By using tdwi.org website you agree to our use of cookies as described in our cookie policy. Learn More

TDWI Upside - Where Data Means Business

Best Practices for Data Management with DataOps

To get the best out of data, follow these five DataOps implementation tips.

DataOps serves as the key that enables business enterprises to extract the maximum value from their data. With a large number of organizations striving to become data-driven and leveraging the competitive advantage data analytics brings, DataOps has become a key driver to unlock the ability to make data-backed decisions and drive real business growth.

For Further Reading:

Data Ops: What Is It? Do We Need It?

How to Build a DataOps Team: 3 Key Team Functions

Reducing Data Friction Just One of Many DataOps Benefits

Andy Palmer, generally credited to have coined the term, believes that at its core, DataOps aims to address the data-oriented needs of the modern agile enterprise.

What is DataOps and why is it needed?

"DataOps is DevOps principles applied to data analytics."

This statement is common whenever someone talks about DataOps. Unfortunately, although DataOps does aim to achieve the results that DevOps helped the software industry achieve, it is a gross oversimplification.

Automation in DevOps guarantees increased feature release velocity, improved quality, and enhanced scale at which software is developed and deployed. It solves many problems that traditional software development and deployment patterns weren't equipped to handle.

More accurately put, DataOps is the application of agile methodology, DevOps principles, and statistical process controls to data analytics. It lends an agile approach to data governance and data analytics development. It prescribes the design, implementation, and maintenance of a distributed data architecture to support production applications that use big data processing frameworks.

It also aims to break down the silos between data analytics teams, software development teams, and IT operations teams. The goal is to allow leading stakeholders in all teams to work with the entire data analytics team.

This last-mile availability of data encourages the democratization of data analytics within the organization. It is an opportunity to allow business teams to solve problems, answer questions, and find meaning in company data by leveraging their own unique perspectives. As the view of data expands beyond the data analytics teams, the quality of insights becomes more business-focused. The output and the efforts of the data team can be directly mapped to business success.

The introduction of industry-specific database engines (such as StreamBase, Vertica, and VoltDB) has also radically improved the performance and accessibility of large quantities of data at great velocities. Google also publicly released its BigTable database on Google Cloud in 2015.

These internal and market-governed factors encourage data analytics teams to adopt a new approach to data management. DataOps represents this new approach. It emphasizes collaboration, integration, and communication within data analytics teams and outside them.

Best Data Management Practices for DataOps

DataOps, despite being introduced in 2015, is still evolving. The focus is on identifying areas of concern for the data analytics teams, and ways to improve cross-team collaboration and to get rid of the silos. To get the best out of data, any DataOps implementation should follow these best practices.

Best Practice #1: Start small and build incrementally

The whole DataOps philosophy is inspired by agile methodology. You want quicker delivery of data and code, but you are not going to get it done all at once.

Building incrementally is the core principle of agile. Agile data processes focus on starting quickly with the data subsets and then focusing on incremental value delivery while incorporating feedback from the end users. The agile data mastering process needs to be incremental, automated, and collaborative to streamline seamless formation of data pipelines.

For Further Reading:

Data Ops: What Is It? Do We Need It?

How to Build a DataOps Team: 3 Key Team Functions

Reducing Data Friction Just One of Many DataOps Benefits

Insist on a cross-functional team structure to improve collaboration. Start with business representation within your data development team. The objective is to steer the function of the data analytics team towards business objectives. To kickstart this process, lay down business priorities for the data team and review them fortnightly or monthly.

Best Practice #2: Build operationally supportive apps

Data analytics teams usually source huge amounts of data that ends up being machine analyzed. Consider cases when these data sources can be directly mapped with operational teams that use insights from this data. Get your data developers to build apps that support a variety of internal operations.

These new apps must be treated and built like software development projects to make sure that data always stays updated. You need people within your data teams who can take data from its source, analyze it, and bring it to a point where internal teams can make use of it. They can then release these insights to the internal departments through a website or a downstream app.

Best Practice #3: Create business data glossaries and catalogs

A glossary attempts to answer various questions about the data itself. These are mostly data-defining questions such as the technical name, definition, and function of a particular type of data in different systems within the organization.

Catalogs are like supersets; they go beyond glossaries. They provide more metadata about the structure of the data. The creation of catalogs presents unique collaborative opportunities with teams that are the end consumers of data. Cataloging helps users understand deeper aspects of the data such as its locations, its users, and best practices for leveraging it.

This adds a layer of self-service to your data analytics team's functions. If someone wants to know more about data or do more with it, they can use data glossaries and catalogs without reliance on the data team.

Best Practice #4: Enable self-service mechanisms for using data

Teams tend to do their own data preparation when they have no data available for their specific use case. They self-source this data and use whatever tools they can find, internally or externally, to prepare it for their use case.

Such self-service data prep needs to be an organization-wide initiative to provide business users with the capabilities to explore, manipulate, and merge new data sources. Instead of undertaking data preparation as a tool for a single use case, access to data needs to be an enterprisewide cultural shift.

One of the primary problems of governance in DataOps is ensuring that data does not just end up being used. It must also ensure that people complete feedback loops and help improve data sources and analytics processes.

A proactive data analytics team goes beyond past and current use cases of data and predicts data needs for frequently and rarely utilized use cases. This can be achieved with the timely collaboration between the business teams and data teams.

Best Practice #5: Use automation to anticipate source changes and avoid downtime

One of the key problems that DataOps teams face is when a data source changes its format or becomes unavailable, affecting apps that use that data. This causes downtime because these apps are often not ready to handle the changes.

For enterprise DataOps teams, handling of source changes in the least disruptive way is non-negotiable. Downtime caused by one source change can disrupt multiple systems and affect multiple teams.

Smart DataOps systems have apps that can work with updating data sources. These changes are detected automatically and mechanisms are built in to safely propagate change information to affected apps with zero (or minimal) downtime.

Final Words

Although DataOps is no longer in the nascent stage, it is still being adopted gradually on a global scale. Automation is the key to democratizing data, but the responsibility to ensure an efficient and productive DataOps process lies in the hands of data analytics teams.

By providing helpful innovative solutions to democratize data and enable self-service analysis, data analytics teams have a huge role to play in the evolution of the practice of DataOps.

TDWI Membership

Accelerate Your Projects,
and Your Career

TDWI Members have access to exclusive research reports, publications, communities and training.

Individual, Student, & Team memberships available.