CHAPTER 16
Big Data and the Zombie Apocalypse
You are probably wondering what big data has to do with a zombie apocalypse. While I was thinking about the rules of an apocalyptic world, I realized that I follow the same thought process I use to think about what the rules are going to be for the new digital world, particularly in banking. I have said before I am not going to do a book where I try to scare you into doing digital; in fact, what I am trying to do is help to understand how to go digital, and part of it is going to be spending some time going over what data analytics really is.
I am going to recap: If you don't remember why we are going to have an apocalyptic world, let me refer you back to Chapter 13, which discusses the TIONS—interchange compression, cannibalization, digitization, mobilization, and disintermediation. We also talk a little bit about payments and how interchange is not disappearing and is not going away, but it is suffering from price compression, and we're going to see more price compression as time goes on. Alternative payments are going to be cheaper, and that means that as things move to ACH, and we see different models for payments, that's going to cut into our revenue. Lending is going to change. FICO won't be enough. We're going to see higher loan fraud from auto loan fraud rings—more bankruptcy changes, right? Student loans are in the crosshairs with that, as well, as we're continuing to see the CFPB hand out damage awards. Now, the CFPB may not exist in its current state, given the administration that's in power right now; however, over time I do believe that we'll continue to see more scrutiny in the banking space.
In this new, apocalyptic world, how do we survive? I started thinking about it in terms of how characters in zombie movies and TV shows survive. First, they're often transient. This is similar to how we're going to have to learn how to be more transient—more mobile—particularly with our data and our services. This is where the cloud comes in.
In the zombie apocalyptic world, you also have to make a little go a long way. I like to think of what a world of scarcity means in banking terms. In a world where scarcity becomes the new law, banks are not going to be able to count on catching an elephant, one big client that will provide a windfall. Banks that do manage to catch an elephant will have to figure out how to store it for the future so it will last. More importantly, you are going to have to figure out how to survive on rabbits, and it's going to take a lot more of them to sustain the bank.
Another thing to keep in mind is that in the world of the apocalypse, the law is not a primary factor in society. The old rules just don't apply anymore, particularly the rules about being nice to each other and playing fair. Banks will have to fight for their turf, and that's something that we have great difficulty with in this particular industry. We're going to have to learn to really make cases for ourselves in the light of regulation to make sure that we don't lose ground to fintechs and other folks.
Apocalyptic Risk
We need to identify risk quickly. It's important to note the difference between identifying risk and avoiding risk at all costs. It's OK to have risk. It will never be possible to avoid risk; that's an impossibility. However, if financial institutions can identify risk quickly and react, they will have the ability to survive. As discussed later in this chapter, data are critical for identifying risk and deciding how to respond.
My father was a military man. He liked to say high speed, low drag. That means he liked to move fast and jettison anything that held him back. In this world, opportunity is going to move fast, and what was here today might not be here tomorrow. For example, I know many financial institutions that have a team that's set up so that when a person calls and says, “I would like a 10-day payoff on my auto loan,” they'll call that person back. Sometimes it takes them an hour, though. In today's world, an hour is too long. By then, your customer has probably completed the loan, been put through the process, and they are driving off in their car. Meanwhile, your loan has been paid off, and your customer has a loan with the car company or someone else. We have to learn to react faster, and we have to learn to get rid of the things that are dragging us down.
In the process, we also have to learn to travel light, which is not a common practice in our industry. We live in a world where, if you've got a channel, we don't deprecate channels. We don't get rid of things. Drive-throughs, teller lines, ATMs, voice-response units, home banking—we still have all of it, and have no plans to phase anything out. We are now adding mobile, and piling on. The problem is that more channels require more resources. Without focus and planning, this will be unsustainable—and data will show us the way.
Staffing in an Apocalypse
If you think about the future in apocalyptic terms, you also need to consider the people who work in your organization. There are likely a lot of people who were helpful in the past but are not cut out for the future. To pick up our zombie analogy, what skills does a florist have that can be useful in the new and frightening world? Flower arranging won't matter at all, but perhaps she will know about edible plants that you can collect in the wild. Perhaps, her skills won't have to do with her previous career at all. Maybe the florist is really most valuable because she knows how to drive a stick shift or do tae kwon do. I consider moms to be pretty valuable, versatile employees. They are a little scary and they are also very resourceful, so I'm always looking for mothers to join my team.
