The Data Chief | Episode 21

Invesco's Jim Tyo on data governance best practices

Jim Tyo



Current EpisodeEP21: Invesco's Jim Tyo on data governance best practices

Episode Overview

Joining Cindi today is Jim Tyo, CDO at investment management firm Invesco, and the former CDO at Nationwide, where he was responsible for overall strategic vision, planning, execution, and management of all activities related to the operation of the Enterprise Data Office (EDO).

On this episode, Jim and Cindi discuss best practices and the evolution of data governance and information quality over the past decade, taking the 'digital twin' approach from the world of manufacturing and applying it to the world of data, the ethical nuances of navigating privacy laws and nontraditional inputs for complete datasets without introducing undue biases, advice for people aspiring to pursue a career in data, and much more

Key Takeaways:

  • Strategies Jim has leveraged to properly balance data governance and information quality (for which his team has won awards), and how the conversation around data governance has changed during his decade in the field.
  • How the concept of the digital twin transitioned from the world of manufacturing to the world of data, and how organizations can use it to get an even more complete picture of their organization.
  • How Nationwide created an internal data marketplace for people to more easily find all the data available within the company.
  • The ethical considerations of navigating privacy laws and using nontraditional data inputs like fitness wearable devices and social media to ensure fair coverage without bias.
  • Why good teamwork is critical in analytics, a world where nobody knows everything.

Key Quotes:

Jim leads teams that win industry awards in data governance and information quality (DGIQ). What's his secret?

"I sometimes joke that one of the hardest things that a data officer could tackle is an enterprise-level data governance program. You know, governance sometimes can be a four-letter word for many corporations where it sometimes can detract from the priorities of the value creation and business and growth of the business outcome you're trying to drive. So I learned over the course of working at multiple companies how to put more of a service-oriented approach into the governance framework and risk management.
I know I scare my senior leaders sometimes when I say we're going to invest the bare minimum in governance. And the reason I say that is [if] you invest any more, you're taking away from those business-outcome, capability-driven opportunities. But of course [if] you invest too little in your governance or risk management, you're putting your members, your partners, your associates, even, at risk. So investing the right amount, having the right priorities, and then coming in with more of a service-oriented approach with governance and expanding that across the enterprise, that's really what that best practices award was for.

Has the conversation around data governance changed during the time Jim's been working in the field?

I think it has changed over time. You know, I've been in a senior executive data leadership role for over a decade now, and I've seen the business partners be more appreciative and approachable about governance and about data risk management, up front than they've ever been. And I think a lot of that has to do with the evolving privacy standards and they're hearing and seeing the horror stories of breaches and other data-centric issues that have happened to damage brand, damage their member and partner relationships. So our business partners are much more open to the conversation, initially. What you have to do is earn the right to have that conversation with them about taking ownership of their data: the business owns the data. Take that ownership, provide that stewardship, be very specific around the scope and the compelling case for the rules and controls that you put in place rather than trying to come in with more of a command and control and policy for the sake of policy.

How does Jim encourage a changing of the guard from IT to a business itself when it comes to taking responsibility for owning the data under its care and the means of protecting that data?

It comes from both sides of the equation, honestly, where sometimes you hear business partners say, 'No, IT owns the data. They own the systems that generate the data. They own the analytic platforms that house the data. They own the business intelligence and analytic utilities that consume the data. Technology owns the data.' And then sometimes you hear from technology, 'No, we own this and it's part of our ecosystem.' But when we start the conversations, we start with those business outcomes. What are those business outcomes in mind that we have with our data assets and establishing that education and training for the stewards and business process owners for data management and data governance within those business units? That really sets the stage for what ownership really means.
What we're asking them to do in order to own and manage data appropriately and what we're going to provide in technology or in the data office as a shared service, what we're going to provide them is the tools and capabilities to help them manage and own that data appropriately. ... We kind of are the data custodians. And if we believe that data is the digital manifestation of our members and partners, our business owns the physical relationships with those members and partners; why wouldn't they own that digital relationship? We can be the custodians for that for them, but they really own that relationship.

How the concept of the digital twin has made a transition from the world of manufacturing to the world of data:

The new CIO had come into Nationwide; it's been close to two years. and Jim Fowler came from GE Digital, and GE of course has a very, very large power presence. And they had been pioneering some digital twins; when they were describing that, though, it was more of a physical manufacturing mindset. Let's take jet engines for instance, where they could create this digital representation. So instead they could turn it around and look at 3D views, look through the engine itself so that they could hone in on parts or components that they may think would fail quicker than others to keep them on the planes, keep them out in the field. So when, when Jim came, we started to talk about data, the data management structure, about Nationwide as a whole. Nationwide's mission as a protection company is to protect people, businesses, and futures with extraordinary care. In order to do that, you have to know your members and partners and associates.
And so we started to take that digital twin mindset for manufacturing and apply it towards our data. I have always talked about data being the representation of our members of our partners, and so if we can construct that clear reflection using that, that digital pieces of our members person, we can then meet them where they are. We can better anticipate their needs in the future for the mutual benefit of both them and the companies. That's how we've been describing digital twin insight. Nationwide is more so from a data construct of that reflection rather than more of a physical, manufacturing construct.

