag/devops/) [#testops](https://sdclabs.com/tag/testops/) [#ci-cd](https://sdclabs.com/tag/ci-cd/) [#flaky-tests](https://sdclabs.com/tag/flaky-tests/)

[Mykhailo PoliarushFounder and CEO](https://sdclabs.com/author/poliarush/)

## Related posts

[
### 48% of mobile sites pass Core Web Vitals, and that is the good news

Only 48% of mobile sites had good Core Web Vitals in 2025, and 95.9% of top home pages fail WCAG checks. Frameworks, speed, accessibility and AI in 2026.

Product engineering · 2026-10-19

](https://sdclabs.com/web-development-in-numbers/) [
### AI pays off only on a good platform

90% of organizations run an internal platform, and DORA finds AI lifts performance only when platform quality is high. The data, the challenges and how to start.

Product engineering · 2026-10-16

](https://sdclabs.com/software-platform-development/) [
### Most outages are preventable, and maintenance is where you prevent them

Outages are rarer but costlier in 2026: 57% of Uptime respondents say their last major outage cost over $100,000, and 87% say better processes could have avoided it.

Product engineering · 2026-10-15

](https://sdclabs.com/software-maintenance-and-support/)

## Get an estimate for your product.

[Get an estimate](https://sdclabs.com/contact/) [Book a call](https://sdclabs.com/contact/#call)
