The CDO to CDO podcast is hosted by Chris Knerr, Chief Digital Officer of Syniti.
Subscribe Here: Listen to all the episodes in the Syniti Podcast Series
Chris Knerr (00:23):
Hello, and welcome to the CDO Magazine interviews series. I'm Chris Knerr, chief digital officer of Syniti, a world leader in enterprise data software. And we're partnering with Chief Data Officer Magazine, MIT CDO IQ, and the International Society of Chief Data Officers to bring you the series of interviews with global thought leaders in the field of data and analytics. Today, I have the pleasure of talking with Diego Vallarino, chief data officer of Scotiabank. Welcome Diego.
Diego Vallarino (00:52):
Hi, Chris, how are you?
Chris Knerr (00:53):
Excellent. Great to see you. So Diego, you've got a really interesting background and I thought maybe just as an intro for the audience, you could start with sharing a little bit about your career journey, how you came to Scotiabank. You have a PhD in a field called cliometrics, which I understand is a subfield of economics and econometrics. So maybe you could just share a bit about your journey and how you ended up being chief data officer of Scotiabank.
Diego Vallarino (01:23):
It is a particular journey. I study a Ph.D. in cliometrics. Cliometrics is a combination of history, economics and econometrics. I love history and economics and mathematics and statistics. And I start to work for different companies in software sector. And after that, I start to work as a director of data analytic for Equifax. Equifax in Uruguay have a huge operation work for all of the different player in the markets; banks, retails, Telco.
And in this position I understand really how to work with data, how to manage different project with data and start to think to change my career to a position more related to internal entrepreneurship. And I accept the proposal for Scotiabank to develop a new area in analytics. And for me was a very interesting opportunity to apply not only what I know about consulting, but also to make decisions related to data to business and so on.
Chris Knerr (03:04):
Interesting. So yeah, there's a lot there that hopefully, we'll get a chance to explore a bit further. So let me start out with something that is kind of top of mind for me in the work that I do and I think top of mind for a lot of folks and our audience, which is you've had these different experiences: academic background, consulting, internal entrepreneurship, a variety of global companies, how do you think companies can best add value through data and through taking a more strategic view of data?
Diego Vallarino (03:37):
This is a very interesting question that we have to answer and the answer change every day. Different focusing in value depends on the sector of the company or dependent on the sector of the economy which the company is on. It means if you talk with the C level, they understand that the data is very important and they know that the data is [inaudible 00:04:07] for example. But when you start to talk with them, they don't know exactly how to implement this strategy and how to impact in the value of the company, the value related to revenue or related to a user cost.
For me, it's very interesting to use a framework, perhaps you know this framework is a combination of, four D plus four E. For me, it's very important to explain this for the management that imply that you have four D related to data science. The first D is related to define the problem, the pain point, and defining the company. The second D related to data. You have to understand that you have data inside the company, but a lot of data outside the company, structured and unstructured, that develop value only to access this alternative data.
The third D is develop; is related directly to algorithms and models and so on. And the last one is delivery. Delivery depends the channels you will use this model. It means you can use in an app on the smartphone or in the web, or perhaps one and once when you define the commercial strategy for example, I don't know. But you have to combine these four D with four E. That means that this all have to be easy to implement, easy to use, easy to revenue.
It's very important to understand that the algorithm is very fashion and the company don't understand they don't use. And it's very important to implement easily to use and to make revenue will ethics. For me, ethics is very important. I know you share this idea and it's very important to improve the ethics in the data science team.
Chris Knerr (06:45):
Yeah, interesting. I liked that model a lot, and I think that just maybe a couple of things to draw out for the audience, and this is very familiar to me in my work as well, is one is that very keen focus on business value and on revenue. And I think you articulated that in both the Ds and the Es, defining what that is, what the pain point is, and then ease of generating revenue or generating some kind of measurable results. So I think that's a really key thing to highlight.
The ethics thing I agree, and I think that might be given that you're in financial services. And there's just so many different aspects that I think your industry is ahead of many others in terms of dealing with the impacts and kind of having a framework. So I think we should talk about that a bit more. But maybe before we come to the ethics, there's a couple of points you made in there which I think are very important.