Around my house, we have a saying: we're always talking about whether or not you want to be on our team. When the zombie apocalypse comes, we try to decide who we want our team, and who are the people we want around that we would be able to survive with. What I would ask you is to look around and decide, who's on my team? You might be surprised at the different people that are around in your organization that can be helpful in this new world.
Just to give you some examples of new titles that you might see, a chief analytic officer. We discussed this in Chapter 10. This is someone who's going to oversee this group. Statistical and optimization engineers—people who are optimizing algorithms to get the most out of them and doing statistical analysis to make sure that they're correct. Data scientists—There's a stat that says that there are only 13,000 qualified data scientists in the entire United States. How are we going to be able to bring those folks in and get them to work with us? Behaviorists—I read a great article about how folks in the digital world now are looking for philosophers and people with philosophy degrees, because understanding people's behavior is far more about understanding their philosophies and understanding their habits than it is about technical expertise. User experience specialists—people who can translate oddball data into changes to your digital front end that create opportunity for you. And then you are going to need a whole project management group filled with scrum masters and project managers and QA and all of these things, and it just seems enormous. What will you do? Well, here's the good news, you don't have to have all these things at once, but you do need to understand what these people do. You do need to understand what these roles are so you can grow into it.
Having these positions filled with the appropriate people will keep financial institutions in a good position for the future. In terms of resources—human and other—organizations will need to be well stocked. In the financial world, being well stocked requires people who have data and know how to use it. And, believe it or not, financial institutions are very well stocked with data. It's the people who already understand it that are very valuable. These are the people who know how to use analytics and machine learning to increase their marketing. That's going to be a huge opportunity. It's great that we have data and know what to do with it; it's great that we're willing to share it, but we're going to have to have some insights. This goes back to philosophy, poli-sci (political science)—those types of degrees where people can look at a string of data and make an assumption or an intelligent guess based on the human mind.
People who can collaborate will also be vital to organizational success. In the banking world, we're going to have to collaborate some, whether it be with other fintechs, which I think is probably likely, or whether it be with other like-minded financial institutions, we're going to have to be able to collaborate.
It might seem contradictory, but these collaborators will also need to know how to fight. In this world, if you don't know how to fight and if you haven't taken a punch recently, you're not going to be useful. After all, we cannot run away from risk, nor can we plan around risk. What we can do is plan for risk and how to respond to it. We have to be willing to fight for our ideas. We have to be willing to fight for our consumers and customers, and more importantly, we have to be willing to fight for our ideals. If we stick to those, we have an opportunity to win.
We also have to be willing to break the status quo. If you look at the main technology players, you look for example at AirBnB, arguably one of the largest hotel chains on the planet without actually ever building a room anywhere, you look at Uber, you look at Lyft, the sharing economy is a great example of breaking the status quo. What are the banking status quos, the clichés that our industry must break to bring people to us?
How will analytics help? If we really look at it, analytics has the power to bring us new engagements, to help us max out profitability, create cost-effective services, find new sources of income, help us to deepen our relationships, create digital relationships and, more importantly, help us to understand the habits, the trends, and the needs of our consumers.
Where will we find people who are a good fit for what we do? If you're a commercial bank, where will you find entrepreneurial people who want to use your services? Analytics can help with that. Analytics can help you by finding opportunities in streams of data that you did not know were there. Maxing out profitability—we talked about this before. We can't just drag an elephant back from the jungle, take one bite, and then throw it away. We're going to have to figure out how to use all of the elephant, and what that means in bank-speak is, we've got to do our best to engage our members and get them using all of our products and services.
Creating Value
Now I'm going to dispel a myth right now. I'm going to call it the myth of the primary financial institution. I've heard the PFI so many times that it makes me want to cry. That is not a reality. No one is going to bank with just one place. I do not believe that that's the case anymore, and if you want to be nostalgic and look back, you can certainly can find a time when it was like that, but in the new world, people are to have more than one financial institution to back them up. They're going to have a credit card from one place, an auto loan somewhere else, and a mortgage at a third. What you can do is look to be three out of maybe four other viable products in their wallet. You could be their auto loan, you could be their checking account, you could be their CD or their investment services, but you might not be their mortgage, and that's okay. What you can't be is just their checking account, particularly at the rates they have today. We have to find a way to drive up that profitability. We have to find a way to make the right offer at the right time to the right person for the right place for the right amount. That's what analytics will bring to you.
