Data Science Leaders | Episode 08 | 43:06 | June 22, 2021
The 3 Biggest Jobs of Any Chief Data Officer
Heidi Lanford, Chief Data Officer
Fitch Group
Data Science Leaders | Episode 08 | 43:06 | June 22, 2021
Heidi Lanford, Chief Data Officer
Fitch Group
As more organizations recognize the power of data to transform their decision making (or for it to become a product in its own right), the role of the Chief Data Officer has become critical.
So what are the biggest challenges facing every good CDO? And where do data science teams intersect with that work?
In this episode, Dave Cole is joined by Heidi Lanford, Chief Data Officer at Fitch Group, to discuss strategies for cultivating partnerships between data science leaders and the Chief Data Officer.
They also explored questions such as:
Welcome to the Data Science Leaders podcast. I am your host, Dave Cole. Today, we have Heidi Lanford, who is the Chief Data Officer at the Fitch Group. And I will let her describe to you all who do not know what the Fitch Group does or is. And previous to her current role at the Fitch Group, she was—let me get this right—the VP of Enterprise Analytics and Data at Red Hat.
Today I think what is really interesting here about Heidi’s background is she was a data science leader at Red Hat, but now she’s the Chief Data Officer. And we’re going to be talking a little bit about, on our agenda today, the difference in the ways in which a DSL should partner with a chief data officer role. And hearing from somebody like yourself, Heidi, who’s done both, I think will be super interesting. So welcome Heidi to the Data Science Leaders podcast.
Thank you! Thanks for having me. I’m excited to be here.
Great! So, before we get going, just tell us a little bit about the Fitch Group.
Sure. So the Fitch Group is essentially a ratings agency. It’s considered to be one of the big three credit ratings agencies, and the other two being Moody’s and S&P. We are a global company and we provide ratings on things from corporates, to sovereigns, to structured finance, financial institutions, you name it. And we also have a solutions business where we try to really take all of the data that we’re looking at on the viability of those types of companies and their ability to essentially pay back their debt. And we use that to create new products and that’s part of our solutions business. And we also have a learning business that does some training for companies and folks who are in related industries.
Got it. Well, that’s great. The topic today as I teased it already is to talk a little bit about that partnership between the data science leader and the chief data officer. So first of all, at the Fitch Group, I assume you have a DSL who is not yourself, correct, like your counterpart?
Actually, our data science leader at the Fitch Group, we have probably two roles that fit into that category. We don’t have a chief analytics officer, but we have a leader for emerging technologies, which includes, artificial intelligence, machine learning, RPA, and data science.
What is RPA?
Robotic process automation.
What is that?
Taking things that are super repetitive and happen in large volumes and basically trying to extract any kind of human interaction and process movement out of that and doing it in a robotic way. So using bots, writing code, scripts that take all of that and make it automated.
Makes total sense. So, really tapping into the fear that artificial intelligence is taking away jobs and one day we’ll make all humans redundant, in a way. I kid, of course! It’s about efficiency.
Yeah, so RPA is related, but there are really specific criteria in terms of volumes and repeatability that it requires. Whereas AI and machine learning are much more flexible and learning and continually evolving.
Got it. So you have a leader of your emerging technology group?
We have a leader of emerging tech and that leader actually reports to me in the organization. And we’re not a super large organization, so we have about 4,000 globally. Given the size of the organization, other companies might be different in terms of the type of roles that they would have. But we also have a head of strategy and business insights, business intelligence. And I am a partner to that leader, providing tools and hopefully—I’m fairly new in this role, I’m about four months in—an improved data platform for her team to really maximize their value and minimize the time that they’re using to have to corral and integrate and massage data to conduct insights. That team is focused on a lot of the operational business intelligence that we need for the company.
So things like product adoption or potentially pricing optimization and share of wallet, upsell, cross-sell opportunities, things like that. They are a somewhat central team that can provide services to the rest of the organization that does that type of work. My role is focused a lot on the exploitation of our data to build products because we’re a data company, ratings being our core product. And my role is to really partner with the business to transform the way we leverage data across our enterprise to enhance current products and applications and things like that. I think you’ll see some CDOs work on that operational data component. My focus right now is working on data as a product.