So in my experience, senior executives understand how to run a business. They understand what a value prop, their eyes sort of glaze over when we start talking about algorithms and models. So I wonder in terms of as a senior executive, but then with this career progression, are there some good warnings you have around how to explain the work that we do and make it intelligible so that the senior folks kind of know enough to be supportive, but don't get lost in the details too much.
Diego Vallarino (08:24):
Yes, it's very important too, for the CDOs role to understand that you have to explain difficult issues in the simple, keep simple. For that it's very important to translate then what we do to the business worst. It means we have to talk out revenue. We have to talk about cost. We have to talk about value as you mentioned before, that's the way the management team understand how we work with data and nobody at the bank or Equifax, or what else asked me which algorithm I use. Nobody asked me which data we use. They asked me how improve our revenue or how we can reduce the cost or do two at the same time.
It's very important to understand that. If we talk, you and me about algorithms, it's one way but when we go to the board, we have to talk about business; how they make the decision, how to improve informing the decision. For me, it's very important to keep simple and talk about business, not only data or algorithms or something like that.
Chris Knerr (10:19):
Yeah, I think that's very good advice. And for our audience, this is a theme that's come up in a number of my thought leader interviews in the series here both that focus on business value and explaining things and kind of terms that business leaders can understand. But there are a couple of things though, I want to connect that I think you'll have a really interesting view on.
So let's come back to this question of data ethics and the regulatory landscape and I guess, a couple of things that would be worth talking about and we can kind of take these one at a time or you can pick in any order you like. So one is as a financial services' company, I think you're on the forefront and your industry has done leading work on PII data, data privacy, data security. There's another whole consideration and kind of what triggered this thought is we're saying, "Well, we want to explain things in business terms to senior executives, but now there's this whole question about accountability and bias in algorithms."
So maybe if you could share a couple of thoughts, one on the overall direction of regulatory, what you see in financial services and maybe what other industries can learn, and then kind of, what's your perspective on how you start building organizational and also technical capabilities that can address issues of bias switch. And just for the audience, I'll explain this hopefully correctly, for example, in algorithms that do credit approvals, there's a significant concern that people don't understand how those algorithms work and that they may be including an implicit bias by the designers of the algorithm or in the data.
So this is kind of a huge and really interesting emerging area. So I know that was about 18 questions wrapped into one, but let's sort of see if we can start peeling this apart one way at a time, because I think there's a lot here that's of really high interest in value.
Diego Vallarino (12:36):
Yes, it's very interesting the issue about the regulatory and the ethics and biases and decisions. And particularly, in the financial services, on health healthcare, on Telco have the same problem. And it's very important to divide the problem into regulatory and ethics. Regulatory is clear today after the European Union have a new law, a new regulation about privacy. We have to work to design a product and algorithms and with this in mind, for me, it's very important to work with a lawyer closely to design new product, to know algorithms, et cetera. But on the other side is ethics.
Perhaps you have the okay to use the data, you have a very interesting algorithm, is a white box and so on. But when you make a decision about the data you have, you have a problem with privacy related to the life of people. Perhaps you are okayed for the law, perhaps you are okay for the regulation, but you make a decision as a manager which remain in the private life of people. For me, it's very important.
My team have these in mind every day. Okay, we have the data, we have the algorithms and which decision we have to make. I offer this product or not. I am very careful for these because you can improve the value with data, improve the value with algorithms but you destroy value a bad decision with this data. Reputation values very important in this day, and perhaps you destroy a lot of value a bad decision.
And if you can see biases in this models or algorithms and so on, you have to think that if you don't fix this biases, you will have a wrong decision to make. When you design algorithms and you find biases, the bias is how people interact each one. And you have the opportunity with these tools, with data management, analytics and data science, to reduce the social biases.
For me, it's very important to improve this, perhaps society by data science. For me, it's a very important point, very important issue related to the impact of data science in society. In my team, you have two moments that analyze biases. The first one is when we access new data and analyze the data. If the data is okay, have biases or something like that.