We also have to create cost-effective services. How do we determine what we're going to provide consumers? Well, we can no longer just put our finger in the air and let the wind blow by and try to decide the direction that customers are going. We are going to be using algorithms. We are going to be using data to look for data-driven results. To do that, we're going to use our history, we are going to use other people's history, and this is where collaboration can come into play. You might want to share your data with others to pull together to make some determinations of the future.
Can you provide valuable insights to your customers? If you can do this, you can create new sources of income. Would it be useful to them to know if they go to Chili's one more time, they are not going to be able to pay their mortgage? Sure, and you can do this, and that's some of the opportunity that analytics is going to bring, not to mention the ability to dig through streams of data and learning from what that data says, what people need. This leads to deeper relationships with customers. Social media is another critical source of data that can support deeper relationships. For example, do you know your customers' Twitter and Facebook profiles? What do they reveal? Have customers ever mentioned or contacted your organization online? Are these folks influencers in your world? Are they evangelizing your organization, or are they detractors? Data analytics will help determine what place data holds in our world.
Digital relationships are vital to an organization's future success. I know that sounds weird to say digital and relationship together, but the reality is you can have digital relationships by creating a meaningful connection with your consumer. That can happen over a mobile device or over a website. You don't have to see customers once a week in the branch like in the old days; you can see them online instead. You can relate your brand to them, and you can understand their needs, their wants, who they are, and who they are digitally. You can understand what services they use. Is there a difference between a person who is enrolled in Netflix versus a person who uses Hulu? Is there a difference between a person who uses Apple products versus someone who uses Microsoft products? You bet there is, but it is up to you to discover the context and what it means to you and your organization.
Digital Insight and Intuition
When we talk about insights, we usually mean intuition that you might have about a piece of data or something that's going.
How do people look at data analytics? Who will benefit, and who will be threatened? There's a great Pew research chart that states that global banks are going to benefit the most from analytics and data.1 Information is really valuable when you have a lot of it, and this is going to take me back to the collaborative portion of this. Our medium-sized or small-sized banks may not have enough data in order to make the intuitions or the insights that they need. They will have to collaborate with others to make sure that they have enough variety in order to make insights into what they're looking for. We'll talk about volume, velocity, and variety here shortly, and what that means.
When we look down at community banks, credit unions, regional banks, they have the most to lose from analytics, and I like to think of it like this, when FIs who are armed with analytics show up and compete against local FIs, they're going to know more about their customers and more about their business than most other organizations. This will result in a cowboys and Indians situation. When the cowboys showed up with guns and sadly the Indians had bows and arrows, and we all know how that went. Adapting to the coming data-driven retail environment will be an important skill that FIs must master if they are to survive in a fintech-dominated banking environment.
Data Is Valuable
Your first action item is this: stock up. Just like you would do in the zombie apocalypse, find a backpack and start stocking up on what you can find. Take an assessment of what you have. Did you know that as financial institutions, we're sitting on some of the most valuable and sought-after data in the world? We can tell you what home someone has. We can tell you what credit cards they like to use. We can tell you what they buy. We can tell you when they buy it. We can tell you if there is a trend to how they buy it. We can use our mobile apps to tell you where they bought it. We can tell you by what kind of phone they use. We can give you their credit reports. We can tell you what bills they pay on a regular basis with bill pay, and in some cases, for those of us that have a budgeting software or a personal financial management platform, we can even tell them what other financial institutions they have. This information is so incredibly valuable yet so unorganized from financial institutions. So, step one is to have someone take stock of your data.
Not only that, but are you deleting your data? Are you deleting the logs from your website on a regular basis? Are you deleting your mobile statistics on a regular basis? You might just be throwing away money. You need to stop deleting now and plan on hoarding. Hoarding is what we do when things get scarce, and in the new world, things are going to be scarce. So, sit down and decide with your team what data you have, and then find out how everything is connected. What's the process if you want a 360 view of one of your customers? Think about what's required if you want to say, “Tell me everyone who bought something from Home Depot in the last three months.” Or, “Tell me everyone who lives in the vicinity of this store.” You have to be able to do that. The challenge is, can you relate the two pieces of data? Can you say, “Show me everyone who has a home loan with us and who has a Netflix account on our credit card?” That's when things get fun. That's when the trends pop out, and that's when the opportunities happen. So, stock up. Those who are well stocked are going to do very well, and you need to know how well your team knows your own data.
