Monetizing Big Data: A Q&A with Wells Fargo's Data Chief
A. Charles Thomas is a rare bird in banking circles.
Nine months ago Wells Fargo made him its first chief data officer, and one of only a handful of such jobholders at banks nationwide. Thomas, who previously held a similar role at USAA, has a doctorate in behavioral science and manages a $100 million budget and a "small team" of 600 people.
In a recent interview, he explained that his team's first priority is delivering more complete and useful cross-selling information to employees on the front lines, to help them make better-informed product recommendations. He also shared his take on what the role of a chief data officer should be, how to get people in different departments to share information, and how big data could be applied to personnel decisions.
The following is an edited version of the interview.
You've compared the role of a chief data officer to that of an orchestra conductor, making sure the different sections play in harmony. What's an example?
A. CHARLES THOMAS: Each line of business and functional area, like human resources, finance or operations, cares deeply about its data. Sometimes they don't look to the left or right as much as they could to see how data that's critical for another line of business could be very important to them.
If you're running a credit card group, for example, why should you care about what's going on in auto loans and deposits? Because there's lots of data that tell us how our customers are engaging with us and many times engaging with other companies that we could take advantage of to help the customer with underwriting on products that might provide better pricing.
The goal is to drive value for the customer that also leads to them rewarding us with more business.
How do you get employees to care about the bigger picture?
The incentive piece is tough. It's my job to show how there's a mutual benefit. Take online bill pay: it tells you a lot about customer relationships. Customers use online bill pay to pay other banks for products they don't have with us. If a customer has a deposit account in online bill pay with us but not a mortgage, and they own a home, they're paying a mortgage with a different bank, so that's why you should care about online bill pay if you have the deposit account. If someone's paying a different credit card with their online bill pay, that's why you should care about their data.
I can show how we miss opportunities every month this number of customers uses a different bank to pay our credit card. Wouldn't you want to know that? Do you think you could win some of that business? They already like us, they transact with a credit card. Perhaps we should give them a nice offer that will give them an incentive to take action with our deposit account.
Bank customers might be identified by name and address in one product database and an eight-digit number somewhere else. Do you feel you have to create one common customer ID?
Yes. We have a program. There are multiple definitions of the customer that's a fact. We're building toward a similar view. The first thing is to at least capture and house all these definitions, and none of them are wrong. Each was designed for one thing, and now we're looking horizontally and saying: how do we match these up and reconcile them so we can have a relevant conversation with the right person in the household, or with the right household or the right accounts that might be owned by a person who lives in the household? Why mail multiple mailings for a product offering to the same household because we didn't have the appropriate definition of customer or customer ID there? The regulators are requiring that we have a common definition and so that's one motivation. But doing the right thing is another great motivation.
Will you need one centralized database for all or most of your products?
About 10 years ago, companies and banks were establishing huge data warehouses. Today the pace of new data being created, such as mobile and social media data or recorded data from complaints or quality control, as you add these new streams, it's hard to generate that uber data warehouse. We have what we call core customer elements including name, address, appropriate IDs and basic product ownership. But rather than try to centralize everything, we're trying to create pipes into the various other data environments that pick up the additional data. So it's more important to know where the data are when you need them as opposed to going into a laboratory and for three years building out this enormous data warehouse that costs a lot of money and then figuring out what to do with it. So if you want to optimize the customer experience across channels, you don't need all the data, you just need specific pieces of it.
Wells Fargo is famous for its Eight is Great cross-sales program, in which every customer household ought to have at least eight Wells Fargo products. It seems like what you're describing could help make that a more targeted cross-selling pitch: instead of asking everybody if they want a home equity loan, they could see you already have one and maybe leave you alone or offer another product that's more relevant.
We have this intense focus on the customer and helping the customer. We believe we have the best products. So cross-sell is something we pride ourselves on. In and of itself, there's good, better and best. People think, if you bring together all this data, does that mean you're going to try to sell me more stuff? My answer is, we should be trying to sell less stuff. In other words, relevance and timeliness are really critical. Say you pick up your iPhone and you want to check the balance on your account to see if you can afford lunch. Is that the right time for us to bombard you with a whole bunch of product offerings that you might buy, but it's not the right time to talk about it? We want to make sure we're relevant and timely.
A lot of bank employees don't really have a real-time view of customer activity. Is this real-time view and recommendation engine doable today?
Going forward, we're going to have to decide which [transactions] need to be real time. Trying to strip out everything you've ever done and replace it with real-time capability is extremely expensive and you wouldn't necessarily reap the benefit. Knowing that someone was at an ATM and then they went home and looked at something on their mobile device or went into a store what strategies are you trying to enable? You might not need to make everything available in real time, only certain data elements. Then you wouldn't have to strip out the entire engine, maybe just the belts and hoses.
It's a wonderful opportunity: think about the delighting factor you could have for a customer if they're looking at their mobile app, they wanted to apply for a credit card, and it crashes. They decide to go to their computer, they open it again and the first thing they see is a message saying, do you want to complete what you started? What a delighter, as opposed to having to repeat the process all over again.
That's a top concern that will keep cropping up in mobile banking, where people feel like they don't know whether a transaction went through.
We have the capability to track where a mobile app fails and how many people were affected. That's great from an operational, diagnostic perspective to see what needs to be fixed in aggregate. However, part of the CDO's job is to say, "How can we take that same information and find out who specifically had the problem when the glitch occurred?"
Imagine that now, from an issue-resolution perspective. The technology is there such that if they call, they can be routed to somebody who knows where in the process it failed and what to do about it, as opposed to, "Hold on, let me go check with the right department, do some fact finding and come back to you maybe by the end of the day." We could ensure that if that customer goes into a store, maybe to deposit money, a rep can say, "We see that you were trying to apply for a credit card, would you like to finish that?" Right now we know these things, but do we pull them together in real time so we can take advantage of those opportunities?
Are there any other areas where you have high hopes for doing something different using better data sharing, quality, analysis and other elements of big data?
The great beyond for us is employee data, to the extent we can legally access it with all the appropriate privacy considerations. Do we know the attributes of someone who's likely to be a successful team member at Wells Fargo? We capture lots of data. We can leverage those data to optimize recruitment and hiring. Maybe you find out someone from a totally different industry winds up doing really well at Wells Fargo. Are there specific college majors that might not seem related to banking, but that [correlate] to doing really well [in this industry]?