Originally published on informationweek.com
Master data management (MDM) has always been important and quite frankly, we’re all sick of hearing about it after three decades. For this and other reasons, some enterprises are unable to get their data houses in order, which is critical now, given the widespread use of AI and data analytics. In short, businesses that want to be competitive better prioritize MDM sooner rather than later.
Customer service, internal efficiency, and automation are still important, but AI introduces a new dimension, and a new level of urgency to this, according to Graeme Thompson, CIO at AI-powered enterprise cloud data management solutions provider Informatica. “It’s one thing to miss out on the opportunity to automate an internal process. It’s a completely different and much more serious thing to miss out on being able to have an AI-assisted customer experience or a fraud detection process.”
One challenge with MDM is that it’s not as sexy as the application-layer stuff, so it can be difficult to allocate the necessary resources to make it happen. While MDM tools can help, there also needs to be a process change, which requires a different mindset.
There is a mindset shift that must happen to get people to buy into the cost and the overhead of managing the data in a way that’s going to be usable, Thompson says. “It’s knowing how to match technology up with a set of business processes, internal culture, commitment to do things properly and tie [that] to a business outcome that makes sense,” he says. “[T]he level of maturity of some good companies is bad. They’re just bad at managing their data assets.”
Some enterprises, such as cruise ship companies, are unable to recognize customers across different cruise lines because their data is still siloed. The result is failing to recognize customers across cruise lines and missing out on substantial financial opportunities. Meanwhile, insurance companies are streamlining the claims process by prioritizing data quality.
“[MDM] has very real business consequences, and I think that’s the part that we can all do better is to start talking about the business outcome, because these business outcomes are so serious and so easy to understand that it shouldn’t be hard to get business leaders behind it,” says Thompson. “But if you try to get business leaders behind MDM, it sounds like you want to undertake a science project with their help. It’s not about the MDM, it’s about the business outcome that you can get if you do a great job at MDM.”
CIOs must also make sure stakeholders understand the cost of failing to act, such as following versus leading an industry, providing substandard customer experiences and risking compliance audits and legal action.
Delaying MDM Is a Recipe for Disaster
Some CIOs are facing serious technical debt when it comes to MDM.
“Everyone wants to bypass the MDM phase. Let’s just get the data right for this one project, and then inevitably, [it leads] to other problems,” says Doug Gilbert, CIO and chief digital officer at business and digital transformation service and solutions provider Sutherland Global. “You’ve taken that contextual understanding, and now you’re doing AI, blindly follow[ing] that data and recommendations for you. Before, you could do a kind of quasi master data management around one or two projects and not think about it holistically.”
Through 2026, Gartner expects organizations to abandon 60% of AI projects unsupported by AI-ready data. “Organizations that fail to realize the vast differences between AI-ready data requirements and traditional data management will endanger the success of their AI efforts.”
This puts the importance of data governance and MDM front and center.
“I see two challenges going forward to put in a master data management strategy and structure because the very nature of [AI] systems is supposed to be autonomous. You must make sure that [the data] feeding it is always clean,” says Gilbert. “I do MDM because we go through so many different audits. It was painful, but I have less breakage, and my systems require less maintenance. I get proper AI outputs and proper predictions when I’m doing analytics. More importantly, my auditability is very easy to prove out because we have the proper controls in place.”
Louis Landry, CTO at cloud and analytics data platform provider for AI Teradata, says in the last five to six years, organizations have walked away from rigorous data governance practices and the desire to automate everything. Instead, they’re having AI agents react to the data they have without that rigorous data governance.
“It definitely feels that we don’t necessarily want to talk about [MDM], but it’s very important and very necessary for the future we’re all planning to live in,” says Landry. “What I’ve seen over the last several years is when you’re talking about data quality and data governance, folks might be willing to spend money on a technology tool, but they’re not willing to spend money on the process and people that are associated with it, and a lot of this is a people problem.”
In older organizations, MDM maturity tends to be unevenly distributed. The core data tends to be fairly well organized and managed, but the rest isn’t. The age-old problem of data ownership and a reticence to share data doesn’t help.
“The notion of data mesh [is] I’ll manage this piece, and you manage that piece. We’ll be disconnected but we can connect, and you can use it, but don’t mess with it. It’s mine,” says Landry. “We’ve known for decades that value acceleration comes when you integrate all this stuff so you can see inventory with customer data, sales data with revenue data — the stuff where magic starts to happen when you bring all these things together. The most advanced organizations have subject matter experts for specific domains. It really improves the overall quality and accessibility of that information and allows data to be turned into knowledge.”
Other Things That Have Frustrated MDM Efforts
In the tech world, whether it’s networks or MDM, there are opposing trends that tend to arise, not the least of which is centralization and decentralization.
“There’s always this back and forth between governance, control and accuracy versus autonomy and agility. I think we’ve been hard tilted towards autonomy and agility,” says Landry. “[W]ith things like generative AI and agents, it looks like we might get a chance at serving both principal needs because you can kind of separate the data management side of it and provide the right kind of governance and control that’s decoupled from all of the autonomy and agility that’s necessary at the consumption and analysis layers.”
He sees the data problem becoming more acute given that every app seems to have its own database and unique version of the “truth.”
“We’re going to see an unimaginable complexity crisis, and I think that fragmentation is something that we’re all going to have to deal with, and the practice of master data management is going to be incredibly important in dealing with that,” says Landry.
The pace of business change and AI innovation is already moving faster than many organizations can manage. And the diverse ways people need and want to consume data is creating a tremendous amount of technical debt.
“We haven’t even begun to come to grips with what that looks like over the last year, in terms of these AI apps and agents and things like that. [Enterprises are trying to] figure out how to monetize their data assets, while maintaining enough governance and control. That’s an incredibly challenging thing,” says Landry. “But I think we often forget that not all these problems are technology problems. The most locked-in CIOs and CTOs that I’ve talked to are absolutely looking at where to apply real energy around people and processes and where to apply technology.”
Landry recommends starting with data domains versus applications, and process automation, not tools.
“Get those things in order, and then as soon as you do, go find the right tools that match the kind of process that you need,” says Landry. “I would invest in expertise in the data domains that matter for you, [who] really understand these things so they can help and guide all of the people that are going to be building applications and agents and tools on top of your data.”
About The Author: Lisa Morgan is a contributing writer who covers business and IT strategy and emerging technology for InformationWeek. She has contributed articles, reports, and other types of content to many technology, business, and mainstream publications and sites including tech pubs, The Washington Post and The Economist Intelligence Unit. Frequent areas of coverage include AI, analytics, cloud, cybersecurity, mobility, software development, and emerging cultural issues affecting the C-suite.