Let’s dive quickly into that, because my first question, and you’ve already anticipated it is, what is a chief data officer? What is their role? I think you perfectly described it already, but when you say data as a product, in layman’s terms, what does that mean to you?
Let me back up a little bit and talk mainly about how I see the role of the chief data officer and what they’re responsible for. And then I can give you maybe some examples of data as a product. I think of our jobs, I would say most chief data officers think of their jobs in three main categories. One is managing the information assets or managing that data. And that could be in the form of, what is our data platform look like? Do we have a data lake? How does the data lake work with big data processing or a self-service environment for people? What kind of data governance tools do we have? What metadata or information about our data and the data lineage that we want to be able to trace how data is transformed along a particular journey or path? All of those things fall into the ‘manage information assets’ bucket.
The second big traunch of work that is the responsibility of most CDOs is delivering insights and enabling others to deliver insights. And I know you want to talk about this in more depth in our conversation, about how far do we go enabling others to deliver insights. Is it really descriptive, prescriptive, predictive? How much of that data science work do we have centralized teams, or even decentralized experts, do versus enabling others to do it on a more self-service type of a model. But that’s the other thing. And when we talk about delivering insights and enabling others, that can range the gamut from providing tools and technology, that one can deliver insights on.
It could be dashboarding and data visualization tools. It could be reporting tools. It could be a data science environment. It could be allowing access to AI and ML technology and tools. So all of those kinds of things, I think fall into that ‘deliver insights and enabling others to deliver insights.’ And then the third big bucket or traunch of work under a CDO, and this one’s gaining more and more traction and there’s more emphasis on it, is generating incremental value. And in my world, where we talked about data as a product, that’s really about scaling data at Fitch for new product development. So we’ve got lots of data, data, data everywhere. How do we monetize that in an even bigger way than what we’re already doing?
That comes in a lot of different formats of not only storing it and housing it in a way that people can then go and do analysis and access it and create and innovate new products, but also that unstructured data component, and being able to have access to maybe, say, public data sources that we don’t create ourselves, but might be available out in the data ecosystem. How do we bring those valuable data sources into our domain and make them available to create net-new products essentially for us?
Got it. So I’m going to try to dumb it down for myself. The three main goals or jobs of a chief data officer are 1) the management of the data itself. So that includes governance and access to data. Number two is, enabling the team that is responsible for coming up with the insights. And I also heard you say actually delivering on the insights as well. I’ll circle back to that in a second here. And then last but not least is looking for ways to actually monetize the data and turn it into a product, so that you can actually sell the data as a product. And certainly at the Fitch Group that is something that you all do, it’s core to the business itself.
Not every company is used to selling data. I imagine the chief data officer, if he or she’s doing the right job, is thinking of ways and thinking outside the box and finding ways which to potentially create data products. But that makes total sense to me. Now on the second one, let me circle back. Do you see the chief data officer…by delivering the insights, do you mean you actually should have a team of analysts or are you talking about productionalizing models that are created outside of the CDO’s umbrella?
That’s a great question. And I think a lot of it depends on the skills and background of your chief data officer, number one. And number two, and not any less important is, the organization’s desire and willingness to want to have some things at the core and other things that are specific to a functional area or a product line as an example, to be done on the periphery. I think it’s a balance between the two. I know some CDOs that are very, very technical. They grew up in IT organizations, they’re experts at data warehousing, as an example, and database design. And then you have other CDOs that have grown up more on the data scientist world and data analytics. And that’s actually my background.
So my strengths are in understanding how people use data to actually do their job differently, in marketing applications and finance applications and supply chain applications. And where I need to augment my team is with those that are great technologists. I actually think you need almost a cabinet of a lot of those skill sets together. Where your chief data officer tilts on the spectrum is also going to determine who they need to bring in to support them. As well as, what kinds of things they can do for the organization. Now, I’m a big believer in being able to know enough on both sides so that you truly can design…I think you need to be able to design all of those things we talked about. Like your data infrastructure, your playground for data scientists, what types of tools you need.
