Usama Fayyad is chief data officer and group managing director at Barclays. His work includes building and operating data infrastructure and customer/client value-driving applications in the group. He holds over 30 patents and is honorary chair of International Conference on Knowledge Discovery and Data Mining. He is also executive chairman of Oasis 500, a startup incubator in the Middle East.
I caught up with Usama at Sibos 2015 in Singapore to ask him about big data, automation and the future of fintech.
I know you have a very busy schedule at Sibos, including speaking sessions, and it’s great to have your time to talk fintech. Can you talk a little about your role? We know you as the first person with the title of chief data officer, ever, but it will be great to know more about your role at Barclays, your priorities and current initiatives.
I am group chief data officer at Barclays Group and I work with all the business units, including our Personal and Corporate Bank, Investment Bank, Barclays Africa and Barclaycard. I run the teams that handle our data systems and try to elevate our data systems so they deliver insights to businesses. The other half of my work that’s now becoming much more important is around data governance, data standards and data architecture.
Over the last couple of years at Barclays, we’ve learned that if you don’t get data right, you get a lot of unnecessary complexities downstream, whether you’re trying to do risk, computations or compliance. The messier your data is, the messier your systems are. Great simplification can be achieved by getting the data right, and it can help drive a lot of revenue by understanding our customers better. In fact, our CEO of our Personal and Corporate Bank, Ashok, always talks about the customer-to-client continuum; about clients who are institutions, who are actually made up of customers, and understanding that interaction is critical to improving service.
You touched upon some key topics in your speaking session at Sibos, about using data in the right manner and about data privacy. Who owns the customer data and who’s responsible for data? Who should be?
Who should be, so that’s a big question! As a bank, Barclays, generally speaking, has a very conservative view on data privacy. Even if we feel the data belongs to the bank, we always question it. Compared to my previous industry and the internet space, the approach is much more conservative and much more on the side of caution.
We have three basic principles when it comes to privacy. The first one is about opt-in: if we’re going to use the data of a customer or a company, there needs to be very clear disclosure and very clear opt-in. A clear opt-in leads to the second principle, that you have to have an economic exchange of value so when a customer or a company actually opts in, they have a tangible benefit. I will give you a quick example: In the personal bank, we have relationships with certain providers such as travel providers, energy providers, and so on. The way we utilize the data and allow our consumers to benefit from it is by saying, for example, we have these relationships and if we think from your data that you’re interested in travel or you travel often, and if you book through let’s say Expedia, you actually get 5% cash back in your account, once you follow the links through us and do the trip. We will say, we can actually save you direct money on your bill with the energy company if you actually go through this relationship.
The third principle is one where we always stay on the side of preserving anonymity. One of the services we provide to our small businesses and our corporate clients is insights into what their customers are doing. Let me start with a public example of members of the parliament, where we actually generate reports in the UK to talk about their jurisdictions and spend data in the local ecosystem. Barclays sees so much of the spend data because of our point of sale presence and our credit card and debit cards (in the UK, consumers use a lot of debit cards, much more than credit). We actually get a nice view of this spend, the trends, where the consumers are going, what’s picking up, what’s coming down. We provide that to the members of parliament as a service that says, within your district, this is what’s happening – there’s more spend on these categories, and so on. But if we report the average or aggregate spend dropping below 50 individuals, when usually they are in the thousands, we just don’t report, because the fear is that at a level of less than 50, you may be able to start triangulating on one individual.
I can tell you theoretically from a data perspective that limit is too conservative. My usual argument is you’re safe at five or 10, but being a conservative 300-year-old bank, 50 is extremely safe.
That’s hard, because there seem to be so few services that provide an opt-in-based customer experience.
Having worked in the EU and the US, one of my well-learned lessons is that opt-ins are one of the most defensible mechanisms, as the consumer is choosing to do something, and an example is frequent flyer mile programs. Technically, you’re telling the airline, track me and it’s OK for you to know when I take off, when I land, where I’m going, where I’m staying, and so on, in return for an economic benefit. Not only that, the consumer will go out of their way to make sure and check with the airline: have you got the right number, are you tracking me, are you tracking my children, and so on, because there’s a benefit.
In my experience, both with technology companies (whether it’s Microsoft or Yahoo) or my own startups, this is the ultimate defensible mechanism. The EU has the best standards when it comes to privacy because they insist on opt-ins and they insist that that opt-in is meaningful; that there’s some return given on the value against it.
It’s interesting you talk about data as a service to your end-consumers, but data tends to become quite tactical in conversations, and is about the infrastructure, the mining and the big data. What are the value-added services or use cases that you can create that could be exciting for the average consumer?
I think big data is huge, and especially in banking – sometimes a late adopter of technology – there are huge economic drivers of big data as a technology. Then there are strategic drivers, such as the MP reports we talked about. That’s powered by big data technology. I also want to talk about the tactical aspects I think the market doesn’t talk much about.
The reason big data matters to a bank isn’t just the technology, not the fancy stuff. It’s much more basic. Cost of storage in a typical data warehousing scenario is extremely expensive in a bank. The cost per terabyte per year can be brought down by a factor of 10 to a factor of 40 and 50 spending on the platform, just by going to something like Hadoop, which is open source and most importantly utilizes commodity hardware and storage. I’m unaware of any financial institution whose data appetite is dropping. All businesses are going to have more data next year than this year. In fact, next week compared to this week.
