Data Quality as a Forcing Function: Breakthrough Conversation with IIA Expert Liz Marsh

In this blog series, Jason Larson, head of content at IIA, sits down with IIA experts who serve as sparring partners for Research and Advisory Network clients. IIA’s RAN expert community, with over 150 active practitioners and unbiased industry experts, is dedicated to advising data and analytics leaders on key challenges unique to their enterprise.

In the digital age, the quality of data can make or break an organization. Liz Marsh, who leads master data management and analytics at Eaton, understands the stakes. Her role involves navigating the complex landscape of data integrity, ensuring that the information her company relies on is accurate and reliable.

In this conversation, Marsh discusses the critical nature of data quality, the common obstacles organizations face in preserving it, and the evolving trends that are shaping the future of data management. This discussion has been edited for length and clarity.

Liz, I’m excited to catch up with you and talk some data quality. As many companies dip their toes into enterprise AI, there’s little debate that the data foundation of these enterprises is one of the most important factors in advanced analytics and AI success. Before we get too far ahead of ourselves, let’s just do a quick level-set on why, exactly, data quality is such a critical issue for organizations on their analytics and AI journey?

Liz Marsh: It’s great to be here! Well, the obvious answer is analytics and AI are really hot topics across organizations right now—and it’s really hard to have a serious conversation about AI without talking about the integrity of your data. But what's not always so obvious is that people don't realize to create good analytics and to create good AI tools, humans are still needed to do that initial coding. You also need a strong underlying infrastructure that makes sense of what the data is trying to communicate so that humans know what to code and analyze. If you have bad data quality, your analytics and AI solutions won't be as successful. We know there's ways to train the tool over time, but it's really hard to quantify the mistakes made along the way and what the impact of these mistakes are until you can trust your dataset.

You’re a seasoned practitioner and I’m sure you have many war stories around this. What are some of the biggest challenges you've faced in maintaining data quality?

Liz Marsh: A big challenge is unclear or unaligned definitions. People from different backgrounds might use the same terms but mean different things, making conversations difficult. Clear documentation is a must for alignment.

Another issue is limited access to subject matter experts. A lot of data quality understanding starts with human intuition before it can be programmed into a data catalog or checked programmatically. At the end of the day, a lot of times it’s the humans that understand what the data means and whether or not it’s high quality. Most data have subjectivity woven into the fabric of their DNA.

Then there's the challenge of a lacking data-driven culture. I know that term can feel nebulous sometimes, but what I mean is that many organizations haven't invested in the talent needed to make data management a full-time job. You need data owners who wake up every day whose job it is to make sure their data domain is healthy. Otherwise, data quality becomes a semi-regular exercise when it’s time to update reports for the executive team, and then it’s a mad scramble to make sure the data is good underneath the covers. This is all preventable but only if the company invests in it.

Also, there's what I call an ego-based fear around documenting someone's role or the information that makes them successful. People are sometimes hesitant to do it, especially with the rise of AI, because there’s concern about documenting processes leading to job loss. But it's important to see documenting data processes as beneficial for the enterprise, not as a threat to individual job security. It's about treating data as an enterprise asset, not as personal property.

That is a big cultural shift for many organizations. It’s clearly not as easy as saying, “Team, from this day forward we’re treating data as an enterprise asset.” So, how do you actually confront the cultural barriers associated with data quality and management?

Liz Marsh: Of course, it's not that simple at all. How I've been successful with it is showing what good looks like. Create a process and implement a solution that people can easily leverage and train them. Get them on board incrementally with the ideas, then the process, and then hold them accountable.

At the same time, I tie it into annual goals and make sure that people are talking about the data-driven culture at every level, from the C-level to front line. Now it's part of the journey and your job to get this information into a data catalog, for example, and to maintain it. People find their job becomes a lot easier when they do this because they don't have to be pulled into every single conversation. There's a place for someone else to go, and if they have questions, they can keep asking further.

How do you make this process stick? Like many initiatives, I could imagine success early on and then the discipline and alignment fall away?

Liz Marsh: Indeed, this is a big challenge. You have to be methodical about how you roll out enterprise-wide initiatives. Start with the core team in the area you’re impacting—your domain SMEs—align on definitions and set the expectation that they push these definitions down their channels and hold people accountable. Fundamentally, it’s changing the way we talk about data. If somebody brings up a different way to interpret a term, go back to the data catalog with them and make sure everyone is aligned. You’re giving the team the tools to ask: “What is the truth?” And this is learning how to talk differently about a company’s data.

Going back to my ego comment earlier, it’s not uncommon to be in a situation where the same people have been maintaining a data warehouse for a very long time. So, sometimes it helps to start with a best-in-class data model in retail, or whatever industry you’re in, and use this model as a sort of third-party arbiter. Implement this neutral model and adjust as you test and learn. This can help position the model as the company’s idea, as opposed to pitting individual ideas against each other.

I’m curious about the role SMEs play in all of this. They seem like such an important factor for institutional knowledge, MDM sustainability, and so on. The paranoid in me goes to how a D&A leader such as yourself deals with SME turnover and the like?

Liz Marsh: You just nailed one of the biggest pain points I’ve experienced in my career—and I’m sure my peers will identify. Culture is central to this discussion because I've experienced organizations that lack a culture of collaboration. They operate in silos, more concerned with protecting their own interests rather than embracing a company-wide perspective. However, my current environment is the opposite; it's highly collaborative, which speaks volumes about the leadership culture that's been cultivated over the years.