And the second is when we finished the development of the algorithms, we have to analyze different variables and understand if the variables is correlated or implied. If their variable imply consolity, is not a bias is, it's okay if you have a good algorithm. But if you have a different value base that imply correlation, but not a consolity, perhaps you have a bias in the algorithms and you have to improve this algorithm.
Chris Knerr (17:20):
So that is really fascinating. And I think extremely interesting and relevant to our audience. I want to just unpack a couple of things and make sure that I understood some key points. So I think one point that I take from that is that, and I think I was saying this in a loose way, like regulatory and ethics. I think what you're suggesting is that as business and data leaders, we're at the forefront of looking at the ethics.
Regulatory in a sense is ethics that have been codified so they're known, and they then have legal implications that we have to deal with differently, but we should take a view of ourselves as leaders in the industry as being the leaders of ethical use of data and not count on the regulators to do that, the regulators are going to catch up jobs.
And I think if I have that correct without kind of calling anyone out, I think there's some very interesting and well-known examples of technology companies that perhaps have not done as good a job as they could have, and that's led to unfortunate results socially. And I also, and I think this is important and reputational damage for the companies that have not done that.
And again, I'm sure everyone can speculate about who some of those companies might be. So that's kind of a big business, social leadership point which I think is absolutely key. That's sort of at a macro level. At a more execution level, I heard something very interesting too. And I want to just translate this into my own words, maybe for folks in the audience who are less technical.
So almost like what you're saying is that there's a capability to look at better performance of algorithms and better ethic. Almost like the way in a factory, I inspect my incoming raw materials to make sure they're the right quality. That's the moment of looking at bias and the underlying datasets.
And then before I shipped something to a customer, I put it through a final quality assurance step to say, "Okay, this meets all my standards." And that's the same way at an execution level that you've kind of coached and built a capability for your team to work with the algorithms and the underlying data sets.
Diego Vallarino (19:47):
Yes. It's very important what you have said. We have to work closely with our teams to understand this is not only the point of view, but also the ethics. The ethics for me, is very important in our job as our data management and really important. Because we have a very important role in the today world. We work with a lot of data than you normally be, you will have more data. And we have to understand that the ethics related to this data management is very important.
Chris Knerr (20:40):
Yeah, absolutely. I think that's so interesting. And that's again, given some of the challenges, I think we've had probably more in pure play tech companies than in industries that are starting to learn how to use data, but you think as these are kind of getting into mainstream business culture, that we can apply that kind of right ethical framework and leadership. I think that's amazing that you're doing that.
Let me shift gears a little bit and kind of come back to part of the first part of our conversation which is around, if I can distill what you said and I believe this, and I've heard this from a lot of thought leaders as well, is focus the data work on valuable business outcomes.
Now as we discussed briefly at the beginning, you've worked in a variety of roles within financial services and a variety of different kinds of companies, are there some characteristics you think of companies and leaders that do data well, how can we predict who are going to be the winners and who are going to be the losers in terms of capabilities or mindset that those companies should be building?
Diego Vallarino (21:55):
Another interesting question.
Chris Knerr (21:58):
I saved all the easy questions for my other interviews, Diego.
Diego Vallarino (22:03):
Thank you very much. For me, the winners and the losers depends directly on how to use data and frameworks related to business in this area. And perhaps in healthcare and our industry is different, but in business, you have to combine this. And for me, it's very important to understand the behavior related to the companies or people. That imply that the the data is an instrument and the software is the tools, but the objective is to understand their economic behavior, their social behavior of people for example.
For me, it's very important number one, to understand the financial health of each client, to improve its health related to the financial decision they have to make. If I understand the behavior of the people or the companies using data is for me, the winner will pass to the industry. Today at Scotia, we work and focus to understand the behavior of little companies and the behavior of people in COVID-19. And you have to know that the behavior is changing in the last six months, we change and our behavior change, and you have to understand this using data in our case. It's very important to relate to behavioral economics with data.
