“The title ‘Chief Data Officer’ started out as a joke!”, says Usama Fayyad -the world’s first CDO- and continues; “Now it became a serious thing, I guess”.
Kate Strachnyi interviewed IADSS Co-founder Dr Usama Fayyad after his keynote speech at the ODSC 2019 East event. An important part of this funny interview was about the evolution of the title “Chief Data Officer” and the role it plays in driving data literacy and culture in companies. This interview is also published on IADSS YouTube channel. See "Data Science Career Talks - Usama Fayyad".
Kate Strachnyi: You were the first Chief Data Officer at Yahoo, and more recently, you were the Global Chief Data Officer at Barclays. Can you tell us about how these Chief Data Officer role has come about and how it has evolved? I know you said it started as a joke. Tell us.
Usama Fayyad: Yeah. It started out with the executive team and Jerry Yang assembled to welcome me to Yahoo, and asked, "Okay. What should we call you here?" And the title Chief Data Officer came up, and everybody in the room actually starts laughing, they thought that title sounded funny enough. Yahoo actually had this culture for creating irreverent titles, Jerry Yang called himself Chief Yahoo, for example. So that was the first definition of what it means, and we also defined Chief Data Officer as an operational role that had to worry about standards, data governance, making sure everything is taken care of and well-architected, and so forth. So it had this dual role for me, at least in the beginning.
Five to six years later, I left Yahoo and Joined Barclays, which is a 325-year-old bank, one of the major banks in Europe, based in London. There, the contrast was significant because of how the field had matured. I remember the first time I gave a public talk as a Barclays executive, somebody walked up to me and said, "You know there are about 1,000 Chief Data Officers in the industry?" Of course, at the time, I didn't say, "Oh, yeah. This title CDO started out as a joke." Nowadays, a lot more organizations took it seriously. A lot of organizations realized the need.
At a bank, there's a lot more emphasis on the data governance side, on the policy side, and a lot on the privacy and the proper use of data, and especially for a European bank, where a lot of the laws are a bit more evolved and advanced than in the U.S. in terms of how much data you're allowed to keep, how much data you're allowed to use, etc. The differences, I would say, were very significant, in terms of what the role is like, and where you spend most of your energy.
Kate Strachnyi: Does every bank, at this point, have a Chief Data Officer role? Are they mandated to?
Usama Fayyad: It's almost getting there. I think many banks fear getting in trouble if they don't have a Chief Data Officer now. The industry basically demands an accountable individual, who is looking and seeing whether that organization is using the data properly, how they're using it, do they have the right limitations on the usage of data. I used to call my role at Barclays the responsible use of data, but before you can use it responsibly, you need to make sure that it's usable, and available, and can be actually applied in a lot of these innovative applications as well, while staying on the safe side. So one of the big programs I pushed there was what I called KYC, which is in banking, is to Know-Your-Customer, a significant area of spending. I called it The Journey from KYC to UYC, where UYC stood for Understanding-Your-Customers, which actually shows the business value rather than just doing it for a regulatory reason, which is what KYC is all about.
Kate Strachnyi: You mention a skills gap, which is validated by The Data Literacy Project led by Qlik, who says that only 24% of the global workforce is data literate. What is the first step a Chief Data Officer should take to drive data literacy and data culture within their company?
Usama Fayyad: That's a great question, after engaging with some of the largest organizations where I was personally involved as Open Insights, we quickly realized that data literacy and data culture is a big part of making the data useful and usable. To that end, we actually launched something we call Data Academy in a lot of these big companies that target to achieve data literacy. There's literacy at two levels. There is what an average employee should know about data, why it's important, why it's important to be safe with it, why it's important to be sensitive to it and why it's important to make sure data is kept safe.
The Data Literacy Project led by Qlik, says that only 24% of the global workforce is data literate.
Then there is the part of what's possible with it, which goes more into the analytics. A lot of data analysts and business analysts don't even know the art of the possible on what you can do with data mining algorithms, with machine-learning algorithms. A great example in monetization is when I joined Yahoo within two years, by utilizing the right machine-learning systems, data systems, data management regimes, and some of the big data technology, we were able to generate, without much work, $800 million of additional revenue derived from targeting for Yahoo. Basically, they were selling the same ads at 10 to 20 times the price that they used to sell at before. Because now, they could actually do targeting that they could prove to their customers and say, "Well, this well-targeted. This is reaching the right audience, which results in a much better dynamic!” both with the advertisers willing to pay more, and the consumers being slightly happier. Nobody loves to see ads, but if you see ads that are relevant, it's a better experience than ads that are completely irrelevant or shown at the wrong time. This created a good dynamic that actually allowed us to create value. So it's very important to have that data culture, and that awareness, and part of that is due to both the data literacy as well as the data science literacy.
Kate Strachnyi: Let's say you have a big company like Barclays. How do you actually implement something, where let's assume someone such as the average admin or somebody who might touch on data once in a while, but is not really a data analyst or really specialized to work with data, how much do you teach them? How much should they know?
Usama Fayyad: Yeah. It has different levels. We ran these data academies at Barclays. We ran it at Barclays Africa, and at the MTN as well, which is Africa's largest telecom, we also ran it at many companies in the U.S. and in Europe. The way you do it is you create different tiers for different levels of awareness and different concepts that you want to emphasize. So for somebody who hasn’t touched on data a lot, we probably want them to be aware enough to understand that KPIs matter, which KPIs they should pay attention to. Our philosophy is every KPI or key performance indicator should have a level that goes all the way from the board to the CEO, all the way down to the lowest level in the company. If you don't have that, then something is wrong in the way you're-
Kate Strachnyi: You mean it should be the exact same KPI or-
Usama Fayyad: No, no. It's at the different level of granularity, right? The person in operations would probably want to see a lot more detail. The person on the board doesn't want to see any detail, but wants to see the big signal and where should they pay attention to what's going wrong, like which region is veering off estimates and so forth. That insistence, culturally, has a saying, "Every report has a version that goes all the way from the board to the lowest level, even though it may never reach the board." It instills the culture to think about it that way. Most of the training is about awareness, and as you get down to specialized roles, it becomes more technical and more around what's possible because those are the people who can actually make a difference, and help you make stuff happen.
I've been involved personally in many data science projects, where we would work hard to come up with an amazing predictor of something. One example that comes to mind was working with one of the very large car manufacturers where we were trying to predict the sales by car, by model in different micro-markets. The problem was very hard. We cracked it in a very innovative way, and in the end, we discovered that the executive team was unable to do much with those predictions, even though they were super accurate.
In those years, incentives on automobiles were a big deal. We figured incentives, where you pay consumers money back for buying a car, was the easiest thing to reprogram, right? You could change them by market, by demand, etc. It required us working with the executive assistant, who prepares the reports. So we worked with the Executive Assistant to take the spreadsheet that goes to the executive team, and permit us to color code certain cells as red indicating that you're probably overpaying. If they are green, they're great. And these need attention or change. Just from color-coding, we went from big predictions, where people were just watching them as a spectator sport to saying, "Oh, now, I can act on it. In this market, I need to change it up or down."
Now, it became actionable. That's the importance of involving everybody because the whole supply chain of data and the consumer chain is reliant on a lot of people doing a lot of different roles.
Help Set the Standards for Data Science & Analytics Professions
IADSS Research Study continues to set standards and framework for analytics & data science professionals. Click here to contribute to the research study and help us!
You may follow IADSS on Twitter, YouTube or join IADSS LinkedIn Discussion Group.
Comments