
DevOps Maturity Model: How to Assess Where Your Team Stands
A DevOps maturity model is a staged scale that shows how reliably a team turns code changes into working software in production. A DevOps assessment scores your practices against it. This guide gives a five-level model, the dimensions and DORA metrics to measure, and a six-step method for running the assessment.
- Zeeshan SiddiquiEnterprise Architect: AI Platforms
In this article
- What Is a DevOps Maturity Model?
- Why Run a DevOps Assessment?
- What Are the Five Levels of DevOps Maturity?
- Which Dimensions Should a DevOps Assessment Cover?
- Which Metrics Make a DevOps Assessment Objective?
- How Do You Run a DevOps Assessment Step by Step?
- What Should You Do After You Know Your Level?
- What Mistakes Make a DevOps Assessment Fail?
- When Should You Bring in Outside Help?
A DevOps maturity model is a structured scale that describes how far a team has progressed in building, testing, releasing and operating software, from manual, ad hoc work to automated, measured and continuously improving delivery. A DevOps assessment applies that scale to your own team: you score current practice against it, find the weakest areas and decide what to fix first.
This guide gives you a five-level model you can use today, the dimensions a DevOps assessment should cover, the delivery metrics that make the scoring objective, and a step-by-step way to run the assessment with your team. If you want an outside review instead, Sphere's DevOps and platform engineering service starts with exactly this kind of assessment.
What Is a DevOps Maturity Model?
A DevOps maturity model answers one question: how reliably and how quickly can this team turn a code change into working software in production? It does so by describing a series of stages, each with observable behaviors, so that teams can place themselves honestly instead of relying on opinion.
Maturity models are descriptive, not prescriptive. They do not tell you to adopt a particular tool or to reach the top level everywhere. A payments platform and an internal reporting tool need different levels of rigor, and a good assessment reflects that. The goal is a clear picture of where you are, which gaps hurt the business most, and what to do next. The same logic applies to information, as our comparison of data maturity models shows.
Why Run a DevOps Assessment?
Most engineering leaders already sense where delivery hurts: releases that need a weekend, outages that take hours to diagnose, or testing that happens only at the end. An assessment turns that sense into evidence you can share with finance, product and the board. Teams usually run one for four reasons:
- To find the real bottleneck. The slowest step is rarely the one people assume. Measuring each stage from commit to production shows where changes wait.
- To set priorities. Automation, testing, security and monitoring all compete for the same engineering time. A scored baseline shows which gap is most expensive to leave open.
- To justify investment with data. A baseline and a target let you show progress quarter by quarter instead of arguing from anecdotes.
- To prepare for an event. A funding round, an acquisition or a cloud migration puts delivery practices under review. Our technical debt assessment guide and cloud migration strategy cover two common triggers.
What Are the Five Levels of DevOps Maturity?
Models use different labels, but most describe the same progression. The table below uses five levels. Read each row as a description of typical behavior, then place your team at the highest level where most of the signs are true, not where one or two are.
| Level | Name | What delivery looks like | Typical signs |
|---|---|---|---|
| 1 | Initial | Builds and releases are manual and depend on a few people. | Release checklists in documents, environments that differ from each other, fixes made directly in production. |
| 2 | Managed | Version control and basic build automation exist, but each team works its own way. | A build server runs, tests are partly automated, deployments still need manual steps and approvals. |
| 3 | Defined | A shared pipeline moves code from commit to production with consistent steps. | Automated tests gate every change, infrastructure is defined as code, releases are routine and scheduled. |
| 4 | Measured | The team tracks delivery metrics and uses them to make decisions. | Lead time, deployment frequency and recovery time are reported, monitoring drives alerts, security checks run in the pipeline. |
| 5 | Optimizing | Improvement is continuous and owned by the teams doing the work. | Small changes ship many times a day, failures lead to blameless reviews, internal platforms remove repeated toil. |
Two cautions apply. First, levels are not a ranking of teams. Level 3 done well for a regulated core system can be the right target. Second, a team can sit at different levels for different dimensions, which is why the next section matters more than the single overall score.
Which Dimensions Should a DevOps Assessment Cover?
Score each dimension separately. A team with excellent pipelines and weak monitoring has a different problem from one with strong monitoring and manual deployments. The dimensions below map to the practices that the DevOps Research and Assessment (DORA) program lists in its capability catalog (opens in new tab), including continuous integration, continuous delivery, deployment automation, test automation, monitoring and observability, pervasive security, trunk-based development, loosely coupled teams and platform engineering.
| Dimension | What to ask | Evidence to collect |
|---|---|---|
| Source control and branching | Do all teams work from one main line, with small, frequent merges? | Branch age, merge frequency, size of the average change |
| Continuous integration | Does every commit trigger a build and automated tests? | Build duration, failure count, time to fix a broken build |
| Test automation | Can the team trust automated tests enough to release without a manual regression pass? | Coverage by layer (unit, API, end to end), flaky test count, manual test days per release |
| Deployment and release | Is deployment a repeatable, automated action that any trained engineer can run? | Deployment frequency, manual steps per release, rollback method |
| Infrastructure and environments | Are environments created from code and consistent from test to production? | Share of infrastructure defined as code, time to build a new environment |
| Monitoring and observability | Does the team find out about problems before customers do? | Alert quality, time to detect, access to logs and traces |
| Security in the pipeline | Are dependency, code and container checks automatic and enforced? | Scans per build, time to patch a known vulnerability |
| Team structure and culture | Can a team ship its own changes, and are failures treated as learning? | Handoffs between teams, change approval path, incident review practice |
| Platform and tooling | Do teams share paved, documented ways to build and run services? | Onboarding time for a new service, tool sprawl, internal documentation quality |
For the testing row, our article on automated testing as a component of agile DevOps explains how test layers fit into a pipeline. For the security row, the DevSecOps service describes how checks move earlier in delivery.