Chris Knerr (24:13):
So I want to stay with that thought and just you can probably add more of this, but just for folks in the audience who aren't familiar with behavioral economics. Behavioral economics is the branch of economics that deals heavily with how decisions are made in the real world. And in particular with kind of back a little bit to the ethics discussion, how different biases creep into decision-making. Probably one that most people have heard of is a thing called confirmation bias, which is our tendency to like evidence that reinforces what we already think and ignore evidence that contradicts what we already think.
So just to unpack what you were saying a little bit, if I think about sort of the data mind and the data mindset, are you suggesting in a way that part of what makes a winning organization around data is not doing that. So actually, looking at the data in an objective way and letting the data drive the decisions, rather than sort of some theory that some executive might have.
Diego Vallarino (25:23):
Yeah, exactly. You have to understand the company strategy and you have to improve its strategy with data, with analytics, with machine learning and different tools. But you have to align the data strategy with the business strategy. It's very easy to say it's very difficult to implement, and you have to work hard anyway.
Chris Knerr (25:58):
And this is a really interesting topic. So in my professional career where I feel like I've had a collision around this is where I've been involved in analytics work that let's say, impact the awarding of incentives. So, I mean, I won't pick a real example, but you can imagine that you have say, executives in a sales organization who are getting commissioned based on metrics. But then you do the work and you find out that the metrics are wrong. The metrics have been designed poorly or say back to our conversation, there's a bias in the underlying data.
So does the chief data officer or should the chief data officer over time, or even now start to take on a role and looking at how those incentive systems are designed? Just in simple language, what I've found like mostly the reason that people fight about the metrics is because somehow it's threatening their incentives, which is a very much of a behavioral economics type of question. I'm really curious what your view on that kind of our role in even incentive design itself should be.
Diego Vallarino (27:14):
Yes, it's really very interesting issue. And incentive for me, is very, very important and because we have to understand that we are in a changing world, in a changing business, and we have to align the incentive to that change. I mean, as a CDO, we have to understand that different players in the company have different incentives to do different tasks. If we don't understand that in economics is related to political economics, mapping the company and identify which people have which incentives to earn its bonus.
Chris Knerr (28:19):
That's it, right. Yeah, it's all about the bonus.
Diego Vallarino (28:22):
Yeah. It all depends on the bonus. And if you help each one to earn its bonus, you have a very friendly environment to develop the data strategy and you have to develop different algorithms or data management and you will have blood shed and you will have quick wins because the different figurative implement your models with data. And perhaps in our world, we don't take care of how the incentives, we focus on technology or different instrument, but not in the political economics related to the organization. For me, it's very important.
Chris Knerr (29:26):
So just to play that back, you think that it's super important to understand the incentives and basically who's going to get what bonus based on what they do, but maybe at least for the moment, maybe in five years, we'll be having a different conversation and I think this is important. This is a relatively new function in large organizations. So in a sense and maybe we'll come back to this, or maybe we kind of touched on this, there's a lot of education that has to be done.
Everyone says they're data-driven, but almost no one is, has been my experience. "Oh, we're a data-driven company," but then you go and you talk to people and they're doing their confirmation bias. They're doing what they've always done. So your view would be kind of with a pragmatic lens. We're doing the data work, we're doing the algorithms, we're looking at the data, we've got a story about business value.
We understand so that we can have a smart conversation with stakeholders how that may impact their incentives, but maybe it's for the next generation of chief data officers to start playing a more economic role in the definition of those incentives.
Diego Vallarino (30:38):
For me, it's very important the next wave of chief data officer is related to our machine learning. The technology will do the different Ds that I mentioned before, develop the algorithm. But the CDO for me, it's a role that is permanent in the midterm, I think, because you have to understand the role inside the company to understand the people, the client, the customers, how to transform customers or client in customers.
It's very important to understand the cost structure of the company and how the data or analytics can improve their use of the cost. And it's very important to understand the soft skills of the role. Not only technology, not only statistic, but also the soft skills; creativity, leadership, innovation, for me, it's very important to a person who wants to work as a CDO in the next year.
