Verizon, one of the world’s largest telecommunications companies, serves over 140 million wireless retail connections and operates a vast data-rich infrastructure that spans across the U.S. and beyond. With petabytes of data flowing daily from devices, networks, and customer interactions, the company’s ability to derive meaningful insights at scale is critical to both its operational excellence and future readiness in AI.

Kalyani Sekar, Verizon’s Chief Data Officer, has spent over two decades at the company and now leads its enterprise-wide data and AI strategy. Sekar sits down with Yali Sassoon, Co-founder and CTO of Snowplow, to break down the evolving data landscape at Verizon, from embedding governance early, to aligning AI initiatives with ethical frameworks, and building observability into every step of the data lifecycle.

In this first installment of a three-part series, Sekar reflects on Verizon’s journey to develop high-quality, trustworthy data — a journey marked by hard-earned lessons, strategic realignments, and a growing focus on the “shift left” principle for data quality. She also unpacks how governance structures are being adapted to meet the rising demands of AI-powered decision-making.

Edited Excerpts

Q: As a data leader, how do you ensure high-quality data from the moment it’s collected, and why is that critical to Verizon’s long-term success?

Verizon is dealing with petabytes of data that come from various devices in our network, as well as the touchpoints customers leave across our channels. It’s very important that data quality is tracked right at the source and implemented using standardized procedures and frameworks, with strong observability built into every pipeline, from the point of collection all the way to the data warehouse and through to consumption.

It’s equally important to clearly define the roles and responsibilities of everyone involved with the data – those who generate it, those who manage it, those who govern it, and those who consume it. Each group needs to understand what “quality” means from their vantage point and ensure that standard is upheld as the data moves through their hands.

Data quality is foundational to Verizon’s long-term goals. As we continue to roll out more analytical use cases, from predictive and prescriptive to generative and agentic, it’s critical that these AI and analytical capabilities, including the reports built on them, are grounded in data that is trustworthy and of the highest quality.

Q: You now have a strong, trusted data foundation at Verizon to support a wide range of use cases. Was it always like this, or did you have to build it up over time?

It’s not like it all happened overnight. There were a lot of learnings that helped us mature this practice over time. Let me give you an example. When we first started thinking about data quality, our focus was mostly on the data once it reached the warehouse. That’s where we introduced a lot of quality rules to ensure everything was in order. But after some time, we realized that data quality is only as good as the quality of the data coming from the source.

That’s when it hit us, we really needed to shift left when it comes to data quality. We started thinking about it right from the source. So, now it’s about ensuring quality all the way from the source to the warehouse.

Another major learning came from the way data is consumed. Every use case that consumes data tends to have its own definition of what quality means. For example, metrics are one way to consume data. AI is another, and when AI consumes data, it gets integrated with systems. So when we think about the end use case, where data is used to drive insights and actions, we realized we need to define and manage data quality specifically for that consumption context too.

It’s been an evolution that’s taken shape over time. And if you ask me whether we’ve learned everything, I’d say, on a good day, maybe we’ve learned 95%. On a bad day, maybe 50 or 60%

Q: How does establishing data governance early improve your ability to use behavioral insights effectively, especially as you prepare for AI and future technologies?

Data governance is a critical function within the broader data and analytics organization, and it involves several key components. At a high level, there are multiple layers to this.

First, at the level of data definition, it’s important to clearly define what the standard sets of data are, what standards need to be applied to specific datasets, and how to classify data in terms of privacy, security, and related requirements. This includes establishing a clear taxonomy and definitions so that the data is holistic and understandable for everyone, this becomes the foundation.

The next level is about organizing the data. When we talk about data, we’re referring to millions of attributes. So, it’s important to organize it in a way that’s easy to manage, almost like maintaining a library. We often refer to this as organizing data into domains and sub-domains. For example, broader domains could include areas like customer, sales, or billing. Then, within each, you define sub-domains to bucket the data meaningfully.

Then comes the layer where you define the usefulness of the data. This includes clearly articulating what data reliability means, the types of quality rules required, what data completeness entails, and how traceability is ensured. This is still part of governance, focused on defining and organizing data effectively.

Next is data discovery. Once the data is defined and organized, the question becomes how to make it easily discoverable, like establishing a marketplace where people can retrieve and use data in a standardized way, ensuring the entire organization is leveraging high-quality, consistent data.

Finally, it’s about standardizing metrics and promoting the ethical use of data for any kind of consumption, whether it’s AI, analytics, or other applications. This is the last layer: From defining and organizing data to discovering and consuming it. It’s about ensuring that data is governed and maintained in a consistent manner so it can be used ethically across the enterprise.

Q: You’ve described the foundational building blocks Verizon has put in place to manage data. How does rapid growth of AI and its expanding use cases impact your approach to data governance? Does the same framework used for traditional analytics also support AI use cases, or does it require a different approach?

We’re a little different in the way we manage the data for AI use cases. When we talk about AI, we have a process of creating a registry of all the AI we would like to do. We call it the “AI Registry.” Every time someone wants to build a new model or modify an existing one, they clearly articulate what they want to do in business language, “This is what we want to do.”

Then, they move on to the next level, describing how they’re thinking about changing an existing capability. In the same registry, they also outline which data sets are being leveraged and how that data will be transformed during the process.

The final step is identifying which use case is going to use the transformed data, and how it will help in making the right decisions. You might have one set of data, but the same data can be used across multiple use cases to drive different types of decisions. So, all of those use cases are spelled out.

Once that’s done, the use cases go through a review with privacy, legal, and business teams. Everyone reviews them, provides their views, and only then does the use case move forward for implementation.

That way, we’re very particular about articulating the use case, the purpose, and how it’s going to drive value, making sure we’re not compromising anything from either the customer or the company perspective.

CDO Magazine appreciates Kalyani Sekar for sharing her insights with our global community.