You need to know enough and you need to have actually worked with them somewhat to be able to walk in the shoes of the people who you’re serving. Because if you don’t, then you become an order taker and you don’t truly have that experience. And most people, I don’t think, want to be order-takers. This is a role that straddles between the business and the IT worlds a lot. And so you’ve got to be very comfortable traversing both of those things, but I think it’s also important for CDOs to be very transparent and admit where maybe they’re not as technical and you’re going to rely heavily on your CIO or CTO partner in the organization to help you with those things.
Let’s dive a little bit into your advice in terms of working. Some of it you’re already hinting at, which is, the chief data officer, if they’re doing their job right, is not just managing their data warehouse, managing the tools themselves, but it’s actually trying to really understand the business problems that the business is trying to solve and the data science team is trying to solve and their DSL counterpart. And even the folks on the BI side as well. And if you have that background, fantastic. If you don’t have that background, what advice do you have for a chief data officer in terms of closing the gap? Maybe their background might be more on managing large databases and data lakes, and then having them move a little bit more to understanding the world of a data scientist?
I think you’ve got to get a fan club going. You’ve got to make a few really good friends and focus and deliver and help a few people out because then you’ve got successes. And everybody talks about quick wins and things like that. But I mean like very deep partnerships where you have a business partner who has a really clear idea of what they want. You are great on that technology and deliverability standpoint. And there’s a good problem that you’re solving. Most importantly, you know that what you’re building, they’re going to use it. And that’s the most important thing. You hear stories all the time of multimillion dollar technology investments, whether it’s in tools or in databases or platforms or whatever.
And then you’re asked, well, who’s using this and who’s changed the way that they’re doing things because of all this stuff that you’ve built for them? If you don’t have that, then I throw up my hands and say, “Why are you here?” None of us want to build really cool things that nobody’s using. I think this transcends everything. It transcends databases or data warehouses, dashboards that you’re building, predictive models that you’re building. You want to be able to answer the question, “How is this changed the game?”
It was an interesting story that one of my colleagues told me about a chief marketing officer who would start her twice-monthly staff meeting with a question, “Who’s taken a look at the—fill in the name of it—the CMO dashboard, we’ll call it.” A dashboard that had all these great leading indicators of, how’s our marketing funnel looking? What’s our win rate? How’s our cross-sell, upsell? What’s product adoption looking like? What’s our brand awareness score? Anything like that.
And she would go around the room and ask her staff, her senior leadership team, “What did you learn about whatever your particular area is by looking at the dashboard over the past two weeks? And more importantly, what is this caused you to do differently? Are you reinvesting your funds differently? Are you developing a more targeted campaign because of what this information is telling you?”
And I would say that is key to have people who are passionate about that. And by the way, I heard anecdotally that this leader, if you didn’t come back with a good response, and you hadn’t been using the dashboard, you were uninvited to future meetings if this happened on a repeated basis.
Wow, that’s tough.
You want partners like that. That would be my advice, is to find people who just really get it.
In a past life, I’ve done a lot of work with marketing teams and it’s interesting how data-driven they are. For those of you out there who are working building models and helping with campaigns and targeting and all that stuff, there’s a huge need for data science in the marketing world. That being said, taking the approach of, in your story, of a CMO really requiring his or her leadership team to be up on the data and understanding where some of the bottlenecks—maybe the pipeline isn’t great this week and it’s gone down and being able to answer the question “why?”—is interesting because I’m a big believer…it’s not just, as you were saying, just looking at usage. Looking at, is this platform being used, is this report or dashboard, is this model being used? But it’s the real question, is this actually how we make decisions? Is this actually driving some value, some ROI? It might be getting used a lot. But if it’s not actually changing someone’s behavior because of it, then it’s not really doing anything. It’s just barfing out data, but it’s not really an insight, an actual insight, right?
That’s right. So then, to riff on what you’re saying, there’s sometimes the gap why that doesn’t happen enough. It doesn’t happen enough. And when there’s that gap, I think there’s a couple of reasons potentially why. And one of those reasons could be that you don’t have leaders like that person I mentioned, that would require it as part of the conversation. Another reason could be, there’s the buzzword that we hear a lot now, and that’s called “data literacy.” And I think another big roadblock for people to adopt and implement analytics and data science, predictive models, all that great stuff coming out of the AI and ML space, is that the data scientists and analytics professionals aren’t going to go out there and make stuff happen with whatever models they’re building. They’re relying and dependent on partners to go, as I like to say, change how you do your day job.
