On-Demand Webinar · 60 min

How to Start a DataOps Transformation at Your Organization

Chris Bergh lays out a six-step program for starting DataOps at an organization: a long-term strategy, executive sponsorship, an assessment of readiness, a roadmap, how to organize the team, and how to make the transition. Recorded October 2020; updated August 2026.

Presented by Chris Bergh

What you'll learn 6 points
  • An enterprise DataOps transformation is six steps: Educate on the ideas, Find a first project that can show value, Establish a community of interest, Demonstrate real value in a month or two, Iterate onto more use cases, and Expand with a staffed Center of Excellence or Dojo that sets common infrastructure, tools, and metrics.
  • The case for transforming rests on four numbers: 87 percent of data science projects never get to production, 60 percent of all data analytic projects fail, 79 percent of data projects have too many errors, and the share of organizations describing themselves as data driven fell from 37 percent to 31 percent even as investment in analytics rose.
  • Transformation focuses on four areas at once: decreasing the cycle time of change so deployment is continuous, lowering error rates in production so customers trust the data, improving collaboration inside and between teams so there are fewer meetings and less bureaucracy, and measuring the process so the gains are visible.
  • Beware big bang approaches to change. Small teams working in a DataOps and Agile method come first, then teams are brought online one at a time, because demonstrable proof of success is what makes the next team want in. The soft parts, social proof, pushback, and cultural change, are not optional extras.
  • Budget about 15 percent of staff effort for DataOps activities at the start: testing, deployment, orchestration, test data management, and DevOps for data tools. In some organizations the DataOps engineer is a separate role, in others a shared function.
  • The DataKitchen DataOps Maturity Model scores six areas: error rates, cycle time, collaboration, measurement, culture, and the customer's need for rapid insight. A Level 1 team deploys new data and analytic code in months, develops in production, and runs no automated tests; a Level 5 team deploys in minutes, spins up a new development environment in minutes, and runs automated tests across every pipeline with an alert response plan. The score exists to provoke a discussion, not to shame and blame.

Slides

52 slides

Questions from this session

What are the six steps to a DataOps transformation?

Educate, Find, Establish, Demonstrate, Iterate, Expand. Educate through best-practice presentations, videos, books, and analyst writing. Find a first project by holding in-depth discussions with individual teams about where the pain is. Establish a community of interest with a wiki, Slack, and shared resources. Demonstrate value with a pilot that finishes in a month or two, iterate onto further use cases, then expand with a full-time Center of Excellence or Dojo.

How much staff time does DataOps require?

Plan for roughly 15 percent of staff effort at first. That effort goes to testing, deployment, orchestration, test data management, and DevOps for the data tools themselves. Some organizations create a separate DataOps engineer role; others make it a shared function across the existing team.

What is a DataOps Dojo?

A DataOps Dojo is a place where teams go for short, hands-on DataOps training. The name comes from the Japanese dojo, a safe environment for practicing new skills, and the model is borrowed from DevOps training dojos such as the one Target runs. The point is that teams gain practical experience without risking errors in the production environment.

What is the difference between a DataOps Center of Excellence and DataOps Technical Services?

DataOps Technical Services is a set of software services staffed by a central team that other teams use, in the same way many companies run a central DevOps CI/CD service for software pipelines. A DataOps Center of Excellence takes a large, deep-rooted organizational problem, solves it in a smaller scope, and then uses those small wins to scale the answer enterprise-wide. Both usually sit on the technical side, owned either by IT or by the CDO.

How do you gain executive sponsorship for DataOps?

Start with the problems the executive already notices: whether the analytic teams deliver the insight the business needs, whether business users mistrust the data or the team behind it, and whether a layer of consultants and shadow teams is doing the work. Then map the paths data takes from source to value and look for where errors happened, who owns each part, and how fast the team can respond.

How do you measure whether a DataOps transformation is working?

Measure two areas. Production metrics cover the pulse of the current production build, data provider errors and successes, test result history, timings, SLAs, and ML model metrics. Team and project productivity metrics cover collaboration, deployment frequency between environments, test coverage, and build times. You have to instrument both the production process and the development process to show the gains are real.

Where to go next