Let's talk about some of the things that are a drag. These are the things I hear when I talk to smaller financial institutions that have attempted to do something with data analytics. The first one is, we have too much data and no one knows where everything is. Data analytics is done department by department. I've watched organizations putting together their reporting for the regulatory environment and it gets passed around from department to department, where a person in mortgages adds their portion of the report and someone in the call center adds their data and so forth and so on, until finally there's a report put together, and the reality is that no one person knows where everything is, and in reality and totality, we are not even sure that anyone knows where everything is. This is an important issue, and it's a big part of the drag, right? You sit down to go do something, only find out that you are not organized and you don't have the data you need, and nothing is more demoralizing than having a good idea and not having data to back it up.
The second thing that I see a lot that drags organizations down in the data analytics space is, they feel like they don't have enough processing power. You want to say to them, “Hey, why aren't you analyzing your bill pay data?” Or, “Why aren't you analyzing these various aspects of the data?” And they say, “Well, we'd like to, but last time we ran that job it took us three months or two months or it never finished or it took a whole day.” In a world of quantum computing, or a world of distributed processing, we should not have that issue.
What can we do to deal with that? One, we can collaborate. Two, that's where the cloud could come in—you could buy a really large machine and only use it for a few minutes or however long it takes to run the data and then shut it down. There's a lot of ways to get around processing power in today's world. The key is that we have to want to do it and we have to look for those opportunities rather than just sort of admit defeat when we run into these problems. Another comment I hear is, “You know, we had a guy and he was great, but he left and now we can't find someone else.” The reality is that there is a true shortage of people who have the skills to do it; however, if look at your situation I think you will find that you don't need a full-time data scientist. Outsourcing these positions can pay great dividends. You can easily make a business case around having several data scientists on consulting teams who are able to take your loan projections or perhaps your delinquency projections, bring those down and increase your loans and pay for these engagements in a short amount of time.
Finally, analytics is viewed as a single product by management and not a capability or strategic partnership. That's a huge challenge, and I mention this in the digital space as well when I said digital is not a product, it's a discipline. Well, in a similar mindset analytics, it's not a product, it's a discipline. If you think about it, if someone was really good at QuickBooks, we wouldn't call them our CFO. We can't just buy QuickBooks and check off the accounting function of our organization. We have a CFO, we have trained accountants and CPAs. We rely on these trained professionals in our banking organizations to deal with the complexities of bank account rules. Why would we think analytics is any different? Analytics is a discipline, and it deserves a seat at the table where decisions are made. It is important to understand that in the future, analytics will be involved in every aspect of the financial business. If you open a new branch, you are going to want to understand the demographics of the area that the branch will serve, you are going to want to understand the foot traffic and what people might abandon a branch that they were using to go to the new branch. This is a job for analytics, and they should be involved from the start of the project. If you are designing a new financial mobile application, you're going to want to understand why the people would use it, who to market the application to, and what services the users will expect. Process engineering is also a good place to engage the analytics team. They can help determine what data to collect, how often to collect it, and how long to retain it. I think you can see the pattern here; analytics should be involved in every decision, every new project, and every process.
More importantly, it could be that your analytics group is a strategic partnership headed by someone in your organization who either understands analytics or has expressed an interest in understanding how to manage it. Ideally this person should be an executive with staff that has been involved with the reporting and with a seat of the table where decisions are made. This person might also look for an intern or a junior data analyst to help him interact with the consultants. Finding a consulting partner is a great way to learn what your organization needs in terms of analytics. It may be sometime before the organization is ready to create a full-time analytics group.
Data Is a Discipline
If we are going to survive in this world, we have to understand that analytics is not a product, it's a discipline. Our ability as an organization to execute on analytics is paramount to our survival in this new world.