Chris Knerr (32:10):
I think that's really a really important set of points because it's interesting, you talk to different people in different industries, in different companies. And I think some companies view this very much as an IT, a techie thing. And I think what you're saying and I very much having worked with lots of companies, I prefer this construct that you're supporting which is that chief data officer, it's a combination of understanding the markets, understanding the business, understanding data and technology and providing decision support that kind of brings all those together in a unique way.
That's part of why I agree with your thinking that this is a role that's... I know there's been some writing on like, "Has the moment passed on my personal view?" I don't think the moments passed at all. I think we're kind of at the point where this role is getting better, understood and better established in larger organizations. And to earlier in our conversation, organizations that do this well are really starting to just see the value and see benefits of having people who have that kind of unique combination of skills. So I think that's kind of great advice for folks in our audience who are kind of mid career and who would aspire to have a role like yours.
Diego Vallarino (33:31):
I think it's very important to understand the combination of data and tools. Perhaps you make a lot of value for the company with Excel, perhaps. I appreciate you explain a very interesting point that make new revenue and you use Excel and you don't use algorithms, Excel. I appreciate you explain that for the management and they understand because they use Excel and you accomplish [inaudible 00:34:08] data strategy.
Chris Knerr (34:13):
Well, it's interesting. I think that's important, because I think we haven't talked about this much. But in my own experience, and again, almost all of our colleagues that I've spoken to are very keen on an agile approach and taking kind of small wins that show value rather than do... And that's another way that this is different than sort of traditional big guy IT, which actually formerly was my own background that you do projects that take two, three years and cost hundreds of millions of dollars.
The cool thing about the data space is there's so many opportunities. And actually, I think it's a good point. It's fine to start in Excel. You can do you can do a lot in Excel, on Hive, on NoSQL, on a little Python coding. And then if you're a big shop, you can kind of bake that into a broader workbench.
So speaking of projects one thing that I wanted to highlight and for anyone who has in our audience, I'd encourage people to check it out. Scotia recently won an award for innovation in digital banking for your pandemic crisis response. So could you tell us a little bit about the award in a directional way, if you're able to share what your team's role was that I think that could be very interesting. And this is something you can look up on the internet for folks, if you want to, you know get more information, but I'd love to hear a little bit about that award and what you guys did to enable that.
Diego Vallarino (35:51):
Yes. Scotiabank won a recognition about the improving in its digital channel, it's very important for us because, we understand when our client want, they want to improve the transaction and the speed and the quality of digital transaction. Particularly my team here in Uruguay, the challenge was very important because we have to work in different areas in different issues first related to the spiel of the answers that the Scotiabank give to the market. [inaudible 00:36:45], don't work with machine 10 years ago. Today, the client go to the internet, apps on their smartphone and ask for a grant or ask for a loan.
And it's very fast and change the way we work with clients that imply automatization, different model, operating model [inaudible 00:37:29], the delivery will different channels; smartphones, apps and so on. The other thing is very important is to understand the new client. For us, it's very important to understand the different behavior, economy changed in the last six months and we are conscious of that. And perhaps, one client in the end of last year want to buy a new car and today they have to consume. We have to understand this different behavior, consume different goods, not only a car. And we have to understand this in a huge amount of people.
And the third is to apply a different price model. The price for me, it's changing very hardly because we consider the value different of the price and the value for different things change. The COVID-19 change our point of view for different issues for different behavior for different consumptions. And we have to understand that, and we have to understand that the price related to the value of different issue.
For example, in real estate in industry and we work closely with different players, we understand that the people want to go outside the city, a house, different apartment, perhaps last year was different and we understand this, and we have to develop new product and develop a new challenge and develop new price for these product.
And my team have to understand this with data and the data come very fast because all of the client use digital channels, and we have to process this data. And we have to develop insight about it. We have to develop a new product or collaborate with product to design a new series product and so on. For us, it's very important to change our mindset in the last six months.