At what point is this transition to the digital twin, and how does it compare to what can be discerned from a traditional data siloed structure?

One of the other things [in government] that I hope ... doesn't go away is we did pilots of certain things ... so what that allowed is ... people to test and get to know the tools we created, situations where people could actually see and evaluate a thing, not analytical thinking in abstract, but a real experience and learning through those pilots and then taking the lessons from the pilots helped us then take the next step in doing something bigger, something, maybe agency-wide, something with a bigger impact.
We recently named a Chief Digital Officer, the Chief Digital and Innovation Office has been focused on digitizing key experiences and really anchoring to the journey that a customer, a member, or a partner would take with us. That's where we're focusing. Because you could talk about a twin and it could feel like boiling the ocean. You want to get everything for everyone. We're really focusing and honing in on those key journeys. You mentioned filing a claim, of course, one of the potentially traumatic experiences that a member or partner or supporting member could have interacting with us as a protection company. So if we can create that extraordinary care, that digitized journey, that digital reflection so we know as much as we possibly can of that member or partner during that process, that's really where we're focusing our time is honing in on those key moments that matter.

In what way is the data transformed to be used practically -- such as in filing a claim?

I'll pick a robotic process automation, RPA, because that's something that's probably near and dear to most data professionals' hearts. And we've been talking about how do we apply automation to be more efficient in a lot of the operations and transactional environments. You can apply process automation against the worst process in the world and you can digitize it. Yes, it can go from analog to digital, but it could still be a bad process. And that that's really where we have been very deliberate to say, is this the appropriate process to support the vision of extraordinary care, that member outcome, that member experience, or even an internal operational experience for one of our associates. And we have done some process redesign and even completely broken down a process and added in both traditional and non-traditional data sources. I would say that's another aspect to this. Not only looking at and just not automating or digitizing broken processes, but re-imagining what that process should look like. And the types of data both internally and externally that could help us create that clearer reflection that clearer picture of how we can anticipate needs.

How does Jim tackle the process of transformation without shutting down operations entirely -- especially during a year like 2020 when there are more claims being filed than ever?

If you have a dedicated approach to reimagination, you can run the innovative, creative redesign that you're doing in parallel with what has worked and continues to work currently. And rather than jumping in and trying to change the tires of a race car as it is running around the track, you might build a new component or a new piece of the race car to the side, and you can have that start to incrementally make a change on how that experience is going to look in the future without completely taking the car apart. Because we know business is moving at the speed of light now, and the last six months, if it's taught us anything, it's that we've underestimated how fast consumer expectations and consumer needs can change. They can change on a dime. And so it's important to run those things in parallel.

How nontraditional data is being leveraged for associates at Nationwide as a marketplace experience:

The marketplace has kind of been a vision of mine since I was an analyst very early on in my career as online markets like Amazon and other shopping experiences were coming into focus and it was such a cool way to shop or to buy -- even now [as] we're interacting with virtual doctor visits and things like that. Data, for whatever reason, has continued to be almost locked up, contained; a very small percentage of your overall associates have access to a very small percentage of your data. ... Why can't I create an Amazon-like experience shopping for data within the organization? And I started to think about that when I was in my first role as a data officer, my second role got a little bit farther at Nationwide, We actually made it happen. It's come to fruition. We have a completely virtualized shopping experience ... any associate in the company [can get that data from an internal website].

How was Jim successful in implementing adoption of such a dramatic change in such a short amount of time?

I joined Nationwide at a very pivotal moment. They had a data organization within technology, but really were trying to transform and let that grow into something that could truly turn into a competitive advantage. And as I started to get to know the leadership team pretty quickly -- and it was very early on in my tenure; I'd only been there a couple of weeks -- but I gravitated toward some of the best thinkers and the best thought leaders that I'd been fortunate enough to work with throughout my career. And one of them is now Chief Data Architect, and he and I were brainstorming. And he's a very creative guy, has a background in professional photography, but has grown into a world-class data architect. And he and I were in my office and we were brainstorming.
And I was explaining to him just as I did to you and the audience today about this vision of a shopping experience. And we started to draw things up on the board. And it was just a lot of fun. I love those activities. And he started talking about a guy on his team and he went back and they discussed it a little bit. And I really didn't think something was going to come that quickly, but this was on a Thursday or Friday. By Monday, they both came into my office, big smiles on their faces and I was like, 'Hey, what's going on?' And they said, 'Here, check this out.' And they had a prototype of the marketplace and it is the same platform and basically the same bones -- as two weeks into my tenure at Nationwide -- to what we've grown into a productionalized product and cornerstone to our data ecosystem.
And it was just so overwhelming, I think, [that] they took their own time. They were embracing the vision and they were using open source too, which was fantastic. It didn't cost us a lot to do the MVP, almost nothing. The underlying platform is the same that runs CKAN is the platform. So it wasn't really expensive. We didn't have to go get software licenses. We're completely virtualized and digitized. It was kind of clunky, the first version that you saw, but at least you could feel it, you could touch it, you could see it, and then we've just expanded it from there.