Which Metrics Make a DevOps Assessment Objective?
Self-scoring is useful but subjective. Delivery metrics add an objective check. The DORA program defines five software delivery performance metrics (opens in new tab) in two groups. Throughput covers how fast changes move, and instability covers how often they cause trouble:
| Group | Metric | Definition (DORA) |
|---|---|---|
| Throughput | Change lead time | The amount of time it takes for a change to go from committed to version control to deployed in production. |
| Throughput | Deployment frequency | The number of deployments over a given period or the time between deployments. |
| Throughput | Failed deployment recovery time | The time it takes to recover from a deployment that fails and requires immediate intervention. |
| Instability | Change fail rate | The ratio of deployments that require immediate intervention following a deployment. |
| Instability | Deployment rework rate | The ratio of deployments that are unplanned but happen as a result of an incident in production. |
DORA reports that the metrics are correlated for most teams, which means improving speed does not have to cost stability. That is the main reason to track both groups together: a team that only measures speed can hide a rising failure rate, and a team that only measures stability can drift into slow, risky big-bang releases.
You do not need a dedicated tool to start. Pull timestamps from your version control system and deployment logs, and record incidents from your on-call or ticketing system. Use your own trend as the first comparison: the question is whether lead time and recovery time improve quarter over quarter.
How Do You Run a DevOps Assessment Step by Step?
A useful assessment takes a few weeks of part-time effort from a small group, not a quarter-long project. The sequence below keeps it practical.
A DevOps assessment in six steps
- 11. Set scope and sponsorsChoose the products or teams in scope and name one executive sponsor. A single product line is a better first scope than the whole company.
- 22. Collect baseline dataGather the five delivery metrics from version control, deployment logs and incident records for the last three to six months.
- 33. Interview the people who shipTalk to engineers, QA, operations and security in separate groups. Ask them to walk through the last release and the last incident step by step.
- 44. Score each dimensionRate every dimension from the table above on the five-level scale, and write down the evidence behind each score so others can challenge it.
- 55. Find the constraintMap the path from commit to production and mark where changes wait. The biggest queue, manual approval or recurring failure is your first target.
- 66. Build a sequenced roadmapPick two or three improvements for the next quarter, assign owners and agree the metric that will show each one worked.
Interviews matter as much as data. Metrics tell you what is slow; the people doing the work tell you why. Ask what people avoid changing because it is risky and what they wait on most often; the answers usually point straight at the constraint.
What Should You Do After You Know Your Level?
The right next move depends on where the weakest dimension sits. Resist the urge to improve everything at once; each level mostly builds on the one before.
| If you are at | Focus first on | Why it comes first |
|---|---|---|
| Level 1: Initial | Version control for everything, a basic automated build and a written release procedure | You cannot automate or measure what is not repeatable |
| Level 2: Managed | A single shared pipeline, automated tests that gate merges and environments defined as code | Consistency across teams removes the biggest source of surprises |
| Level 3: Defined | Delivery metrics, monitoring and alerting, and security checks inside the pipeline | Measurement turns a working process into one you can improve |
| Level 4: Measured | Smaller changes, faster recovery and shared platform tooling that removes repeated work | Gains now come from reducing batch size and toil |
| Level 5: Optimizing | Sustaining the practice as teams, products and systems change | Maturity erodes when ownership and measurement lapse |
Legacy systems often cap what is possible, because tightly coupled code resists small, frequent releases. If that is your constraint, pair the DevOps roadmap with a modernization plan; our legacy system modernization framework shows how to sequence the two.
What Mistakes Make a DevOps Assessment Fail?
- Treating the score as the goal. A higher level is not a business outcome. Tie every improvement to a delivery metric or a customer-facing result.
- Assessing tools instead of practices. Owning a pipeline product does not mean the team delivers continuously. Ask what actually happens on a normal release day.
- Scoring in a room of managers. People closest to the work give the most accurate answers. Leaders alone tend to rate their teams one level too high.
- Ignoring culture. Handoffs, approval chains and blame after incidents slow delivery as much as missing automation. Our article on how DevOps speeds innovation covers the cultural side.
- Measuring once. A single snapshot goes stale. Repeat the metrics every quarter and rerun the full scoring once or twice a year.
- Copying another company's target. A team structure that works elsewhere may not fit your size or risk profile. See our guide to the ideal software development team structure for the trade-offs.
When Should You Bring in Outside Help?
You can run a first assessment with your own people, and for many teams that is the right start. An outside assessment helps in three situations: when the team is too close to the problems to score them fairly, when leadership needs an independent view for an investor or acquirer, and when the roadmap needs engineers who have built these pipelines before.
Sphere's DevOps and platform engineering service includes a DevOps assessment and strategy offering that evaluates current maturity and builds an improvement roadmap. Delivery follows four phases: assessment and planning, security and stability, automation and scalability, and continuous improvement. Related work includes cloud and infrastructure engineering and the DevSecOps practice. Scope and investment are defined with each client after an initial conversation about your systems and goals.
Frequently asked questions
Related posts



An understanding of data maturity can guide a company's transformation from being clueless about data to taking advantage of its full potential. In this post we compare different approaches to data maturity.

A technical debt assessment shows what deferred engineering work costs your business. Learn how to measure technical debt and how acquirers price it in M&A.