So, are we empowering them enough if we’ve done a great segmentation model or we’ve built an awesome forecasting model, have we done enough to educate them on how to comfortably and confidently use that output to go change their day job? That, I think, ties back to data literacy, which we hear a lot about now, it’s definitely a big buzz word. And implementing programs like that are becoming more and more of an expectation. And going back to the CDO role, I think that that’s also part of the job description of the CDO, is you’re an evangelist for data and how a company uses data to do a better job. Data literacy programs…in my last job, we implemented a corporate wide global data literacy program.
They’re a lot of work, by the way, you need a lot of folks to participate and help develop content and market the heck out of it and have executives that say, “Yes, you need that.” But, concepts that data scientists and analytics professionals are very, very comfortable in, like the whole nuance between correlation versus causation, or how to deal with outliers, and what are minimum sample sizes that you can trust in order to believe this competitive analysis that you’ve gotten—just making some things up here. It’s not the data literacy is trying to train everybody to be a PhD statistician or a data scientist. It’s training them to feel comfortable with the outputs that they’re looking at so that they can go and implement that and take action on it.
That’s a critical component to getting that adoption and usability that all of us in this profession are seeking more of so that the company gets their return on investment on us and the investments we’re making in infrastructure and tools. And, it’s also job satisfaction. You want to see your work being used.
Totally. I think replaying a bit of what you’re saying is that, first of all, with a chief data officer, one of their jobs is to make sure that the data that is being used is of high quality and it can be trusted. And if you trust your data, the next step is really that data literacy. Do you have business counterparts—it could be data scientists, it could be analysts, it could be decision makers—who really understand the insights that are being created by your data science team or your BI team. Do they have the literacy to understand that when you put a model in production, this is what accuracy means? And if you’re doing a controlled experiment, this is why we are creating a holdout group because we want to be able to properly measure the results of this campaign, etc.
If you do not have that data literacy, then it becomes very difficult to have meaningful conversations that will actually result in changes in behavior. I think you’re hitting on a lot of good points. Do you have any advice to data science leaders and data scientists and data engineers and everyone else who’s listening? How do they start a data literacy program?
A lot of times, it becomes a grassroots effort. I do think it’s critical to have a senior executive champion or sponsor, like a CDO or the equivalent of that is really important for the sponsorship component. You also need a set of champions or executive stakeholders. These are your tribe of leaders that are bought into this as a concept and a need. And they’re committed to making this part of their staff’s learning and development plan for the year. And then, it is important, depending on how full-blown you go…in my last role, we rolled it out to a company of about 15,000. We had four different learning personas that ranged anywhere from a data beginner to a data knight or a data champion.
And we had an online Cosmo-type quiz. It was a fun quiz that bucketed you into which persona you were. Then it gave you a learning path and recommended courses. And they were online learning courses that were digestible chunks of 15, 20, 30 minutes a piece. But to build out all of that course content, I hired a full-time data literacy program leader. And then he tapped into our analytics community, which was throughout the company, and they actually helped contribute course content, recording of videos, doing the little quizzes at the end to see if you understood the content, made specific use cases for different areas of the company. So we would have a marketing example for the correlation and causation track, and a finance example as well.
So we had a very involved community, so there was a little bit of hiring full-time people. I would say it’s probably at the end, we were at like two people, full-time doing this to really get it off the ground. And it was heads down work for a year. And we had some help. There are consulting companies out there that can get you started, you can buy some content from other providers, and you can also create your own.
I imagine that some of the content is going to be fairly generic. Like some of the basics in statistics, the basics in understanding how data is used and the process of building out models in the process of business. But I imagine some, obviously, needs to be very specific to your organization. Because the data that you had at Red Hat is probably very different than the data that your team is using at the Fitch Group. I imagine it needs to be that balance, right?