Let's look at some examples. Let's look at Google. Google processes 3.5 billion requests a day. It is currently building data centers all around the world. It is anticipating an exabyte a day. That's an insane amount of data. Eric Berlow, TED speaker at our latest Analytics and Financial Innovation (AXFI) Conference, was asked what he thinks is behind the analytics boom. Most people think increased processing power and more sophisticated algorithms are behind it. Berlow thinks that it's the volume and the vast amount of data that we have access to now that companies like Google or Amazon or Facebook or Microsoft or Apple are collecting on a daily basis about us, and all of that data means that you can mine it and you can look through and learn things about the data—you can find opportunities within the data.
Facebook collects 500 terabytes of data daily, including 2.5 billion pieces of content, 2.7 billion likes, and 300 million photos. And it has admitted to storing 100 petabytes of photos and video—as of 2012. Amazon draws data from 152 million customers, and if you don't think that it is running systems and things through its AI, you are crazy. These companies are learning from reviews, feedback, what you looked at, and what you clicked on.
I have a favorite story about Amazon. I was riding into work one day and, as a huge Van Halen fan, I decided I wanted to listen to an interview with David Lee Roth by Howard Stern. When I tuned in, there was an on-air testimonial for a product called a Squatty Potty. My curiosity got the best of me, and when I got to work I googled “Squatty Potty” and I clicked on the Amazon link that came up. I learned that Squatty Potty is a tool to allow you to go to the bathroom easier. It's like a footrest you put in front of the toilet bowl. The Amazon page had a drawing of a person sitting on a toilet and using the Squatty Potty. You may have noticed that the things you view on Amazon have a way of popping up in ads on other websites. That's called a super cookie. Well, that image of a person on the toilet followed me around for days, and I could not get it to stop. Now, that said, it's an effective way of doing sales, but in my case, during presentations to hundreds of people, I would pull up some site to illustrate something and there would be that man on the toilet, which was more than a little embarrassing.
This is a great point, and I can already hear you saying this: John, we've been thinking about this, and our board is just not comfortable with using data like this. There's a fear they were invading people's privacy, there's a fear that we'll become like big brother. Recently there's been a lot of stories in the news regarding privacy issues. Edward Snowden, arguably the biggest story about privacy in years, exposed the government's ability to digitally spy on the citizens of the United States in the name of protecting the homeland. Right or wrong, he exposed the government's incredible reach in to our daily lives by using our cell phones and online habits to track people of interest.
Another recent case of privacy involved the terrorist attack in San Bernardino in February of 2016. Investigators collected one of the terrorist's cell phones, which happened to be an Apple iPhone. The iPhone was protected by a passcode that the FBI couldn't break without risking erasing the phone and losing its contents. The FBI then contacted Apple in an effort to gain access to the phone by requesting that apple create a special back door. Apple declined due to a policy to never undermine its own security measures. Most financial institutions regularly work with the authorities to deliver data when requested via subpoena; however, in the future I expect financial institutions to institute similar encryption techniques that may put FIs in the same position as Apple was with the San Bernardino case. Google and Facebook are in lawsuits worldwide, and, of course, one of the most documented and famous data overreaches occurred at the retailer Target.
Target is one of the biggest users of analytics. I recently read Charles Duhigg's book The Power of Habit, where he describes the now infamous Target incident in detail and gives a behind-the-scenes look at Target's extensive analytics department. Target discovered that by analyzing the buying habits of its customers, it could make predictions about their future needs. For example, the analyst at Target found a link between customers who purchase large quantities of lotions, or certain kinds of vitamins, and pregnancy. Target then compared these results to its baby registry to determine the accuracy of the model. Using the baby registry as its guide, they identified common buying trends that allowed them to accurately predict if someone was pregnant and even how far along in their pregnancy the customer was. It turns out that pregnant women are highly sought after by retail stores because statistically, wherever the mother buys her baby supplies like diapers and bottles is where she's going to buy other things like groceries, because it's convenient. New moms value time (and possibly sleep) above all, and because of this, they tend to prefer one-stop shopping.
Target sent an advertising flyer filled with coupons for baby clothes, cribs, and other baby needs to a young lady who had bought some of these items and fit their model. Unfortunately, this young lady was a teenager who was still in high school at the time, and her father took offense to the overt target marketing and complained to the store that Target was encouraging his teenage daughter to get pregnant—only to find out later that Target's model was right and his daughter's baby was due in August. Target discovered that sending ads that are filled with a particular type of item clued people in that their shopping habits were being monitored, and it also discovered (thanks to the very upset father) that people don't like it when they feel like they are being manipulated into buying things. Target changed its advertising so that alongside the targeted items (such as the baby supplies), the ads included other unrelated items such as TVs or lawnmowers. This approach allowed it to target market customers without alerting them to the fact that their purchase data had been analyzed to create the flyer.