Suzette's outcome-driven approach to getting students interested in pursuing a career in data:

If you ask someone in seventh grade, 'Do you want to be a computer programmer?' or 'Do you want to be a data scientist?' They might be [unenthused]. But if you say, 'Do you want to study human machine interaction? Do you want to ensure that our oceans are clean? Do you want to figure out how we have a resilient food supply? You want to be a game graphic designer?' They're like, 'Okay! That sounds interesting.' We have to give signals from private companies around what kind of things they want people to do to be employable. And we have to talk about the cool stuff they get to do to get them interested in all the underlying coursework and study that goes with enabling them to do that.

As one of the few women in data leadership, how does Suzette believe we can help more women and people not equally represented find their voice in the space?

When I interact with women, the very first thing I always start with is: 'Know and understand the business in which you were working.' Those authentic connections give you a really strong foundation. Just start the conversations. Always be at the table. Have a seat. Contribute to the conversation. You're in the discussion because you bring a unique perspective. And demand better behavior of everyone that you're working with. ... We should celebrate unique and diverse approaches to producing different kinds of outcomes.

On the importance of drawing talent from diverse (even legacy) skill sets and backgrounds:

I focus a lot, when building the teams, on qualitative aspects, like an innate sense of curiosity, a willingness to challenge and question, and come from different perspectives. I've had a professional golfer. I talked about a professional photographer. I have had, of course, two PhD statisticians that are data scientists -- incredibly valuable and important, as well. Some of my best data analysts and data solution support and thought leadership had very little secondary education training, but they were perpetual learners. And I have tried intentionally to bring in as many different perspectives as possible, because I found that to be a secret to uncovering some of those really innovative and cutting edge techniques.
The best thing about my short tenure thus far at Nationwide [has been] the opportunity to partner with Columbus State Community College. There are a lot of programs and a lot of training for people that are already in a data profession. We feel like there's a lot of conferences, a lot of webinars, a lot of ways to upskill and continue to grow [as] a data professional once you get there; there's also a lot of great programs in colleges and universities for students coming through that can learn the computer science [and] the statistics.
[It's also helpful for] existing high-value, highly experienced technology professionals or business professionals that may be in a legacy skill set. Maybe some of those highly experienced, highly valuable resources and associates would like to get into a data and analytics profession, but [wouldn't otherwise] know how.

How Nationwide has made the underwriting process and its cost more efficient through automation:

A lot of people talk about machine learning and artificial intelligence now, but it really is grounded. The data scientists work are their prediction machines. They are that, that decision intelligence and getting that in a virtualized mode that can near real time integrate with our systems of record had been a passion for Nationwide for quite a while. Over the last decade that the team had produced in partnership between the analytic group, business group, and the technology area, what they call a model factory, where they've been able to at least streamline components of getting the models into production. The quicker you can do that, the quicker you can gain value, measure the progress because as we all know, models need to be evolved over time. They're not perpetual, because as the audience changes, so does the model need to change.
Our analytics team and our enterprise analytics office have ... been able to produce underwriting models, and I'll use life insurance underwriting -- [usually] a very invasive process -- for example. They've been able to go from 0 percent automated to close to 30 percent automated underwriting decisions. It's much less invasive; the time to decision is dramatically, exponentially shortened. There's not as many tests that are required and their hope is they can get up to 40, 50, 60 percent automated. They're on their eighth iteration of a machine learning multi-ensemble algorithm. So that's one area where I think Nationwide has been very successful because of that, that innovation that they were doing. And ... I think the statistic was that the cost dropped $400 per application when they had instituted the model, which is pretty impressive."