Yeah, it does. And I also think I’m a big believer in specificity and examples. Sometimes if you are taking a course where the content is too generic, you’re also making an assumption that people are able to connect that generic example. Like if you were doing a course on causation versus correlation, and all of your examples were, my favorite one was there was something in USA Today, this was probably five years ago, and it published some, it was a medical journal study.
And it was like people who take sauna five days a week have a 50% less likelihood of having a heart attack. And I was like, “Oh my gosh, I can’t believe that this medical journal published something like this.” It didn’t look at maybe people who take sauna are also very into fitness and health and they eat right. And this was a classic example of a bad drawn out conclusion that you’re presenting to the general public who’s reading this, because it was in a USA Today article, and now they’re thinking…not that there was a rush on everybody going and doing sauna for five days a week, but you get the point.
Yeah, totally. I see it all the time.
You see it all the time, and it’s so bad. Every news association should have a statistician to look at all that stuff before it gets published, I think. But anyway, that’s a side note. My point is that if our course content just showed generic examples like that, which are great to get people to understand the content, but then they don’t go one step further and say, “Now let’s take an example at a Red Hat or at a Fitch where you’re analyzing financial statement data to assess all these factors for credit risk.” And I’m not saying that people aren’t smart and can’t pick that up, but let’s make it easy for them to relate this concept to their day jobs. So that I think they can more quickly then take this and go do something with it.
100%. I think there’s a lot of good information there from you about setting up the data literacy program. I’ve talked a bit in the past in the Data Science Leaders podcast about how it frustrates me sometimes when you look out in the real world and people should be thinking about things like a statistician, thinking about probabilities. I mean, certainly there’s been a lot of examples of that with the COVID crisis and looking at vaccination rates and like, “Oh, the vaccine doesn’t actually work.” Well, if it works 93% of the time, that’s pretty darn good, but don’t forget the 7% where it doesn’t work. You have a story of your cousin Ed who had the vaccine, but still got COVID anyway. Cousin Ed might be in that 7%.
Anyway, I could go on and on about that, but that’s really important when you’re trying to make business decisions and you understand that they’re not always going to be 100% accurate. Your segmentation’s not always going to be perfect, it’s going to have outliers, it’s going to have people that probably shouldn’t have been there. And that’s just part of the business and you’ll learn from it and you’ll get better. I think I’ve learned a lot today about what a chief data officer does. One thing that I’d wanted to touch on briefly is, there is a concept today, there’s an idea that Gartner has around the citizen data scientist, right?
I knew you were going to get here.
Well, I had to. And I think it’s interesting because there’s a lot of schools of thought. When you’re talking about data literacy, how do we bring some folks who may not have that statistical background and actually bring them into the world of doing data science? And as somebody who owns the tools and the platforms where the data science is being done, what are your opinions on the citizen data scientist, maybe even to define it in your own words?
Sure. I think of the citizen data scientist is someone, anyone, who can create or build a predictive model, but their primary job description is outside of the field of statistics or analytics. I don’t like to classify it as, did you have a PhD in statistics or not? I think it’s really, what is your job function and role? Because I’m a big believer that people can learn skills on the job, but a citizen data scientists wouldn’t have the primary job description of statistics or data science or analytics. Now, my perspective on that, so it’s interesting, you should actually interview my father about this because my dad—maybe he’ll listen to this podcast—is a retired PhD, and he was chair of the statistics department at a university.
I think he has always been somewhat amazed, surprised, I don’t know, of what I’ve been able to do in my job and I do not have a PhD. Now I’ve spent my career, and I was a math and stats major in undergrad. And my first job out of college, I was working in a management analytics group for Pricewaterhouse. And so, I did lots of programming and coding and built predictive models and did segmentation studies and learned and trained with a lot of people that had more advanced degrees than me, and had a lot of experience. Now, that being said, I think it’s a stretch to say just anybody can go pick up a data set and go build a predictive model and then, we should all trust it or do our day job differently like we were talking about.
One example that I have, and this isn’t really like a citizen data scientist versus Heidi Lanford example, but I was on a project a while ago at Keebler. So Keebler makes cookies and biscuits and things like that. It was a supply chain forecasting implementation. Something really, really basic, but we were using software to build forecasting models for Keebler. I remember having this heated debate with one of the senior leaders on that project about the formula, like how we should code the formula to measure something. And he was insisting that it was one way. And I was like, “No, it’s not. I know this.”