Target has a lot of data about its customers but that pales in comparison to the quantity and quality of data that a financial organization has about its customers. I have talked to many executives at financial institutions who are very concerned about mining their data and using it to advertise to their customers, for fear of creating a situation similar to the Target incident.
Here is my theory on the privacy issue for financial institutions: Do anything you want to do with data in the banking space as long as it works in favor of the consumer and you're transparent about what you doing. If you're honest, up front, and you are doing something to help somebody save money, make money, or make good financial decisions, then do whatever you want. Financial institutions must resist the urge to use the vast amount of data that they have to manipulate their customers. As organizations become more proficient with their data, the lines between being creepy and doing the right thing will get increasingly difficult to discern.
In Europe, they are particularly sensitive to privacy issues and have passed laws that are designed to protect people's online and digital privacy. The new law is called General Data Protection Regulations (GDPR) and goes into effect May 25, 2018. The law has two main pillars, Right to be Forgotten and Informed Consent.
Let's start by dissecting the part of the law that states that everyone has the right to be forgotten. In Europe, once this is enacted, an individual will be able to contact Google and Facebook and request that they delete your search history and any other profile data that the company has stored about you. The law also states that multinational companies (such as Google and Facebook) will be treated as single entities. As of this year, many of the largest US multinational organizations are spending hundreds of millions of dollars to comply with the regulation. The regulation states that organizations that are not in compliance could incur fines of up to 20 million euros or 4 percent of global revenue (whichever is greater).
As of this writing, a major breach has happened at Equifax, a provider of credit services with millions of accounts and access to very private data. Had this happened after the GDPR deadline, Equifax would have had to pay 4 percent of its global revenue (more than $3 billion), or $120 million in fines. Remember that if you house any data from a citizen of a EU country, then GDPR theoretically would apply to you.
Informed consent means that we cannot put 50 pages of a EULA, which stands for End User License Agreement, and have a button or checkbox that states “I agree I have read and understand the EULA.” GDPR states that an organization must be able to clearly communicate in simple terms to customers what they are agreeing to. Many people discover their data on the internet and later realize that a EULA for a piece of software they installed included a provision that allowed that company to mine their personal data. A company called PC Pit Stop decided to test to see if anyone was actually reading its EULAs and included a provision that promised financial consideration for anyone who emailed an address in the disclosure. It took four months before someone read that part of the EULA and sent the email (that person got $1,000), proving that no one really reads the EULAs. Informed consent provides protections from users against organizations inundating users with large documents that they know no one has the time to read. It also clearly states that the EULA should not contain other provisions that don't specifically pertain to the software or data processing that the EULA is intended to cover, and if they do, they must be clearly spelled out to the end users in such a manner that they can clearly see that the request is out of scope for this particular EULA. This means that there cannot be any loopholes in the EULA that, say, allow the company to sell data about how you interact with its service to a third-party, especially if that provider doesn't have a clear connection to the product or service that EULA covers.
The good news is that financial institutions have plenty of data, we have access to the right people, we have good intentions, and for most of us, laws like GDPR don't affect us. We can use our data to drive results and we can use our data on behalf of those that we serve—as a matter of fact, most of them are expecting it. Speaking of expecting, let's dive into the different kinds of analytics.
Types of Analytics
Let's talk a little bit about what data analytics is and what the tools of the trade are so that when you're starting to access people, software, and tools, you have some idea of what the lingo means. The first thing to understand is the types of data analytics. There are three types of analytics that most people talk about. The first one is descriptive. Descriptive analytics tells us what happened and not what will happen. It's historical data. A great example would be a branch report. Please tell me all the people who came into the branch and used the teller line last Friday. Tell me all the people that used the drive through; tell me all the people that used the ATM. Because these are all things that happened in the past, they are considered descriptive analytics. Reports described something that happened. I find that most organizations are already proficient at this sort of analytics; many have been relying on historical reports their entire career and still use descriptive analytics for decision making. People who make decisions on descriptive analytics are usually looking for trends to decision on. Consider a branch manager reviewing stats from the previous year. She might take note of the fact that in the winter around the holidays the branches are far busier, and as a result would adjust her staffing accordingly. She hopes that the trends she is seeing hold. While this approach has been used for many years, it doesn't account for changes in the environment that might affect the amount of foot traffic that the branch will get. For instance maybe there will be construction in or around the branch that prevents people from getting to the branch and as a result drives down the amount of customers.