On working with data that comes in under the scrutiny of different privacy laws (such as GDPR) and preventing bias based on nontraditional inputs like social media and fitness wearable devices:

We have to wrestle with this idea of ethics and artificial intelligence and data ethics. Let's say, for instance, that we underwrote a life insurance policy [for a nonsmoker]. And then let's say that it is acceptable that a company could use social media data for policy reviews, policy underwriting, or policy servicing, and we see pictures on that particular member's Facebook [in which that person is] smoking a cigar and partying and having a good time. What do you do? What can you do legally? You could potentially cancel the policy. ... We have proven that you didn't answer the application truthfully; we could increase your premium. We could wait until you file a claim and then announce that we've noticed that something was wrong with the application, but what you legally or compliantly can do is sometimes different than what you ethically can do.
We have members on our board that are very in tune to data security, and so they've asked questions about our data risk data management process. We've actually elevated data, quality, sufficiency, and protection as one of the top risks of the company. And ethics is a component of that. ... I don't believe there's ever going to be a completely automated algorithm that defines our ethical standards at Nationwide as a protection company. But if you're having the dialogue and at least you have a mechanism to ask the questions and ensure we're comfortable on a baseline set of principles above and beyond ... the moral minimum of what's legally acceptable, I think that's where you start to ensure that the ethical values of our protection company [are] being applied to our automation.

How transparent should such data be to the consumer it's affecting? Will we see greater transparency being shared in the future?

I think that we are going to get to a place where it's going to be a standard expectation. It would be an expectation of me [as] a data guy. If I apply for a loan or if I go to the doctor or something ... I may ask things that most consumers wouldn't. But more and more, I think our standard members or partners are going to be asking the same type of questions and to understand how those predictions are occurring as far as -- and even including and up to what data was involved. 'What data of mine did you use to make that prediction?' Those questions, I think, are going to be standard expectations.

The developing innovations that Jim is most excited about:

What I want to see in the next three to five years is the data ecosystem. When I talk about our data strategy, it can go in so many different directions. I land on data democratization, data virtualization, and the digital twin -- those three legs of a stool are really what we're going to stand upon to catapult the type of care that we can provide to our members and partners. How do we get as much data as possible available to as many of our associates, partners, and members as we can?
95 percent of our data should be available to 95 percent of our associates. Historically, it's been dramatically lower than that, and I'm challenging a lot of the conventions in saying that. Of course, we're not going to share proprietary contractual data or confidential information, but I really want to push the envelope. Why can't we have as much data as possible available to as many associates as possible that will bring the entire organization's data acumen up together.
[On] virtualization ... how do we completely virtualize and expose data digitally so that, if it's on legacy platforms, that you don't necessarily have to physically lift and shift that data to another place, which gained you zero business value or zero business return on investment? Instead, can we virtualize and play data where it lives?
And there are times when data professionals almost hinge their job on knowing a certain asset or being able to get to a certain type of data -- if ever I need this type of data, I go call Jim and Jim knows how to get it. We have to break that mold. We can't have people be the only expert of our data. We have to be able to free the data [and] let the data speak for itself. Let the people do the analysis. Do the insights, do the data, [and the] storytelling, not be the gatekeeper for a particular asset.

Jim began his career years ago as a business analyst at KeyBank. What words of wisdom might he impart to someone in such a position now that he's gained the perspective of a serial disruptor, visionary, and CDO?

Focus on the key skills and attributes that you want to develop ... [including] some computer science and coding capabilities. And then learn the business; learn what the outcome is that you're trying to drive. And if you can combine components of those things on the qualitative side, learn to tell the story to communicate the message. I was not a great communicator early in my career, and I got that feedback. I spent a lot of time in front of the mirror recording myself and talking about things, and it has helped through the process of moving into leadership.
And [there will be days when you say]: "I can't do this. I don't know what I'm doing. I am completely out of my depth." Quite honestly, because of how fast technology is moving, it's almost every day that I feel like I'm completely out of my depth -- even now. The good part is I've got brilliant people around me. ... What you need to do is know how to leverage your teammates, leverage the people that work for you and with you, and together you will be the best you can be.

At this point in 2020, for what is Jim most grateful?

I'm grateful for Zoom. I'm grateful for Teams. I'm grateful for Skype and WebEx and all of the digital tools that Nationwide and other companies have prepared to get us here. I am an introvert at heart, to be honest; my profession requires me to be more of an extrovert, which takes a lot of energy, but there is something about being able to see people's facial expressions and to interact, and I don't believe companies could have been as successful through this virtual transformation without the great tools, technologies, and capabilities we have to see each other and to be able to interact in the way we are. It's pretty impressive.


Jim Tyo is CDO at investment management firm Invesco, and the former CDO at Nationwide, where he was responsible for overall strategic vision, planning, execution, and management of all activities related to the operation of the Enterprise Data Office (EDO).

With more than a decade under his belt as a senior executive in the data field, Jim has envisioned, championed, and built data analytics organizations, processes, and technologies from concept to design in implementation and through execution. His vision has created significant culture changes while shifting information delivery methodologies — including innovative agile execution approaches — ensuring appropriate controls and governance focused on a data-driven experience and empowering business outcomes.