And I actually went home and brought my graduate statistics books back to the office. And I was like, “Here it is. You’re wrong. You’re putting X minus Y and it should be Y minus X over X,” something like that.
I thought you were going to say that you called your dad…
I did call my dad! I actually did call my dad. And I was like, “Dad, I think I’m right.” And my dad was like, “Yeah, you go tell him, you are right!” And of course my dad wanted to send the whole theorem about the whole thing, which is why I never asked my dad for help on any of my statistics classes in college, because I think he really wanted me to go get a PhD in statistics and I didn’t. So I let him down in that regard. I’d say the other example would be, I love marketing, I love marketing applications. I’m a consumer of marketing.
We all are.
But I am no expert in, say, how to build a brand. I have my own personal ideas. I didn’t go to school for it. I haven’t spent any time my career doing that. I’ve helped brand marketers do analytics projects, but if I were a brand marketing expert, I’d take ideas from people. But if someone was to say, “Oh, I can go build this whole new brand campaign.” And they hadn’t had any training in it, we wouldn’t expect that of a citizen brand marketer. So why is there this concept of a citizen data scientist? I think the key is data literacy.
You know what, I will retract everything I’m saying the day that we have everybody in companies totally completely tooled up on data literacy and they can go do stuff with dashboards that they’re seeing in predictive models and take that and go implement it and take action on it. And when we’ve exhausted that, where we’ve got tons and tons of folks trained on how to consume data and do something with it, then bring on the entourage of citizen data scientists.
Bring on phase two citizen data scientists.
Bring on phase two and I will back down.
No, I think it’s important to have opinions and quite frankly, I know I’m the host and I do have my own opinions, I share yours. I think there’s a reason why there are folks with advanced degrees in the world of data science, and there are also some folks who don’t have advanced degrees who are also very, very, very smart. And, if I were your father, you would not have let me down, Heidi. I think you’ve done incredibly well in your career, but these people they’ve worked extremely hard at their craft. I mean, they’re craftsmen and they’re experts. Sometimes, if you’re building a model and you’re betting your business on it, or you’re making important business decisions on it, you would want to have your experts standing behind it.
Now that being said, I’m going to talk out of both sides of my mouth here, when we were just talking about data literacy, I think also the goal there is to get folks on the business side, who are not terribly well-versed in the world of data and data science to get a little bit closer and meeting us in the middle a bit. So I think tools out there that help the citizen data scientist, I think can help bridge that gap even further. But I think you’ve got to draw the line on how far you actually want to take some of the work that your citizen data science are doing and what problems you want to put them up against. You don’t want to put them up against million dollar problems you want to put them up against fairly basic problems.
Absolutely.
And you always want to have experts on the sideline helping them out.
Yeah. I’m all for more transparent and easy we make data and analytics to consume and access, then that’s the mission that I’m on. I want people to get value out of this, and I think it’s an incredibly important part of their job and how they need to make decisions. And so I want to make that as easy for them as possible, whether they’re a data scientist needing data to build a model or somebody who needs the output to improve how they do their job.
Well, I think we’ll leave it there, Heidi. I’ve learned a lot today and I think the Fitch Group is in good hands. Thank you very much for joining us today on the Data Science Leaders podcast. I really appreciate it.
Thanks, Dave. It was fun!
29:22 | Episode 16 | August 17, 2021
40:04 | Episode 15 | August 10, 2021
38:29 | Episode 14 | August 03, 2021
26:54 | Episode 13 | July 27, 2021
Use another app? Just search for Data Science Leaders to subscribe.
Data Science Leaders is a podcast for data science teams that are pushing the limits of what machine learning models can do at the world’s most impactful companies.
In each episode, host Dave Cole interviews a leader in data science. We’ll discuss how to build and enable data science teams, create scalable processes, collaborate cross-functionality, communicate with business stakeholders, and more.
Our conversations will be full of real stories, breakthrough strategies, and critical insights—all data points to build your own model for enterprise data science success.
Host, Data Science Leaders
