How to empower data teams in 3 critical ways

We need strong, collaborative data teams — not just to solve global problems like COVID-19, but to spur innovation. Consider these three steps leaders can take to help data teams have bigger impact
236 readers like this.
Two employees analyzing data and reports

More than ever, solving the world’s difficult problems requires the power of data. We’re seeing this firsthand as we witness data teams stepping up to mobilize communities during COVID-19.

To best use data, we need strong, collaborative data teams — not just to solve global problems, but to spur innovation. In retail, for example, we can now better tailor and secure shopping experiences; in financial services, we can make smarter, faster decisions that reduce risk; in oil and gas, we can unlock new efficiencies in discovery, extraction, and downstream delivery of energy; and in healthcare, we can leverage predictive modeling and patient tracking to accelerate clinical trials.

[ How can automation free up more staff time for innovation? Get the free eBook: Managing IT with Automation. ]

The days when data engineers and data scientists worked on different teams are ending. I believe the convergence of data teams will be similar to the convergence of development and operations teams and DevOps in software development – or before that, the rise of full-stack engineering teams. Opportunities abound for data teams to make a massive impact, but success will depend on streamlining efforts and empowering them in the right ways.

1. Recognize the limitations of siloed data teams

First, leaders must acknowledge the negative impact of siloed activity on efficient processes. As this McKinsey article suggests, consider this comparison: A single basketball player can positively impact a team’s performance, but teaching the entire team to work together will lead to better outcomes. Similarly, breaking down silos between data teams and opening space for more collaboration increases efficiency.

Siloed work on data teams can not only create internal friction, but it can also reduce the effectiveness of technology. Take AI as an example: Forty percent of organizations making significant investments in AI don’t report business gains from it, and siloed data teams can further impede harnessing the true power of AI. Siloed teams add friction in iterative model development processes, slowing the development of AI tools. Changes to data pipelines can also take months if they require efforts from multiple disjointed teams. Teams must collaborate cross-functionally to agree on data definitions or metrics and high-quality analytics for AI projects that require combining multiple datasets.

We need to break silos down, enable collaboration between data engineering and data science teams, and build a new data team structure.

We need to break these silos down, enable collaboration between data engineering and data science teams, and build a new data team structure.

2. Make room for new roles

I believe data teams will become more vertically focused on business problems and we’ll continue to see more hybrid roles. These new roles include machine learning engineers who can manage an AI application from data preparation to production, or data scientists or engineers with “full-stack” data experience.

Leaders should also consider investing in senior talent roles that encourage collaboration over siloes. For example, we’ve seen a rise in the Chief AI Officer role across the industry, but having a separate AI organization can be ineffective as it assumes that AI is separate from your company’s data strategy. AI is very much dependent on data strategy, and companies should first invest in Chief Data (and Analytics) Officers who can coordinate these projects end-to-end.

Although a unified team structure will not be the right choice in all situations, this type of convergence has happened in computing many times before: Most recently, cloud services combined with tooling for DevOps and frameworks such as Node.js have enabled full-stack web application development.

Ultimately, leaders should choose a team structure that lets them iterate on business problems and deliver value most quickly, and data professionals will need to suggest the practices that let them deliver this value to their organizations.

3. Translate the broader business value

No organizational data strategy is complete without buy-in from business leaders across all facets of the company. Beyond new roles within data teams already mentioned that encourage collaboration, McKinsey suggests that the role of the “data translator” — someone who can bridge the gap between frontline managers and data scientists — will become more important. As this MIT Sloan Management article explains, translators are critical to breaking down complex information for executive stakeholders as well as educating and training employees across the company on the power of data.

By communicating with the C-suite and other stakeholders, translators can also help ensure every data project ties to a broader corporate goal. This is especially critical during a crisis or challenging economic period when funding is limited, and quantifying business value is even more essential.

As we face the global and economic implications of COVID-19, data teams have never been more essential to solving issues across the biotech, government, and financial sectors. They can make a critical impact, but only if we enable and equip them to tackle these challenges.

[ Get exercises and approaches that make disparate teams stronger. Read the digital transformation ebook: Transformation Takes Practice. ]

Matei Zaharia is an Assistant Professor of Computer Science at Stanford University and Chief Technologist at Databricks. He started the Apache Spark project during his PhD at UC Berkeley in 2009, and has worked broadly in datacenter systems, co-starting the Apache Mesos project and contributing as a committer on Apache Hadoop.