That leads us to predictive analytics. Predictive analytics is the art of taking historical data, adding other data, and intuitions, and coming up with future trends. While we can certainly forecast what might happen based on the past, when we add new information to the data set we increase our chances of finding new opportunities. One example of new data is doing comparisons with other channels such as the ATM or home banking to increase the validity of the results and find additional trends. This is where most organizations are trying to get to—they are trying to go from descriptive to predictive, especially in light of new regulation. A great example of predictive analytics can be found in the ABA's change to the Current Expected Credit Loss (CECL), which replaces the current “incurred loss” model with an “expected loss” model. The incurred loss is largely based off of historical data and trends and has its roots in descriptive analytics. This is in stark contrast to the new “expected loss,” which describes an event that happens in the future. The new rules require financial institutions to essentially “predict” risk in their portfolios between 2020 and 2021, depending on whether your organization is considered a public business entity (PBE).
The final category is prescriptive analytics. Prescriptive analytics is about scenario planning. It gives you actionable data, data that you can use to make decisions when choosing between two scenarios. Prescriptive analytics can also be used to mitigate potential future risks For example, if there is a concern of a bubble in mortgages or a future drop in the market, a prescriptive analytics model can be fed the different parameters of the market and it will provide many decision options to address that risk. All of this sounds great, right? It's exactly what a financial institution needs! So how do I make it happen? Well, the reason I put this one last is because it's very difficult to achieve without mastering descriptive and predictive analytics first.
Prescriptive analytics will be the catalyst for implementing machine learning or artificial intelligence. For instance, let's say we were trying to determine which customers are planning on buying a car in the next six months. How would we go about determining this? The current way is to combine historical data on customer behavior and compare it to a baseline of customers and their behavior in the six months before they made a decision to buy a car. With this model, you would get good feedback, but it would rely on human observations, which can sometimes be blind to trends that aren't obvious.
A prescriptive analytics model could also employ artificial intelligence to process the inputs. The human model would look at all the people who have bought a car recently, and then we would comb through their actions in the months before they bought the car and try to use human intelligence to determine what the common trends are. When utilizing artificial intelligence in your model instead of relying on human intuition to discover common trends, you feed it a massive list of people that did buy a car. You feed it as much other information as you have on hand that relates, for instance branch visits, website information, purchase history, and anything else you have lying around, and you let it find the trends. As it gives feedback, you may have to weed out false positives or train it, but eventually it will learn, and as you get more people that buy cars, you keep feeding it these opportunities, and you'll also get a baseline. Then you give it your target list of people that didn't buy cars. The artificial intelligence will then use the classifier it created from the baseline to identify the people who are likely to buy a car in the next six months based on matching behavior.
The more you feed the model, the more it will learn what the prerequisites are. It will begin to discern what the forward-thinking trends are, and most importantly, what the leading indicators are for people who are going to buy a car. It could be that they are expecting a baby (maybe partner with Target?) and they're going to need a minivan. It could also be that someone has spent a lot of time in the car section of your website, which would be obvious, but there might be unobvious things—things that only a computer could find, such as a high amount of payments to an auto repair shop in their purchase history. The only process that could possibly see these trends, because the human mind is not capable of comprehending or perceiving this data, will be artificial intelligence. These sorts of analyses are going to be game changers for financial services in the future.