Number two is banks have traditionally paid a lot of money on ETL – extract, transform and load – very expensive software to access your own data; to move your own data from one database that you own to another database that you own. What’s happening with the big data technology is that not only does it allow us to do this in a flexible manner, it allows us to do it on many more data sources, such as documents, images, recordings from our call centers, and so on, that traditional ETL tools – where we spent tens of millions a year (if not more) – cannot do. So those are economic drivers of why we should care about big data.
The third one is the strategic one, which is the enabling of capabilities that we didn’t have before. One of the big technologies we’re excited about is our Fusion Cell, and our ability to bring data from multiple sources and make it available. In the industry, we hear about ‘data lakes’, where people just dump data together. They become toxic dumps because you no longer understand what’s in them. It’s a mess. Taking the approach where we systematically (in a governed way) allow our business units to sort of fuse data together, yet not lose the data architecture and the understanding of what’s in the data, is I think a huge game changer that allows us to avoid a lot of what has happened in other institutions. You sometimes hear about these big data projects that went awry and cost tens or hundreds of millions
How does Barclays work with fintech startups?
I think Barclays is a pioneer in this area, and not only do we embrace working with startups, we have these ‘open innovation’ days where we invite startups to come in. We review with them, open up problems for them to solve, and they get to pitch us on solutions. We have a unique reward system where we reward our managers for working with startups, and this started with our accelerator program called Rise, and we have accelerators in London, Birmingham and New York that are very successful. We actually measure how many managers from within Barclays mentor these startups, and we use a score card matrix to reward them for citizenship. In my world, citizenship is about working with startups.
Another thing we do is the measurement of the effectiveness of our accelerator with how many contracts the company gets. Our first cohort was 10 companies, and by the time they finished their three-month residency, they actually had six contracts with Barclays, which, to a startup, is more important than funding or mentorship. The effect of that is actually magical, because it’s a way of injecting innovation into the bank, of forcing our teams to be less insular and less risk averse. The groups and managers who actually did contracts with these companies get rewarded. That’s our way of helping the economy by working with more fintech startups.
We’re very, very interested in making London a big fintech hub, which it already is. We want to grow that more. We think the conditions are right with all the banks and the financial industry up here. It’s a more interesting environment than Silicon Valley, for a fintech startup, and by embracing those technologies, we actually address a lot of the problems that we have.
What areas of fintech are you personally excited about?
One is automation, as I think so much of what we do in banks can be done in a much more automated way. We worked with three startups who specialize in the area of artificial intelligence and machine learning, to help apply them to our own operations. If your job is to do SSI (standard settlement instructions), you’re filling out these forms all day long – you’re getting a fax, you’re matching it to the records, and you’re typing up numbers.
As you go through doing the stats automatically without any intervention, without changing how you work these technologies, actually watch what you do and very quickly learn. Ah-ha! When I see a PDF or a fax of this format, they’re looking for these fields and it starts automatically highlighting which field needs to be filled out and starts automatically filling out the fields, and the colleague just has to confirm. Once it’s very sure, it goes through ‘straight through processing’, which is almost an infinite speedup. It goes from a speedup of 10x when it pre-fills the form, to a speedup of a 1,000x or 10,000x, when it automatically does the straight through processing. So, this is one area I’m pretty excited about.
We’re also interested in many predictive analytics companies that allow us to better guess or recommend the next best offer to make to a customer or client. We also utilize the technology inside our internal risk finance and complaints stuff, so you can actually recover a lot of lost RWAs and Risk Queries Assets and help control capital better.
It also helps on the automation side and on the customer experience side, with the right recommendation and the right communication, and on our functions, plus risk and capital reserves.
So where do you think fintech is going? When we come to Sibos next year, what do you think the discussions will be about?
This year at Sibos, I spoke at a session on digital identity and the role of banks in it. There have been global surveys on who we trust the most with our digital identity and the results were actually surprising to me, as banks are No 1, ahead of the government. Interestingly, social networks come at the very end, even though those are the most ubiquitous authentication methods, even used by several ecommerce players. They use it, but they trust it the least.
Banks are in a position to offer identity as a service thanks to our onboarding process, and this is a huge area of interest. Solving identity in a digital world is one of the big fintech challenges, especially in a world of the Internet of Things, where devices have to know who they belong to and work on the owner’s preferences.
I think automation for banks is a huge deal, so utilizing these AIs and machine learning technologies to do automation of tasks that we normally do with people is a huge change.
One of the big surprises to me in coming into the banking industry from the internet world is how little banks leverage their ability online, to do things such as cross-sell and up-sell effectively, and ‘next best offer’. Those areas I think are big challenges for banks, where fintech can help.
In the area of data, the grand category I use is regaining customer or client intimacy through data. If you go back even 30 years, banking was much more local, with the branch manager and the staff pretty much knowing who the customers were and who was going through what; buying a car, buying a farm, losing a job, getting a new job, graduating, and so on. As you scale and you go digital, you lose this intimacy, which means you lose the knowledge. Things such as risk, credit or KYC become a challenge and an expensive activity. If you actually leverage data correctly, it’s a win-win and you regain the customer intimacy.
I think that whole area of leveraging data to regain customer intimacy is a huge area, where fintech – especially artificial intelligence, machine learning and algorithms – can play a big role.