Chris Knerr (40:28):
It's a fascinating project from what I read and just I mean, a couple of things just to kind of draw out of that. I mean, this was a lot of work that was done very quickly in retail banking, which is complicated. Back to some of the things we talked about, there's a lot of data, there're regulatory complexities, suddenly there's all this new demand for different kinds of interaction. And this has been written about a lot, but what I think is cool about this is, it's often been said this year that we took five years of digital progress and compressed it into less than five months.
So it's very impressive that you guys were able to do that. And then it just draws on some of this, the agile thing that we talked about working with complicated data sets. I liked that you highlighted kind of the pricing aspect of it. Because I think if I understand correctly, you wanted to both help people, help customers, real humans who have financial problems because of the pandemic. But at the same time, Scotiabank is a business and it still needs to make money.
And what your team was able to do was sort of bring those two pieces together. And then the other thing and I'd encourage everyone to read about this project because it's an impressive award and it was a really cool project. I think two things happen. One is you did a lot of complicated work very quickly, and it almost seems like it came out of nowhere, but in fact, your ability to do that was in part, because there was an executive commitment been about three years ago to putting in place organizations like yours. So in a sense you proverbially, you made your own luck in this.
And in fact, if that program hadn't been there, I doubt you would have been able to do all this so quickly. So I think there are a lot of good lessons for industry leaders and sort of thinking about like, look, this was event driven, but on the other hand, the bank's leadership had enough foresight a few years ago to say, "We need to start building these capabilities so that we can do better work, have higher returns and do a better job for our customers." And then, there was this kind of a remarkable opportunity where a lot of these things came together with a great result that we saw.
Diego Vallarino (42:54):
Yeah, exactly. Particularly in Uruguay, Scotiabank can respond very quickly because we have a team analyzing data. I worked for one year and a half to develop a team. And this situation imply that the team was consolidate and have a different process, understand that the business and accelerate the digital channel. But if we don't have the team, the process and the area, I want them to think about it.
Chris Knerr (43:48):
Yeah. Super, well, I think maybe that's a good note to kind of wrap up on, are there any final thoughts you'd like to leave our audience with, major predictions in data or data technology for the next five to 10 years? Advice for mid-career data professionals that we talked about, any final words of wisdom that you'd leave us with.
Diego Vallarino (44:16):
I will try to say something interesting. I don't know, the evolution of technology in the next three, five years, but I think that the people who want to work in this data area have to understand the landscape they are working on. It means they have to understand the instruments, the tools related to data, but also to understand that the business, the landscape you are working on.
If you work in a retail or if you're in financial services, you have to understand the business strategy. You have to understand the soft skill you need. We talked about the innovation, leadership, negotiation is very important to work in data. Perhaps it's not common to understand when you study data science that you need a negotiation or creativity or leadership, but when you start to work, you will need.
Chris Knerr (45:32):
Yeah. I think that is a great summary. And I think of a couple of maybe major points that I can just try to synthesize. That was a great summary of that kind of role description for what makes a good chief data officer, which I think is really important. And then couple of other things just to highlight that kind of focus on business value and business language and not data language.
And then the other thing that I thought was just fascinating and I just really appreciate your insight into is this idea of the importance of business ethics, how it's really on industry to lead that and that's going to end up being the regulatory of the future, but we have to pioneer that within industry. So Diego, can people find you on online, on Twitter and so forth?
Diego Vallarino (46:26):
Yes, of course. In Twitter and in LinkedIn, all of us, we are in the internet world.
Chris Knerr (46:35):
Yeah. Perfect. Awesome. Well, this was great, Diego, thank you so much for joining me today, for our audience, I hope everyone found this as interesting as I did. We have some additional interviews @cdomagazine.tech, but I have to say this was one of my personal favorites. So thanks very much. Thank you, Diego. I appreciate it and hope everyone has a terrific day.
Diego Vallarino (46:59):
Thank you, Chris. It's a great chat with you and thank you for the opportunity to present my work.
Chris Knerr (47:06):
Indeed, thank you.
About the AuthorFollow on Linkedin More Content by Chris Knerr