Let's talk about the terms of the trade. Here are some simple terms. You are going to hear some weird stuff, and it's probably good that you at least know what these things are. The first one you hear a lot is called Hadoop. It's an open source large data set manipulator, and has become pretty popular for folks who are trying to crunch large amounts of data and look for trends. It has a cute little elephant for a logo, and if you've run into it you would know. The next one is just a letter of the alphabet, which is R. It's yet another open source tool; however, its intended purpose is to provide statistical analyses and data visualization—if, for example, you wanted to do a 3D plot map on a site to determine which of your members are more likely to go to a certain branch or adopt a certain product based on geography. Finally (and this is kind of an overarching term covering a lot of different products), is machine learning, also referred to as cognitive analysis. It's currently the most popular course at Stanford, and ultimately, it is another name for AI. As we get bigger and bigger and bigger data sets, machine learning is designed to discover trends and point out behaviors that will lead to opportunities that a human would never find. Each of these tools has specific purposes and builds on the three disciplines of analytics that we have been discussing in this chapter: descriptive, predictive, and prescriptive.
Now, let's talk about the data itself, and what are the tools that you need to have to be successful at data analytics. We like to think of them as the three Vs. The first one is volume. The more volume you have, the more likely your organization will come up with outcomes that will be viable. If you don't have a large data set, then it's possible that your outcomes will be tainted by a lack of scale. In other words, if you want to find out a trend about apples, you are going to need more than 10 apples to determine a trend. You are going to need to look at thousands, maybe hundreds of thousands, of apples before you can really make a factual statement about apples. That's volume.
The next one is variety, and that's the diversity of your data. If all you have is apples and you are trying to trend fruit, then you would be missing oranges and bananas and all the other types of fruit that could affect your trend. Having a variety of data is very powerful. The more data variety you have, the more likely you are to determine trends, to detect anomalies, as well as make accurate predictions. If you were looking at a group of customers and you only included data from one zip code in your data set, your outcomes could be wrong due to the regional behaviors of a particular group. Variety fully plays into that.
The final one is velocity. Velocity of data is fast becoming the most important aspect of analytics. Data over five hours old is useless. Case in point: Earlier we talked about someone who called her FI about a loan payoff. When the financial institution called back to say, “Hey, what are you doing? Are you buying a car? We want to help you!” the customer responded, “Oh, I already bought a new car.” Like life in the zombie apocalypse, things move fast, and because the response speed of the FI had not caught up with the speed of making a loan at a dealership the FI lost a customer and a loan. If you're working off a data set that was most recently updated last quarter, it's possible that the situations that you are including such as credit card balances, assets, or financial positions have changed. Not only that, it's even more likely that the environment that facilitated the outcomes you are using in your dataset have changed as well. For this reason, we will see a shift to real time data, or close to it, being used in future analytic models to avoid making offers or assumptions from stale data.
Again, to recap, that's volume, variety, and velocity.
Closing out, what are our opportunities here? We are going to have to become digital services slayers. We're going to need to create digital streams that allow FIs to take the digital services that are bringing us so much data and make use of it to improve our customers' financial health. Notice that I didn't say to manipulate people, that's not what we're talking about. We're going to need AI (see Chapter 4). We're going to have to have the ability to determine, as I mentioned before, the difference in behaviors between someone who owns an Apple iPhone and someone who owns an Android mobile phone. What are their wants, what are their needs, how do they make their living, and what do their life choices say about them and their character? After all, most of determining loan risk is about assessing character.
Finally, we're going to have to monetize our own data. I know this is again an anathema to our industry, but people are already doing it. If you think about Mint, if you think about Yodlee, if you think about all the products that log in and screen scrape through our platform and then pull our data out, others are already monetizing our consumer's data, and the customers are opting in and allowing these third parties to do it. Monetizing data may mean working with digital marketers to advertise cars in our home banking or mobile products, or market local retailers services that are trusted to the customers.
In summary, here are the key points to stay alive in a zombie apocalypse:
- Be mobile. Don't stay in any one place too long. Always be looking for the next safe harbor.
- Stock up. Make sure you are not deleting important data. Take stock of what you have. Start hoarding today!
- Choose your teammates carefully. Make sure that your culture will support a data-driven approach.
- Data analytics is a discipline. It is not a product.
- Act quickly and decisively using data to validate your approach. What was here a few minutes ago might be gone in the next five minutes. Windows are shrinking.
- Learn how to use the tools of the trade. If you can't shoot a gun, then just having one handy isn't going to protect you.
- Determine your capabilities. Know where your organization is operating based on the data analytics models: descriptive, predictive, and prescriptive.
- Monetize your data. If you don't, someone else will.
So that's how you survive the zombie apocalypse with data analytics.