Creating a data-driven culture isn't just a task for those on the front lines dealing with data problems. It's a mission for every level of the organization. It's about developing the people, processes, and technology to truly enable a data-driven approach.

One of the main challenges, especially with SMEs, is high turnover, as you said. Sometimes they move on to new roles, but often I see organizations cut large groups of analysts to save costs, not realizing the impact on their bottom line. They're not just losing headcount. They're losing institutional knowledge—the very fabric that keeps the organization cohesive. Companies do this and then when they're in crisis, they scramble to bring that knowledge back in. Watching this cycle is both fascinating and a bit disheartening.

Considering the challenges of high turnover and loss of institutional knowledge, what’s your proactive approach to retaining the knowledge? What lessons have you learned along the way?

Ownership is key. For each data domain or dataset, it's essential to identify an owner. Then set a clear expectation for them to start building something tangible—like a data dictionary or a business glossary. What are the different fields? What do they mean? How are SMEs translating them? Building artifacts like this helps clarify these questions and the flow of data from beginning to end.

If I didn’t have a data catalog, I’d create one starting now and push the SMEs to it. I’d set a goal for everyone to contribute to it over the next quarter. It's like creating an encyclopedia for the organization's data—how is that not fantastic?! Partnering with different teams, I'd drive adoption and awareness of this new enterprise asset.

We often find ourselves spread thin, juggling multiple tasks. It's easy to get lost in the noise of urgency. So, I try to be really intentional about setting clear goals for myself and my team. Everything we do must align with these goals. If it doesn't, it has to be deprioritized. After all, prioritization means making tough choices. It’s not prioritization until the decision hurts. What's most important to me is establishing foundational building blocks that allow my teams to scale.

I’m going to steal that line, “It’s not prioritization until the decision hurts.” So true. Once you have the foundational building blocks in place—or perhaps as you’re building them—how have you done business differently as it relates to MDM?

As a business leader, my starting point is always the vision of the business problem we're trying to solve. Some may view MDM as solely an IT initiative, but true MDM is about the partnership between business and IT throughout the entire journey. Implementing technology is one thing, but without the business teams and operations, like data stewardship, to maintain it, the technology is basically useless.

I focus on leading with a clear vision for operations, then establishing the organizational structure needed to support the tool. My experience in leading MDM teams has taught me that MDM is often the catalyst for business process excellence. Now, is it necessary to thoroughly document data when only impacting a small area of business? Maybe not, but for business-critical domains like customer master data or product master data, you need to be super clear on your definitions. If not, various applications will be negatively impacted downstream, not to mention executive reporting. Precision is non-negotiable here.

MDM and business process excellence…tell me more.

Liz Marsh: MDM is often the forcing function for business process excellence. It creates that organizational muscle that really cares about data quality. While MDM might start the conversation, it's about building an end-to-end landscape that keeps MDM elements high quality and flowing to all systems.

Going back to this idea of doing business differently. Talk more about your business partnerships. Where have you had success getting buy-in on your vision or delivering wins for the business?

Establishing governance partnerships is my first move. Specifically, setting up data governance channels for various domains and creating an enterprise data governance office to steer the ship. These teams often need a model of what good governance looks like to emulate and adapt for their specific needs. They require a clear structure, knowing what boxes to check, when to check them, and having the authority to act. Data governance teams are ally number one.

Finance is another key partner because making a positive impact on finance data makes everyone happy. It's a complex area from a data perspective, but it's worth the effort.

Marketing is also great partner. They're typically at the forefront, always looking for innovative ways to acquire new customers, sell more products, or enhance the customer experience. They're usually one of the first to adopt new use cases because of the significant value proposition, impacting not just internal operations but also reaching out to the customer base.

What emerging trends or challenges do you see in the field of data quality?

Liz Marsh: One major trend is AI, which is becoming a forcing function for data quality conversations. Organizations have largely not focused on data until it's time to report results. But with AI driving actions within organizations, data quality needs to be considered upfront.

Documentation is key, and I hope we see more companies making business glossary creation and documentation more engaging. Gamifying data cataloging could increase engagement and make sure people are actively maintaining data quality.

Finally, there's a big push for driving data culture change. Data is the new gold, but not everyone knows how to handle it. We might need knowledge sessions to help people understand the value of data and how they can contribute to the journey.

Do you see a future where the “master” in data management is dead?

I don’t believe master data will ever become obsolete, although the term may evolve. MDM is certainly changing. It’s no longer just about mastering a small set of critical data elements but about constructing that end-to-end landscape I mentioned earlier.

Establishing an organizational dedication to data quality is so important as AI takes flight. In my experience, MDM acts as a powerful driver for this commitment. While it may continue in this role, we have to remember that master data is just a fraction of the data our organizations handle. It shouldn’t be the end of our efforts. Perhaps MDM’s evolution lies in sparking broader cultural change, laying the groundwork for expanding these practices into other data areas.

What advice would you give to organizations—what to do and what not to do—as they embark on their data quality management journey?

Liz Marsh: Don't fire your people who know about the data. It's tempting to cut costs, but you're losing invaluable institutional knowledge. And don't feed into the fear machine around AI and automation. It's important to create a culture where people are excited about data and see it as an opportunity for growth, not a threat to their jobs.

Start by driving ownership. Identify the owner for each data domain and set clear expectations for them to create documentation like data dictionaries or business glossaries. If you have a data catalog, push everyone to use it and make it part of their goals to get things documented.

Be intentional about your goals. What are the foundational building blocks you need to establish a team that can scale? Make sure everything you do ties back to these goals, and if something doesn't fit, it has to be deprioritized. Make your decisions matter.