Information Overload: Capitalizing on Big Data Overview
Information Overload: Capitalizing on Big Data Overview
Britt Technology Impact Series
August 17, 2013
The Britt Technology Impact Series is an offering of the Center for Digital Strategies at Dartmouth’s Tuck School of Business. It is made possible by a generous donation from Tuck and Dartmouth alumnus Glenn Britt, CEO and Chairman of Time Warner Cable. In giving the gift, Glenn stated: “The role of business people is to understand the possibilities created by new technologies, recognize unmet consumer or business needs they could fulfill, and determine if the new technology and the customer needs can be put together in a business model that makes sense.” The Center for Digital Strategies structures the Britt Series so it highlights relevant aspects of a set of technologies, examines business models and illustrates how consumer and corporate needs are being met.
The Britt Series focused on big data for the 2012–13 academic year because the amount of information in the world is growing at an unprecedented pace. This ever-widening flow of data is the byproduct of a digital, networked economy. New types of information and new combinations of data sets are yielding new insights. This “big data” is changing how technology can serve consumers and enterprises. The following summary highlights the unique perspectives offered by Britt Series speakers who are at the forefront of working with big data and, as such, in developing new ways to examine and interact with our world.
Thinking big about big data
The 2012–13 Britt Technology Impact Series examined the explosion of information emanating from our digital world. This phenomenon is often referred to as “big data.” This is the idea that the amount of raw information is stretching beyond our ability to manage it and make use of it. The technology that underpins modern society — from smartphones to social platforms to supply-chain networks — is driving an unparalleled increase in data. SAP AG explains it this way: From the start of recorded time until 2003, humans created 5 exabytes (or 5 billion gigabytes) of data. By 2011, humanity created that much data in two days. In 2013, it takes only 10 minutes to generate 5 exabytes. This rapid growth demonstrates new tools are needed to capture, analyze and use huge sets of data.
The executives who convened for the Britt Series revealed myriad ways in which big data is changing everything from how marketers reach consumers to how political organizations canvass voters to how automakers design cars.
The discussion around big data begins with a simple question: How big is big data? There is, of course, more than one answer. One constant is that the perimeter of the big data universe is in a perpetual state of expansion. A deeper understanding comes from knowing what produces big data. Consider that YouTube users upload 72 hours of video every minute; Wal-Mart handles more than 1 million customer transactions per hour; Facebook users click “like” more than 3 billion times per day. Machines talk to other machines. Each of these actions produces a digital footprint that is, at its simplest, data. Analysts predict data will grow by at least 40 percent annually in the coming years.
Executives gathered for the Britt Series offered their own definitions of big data:
- At the insurer Aetna, big data means handling 500 million to 600 million claims per year while maintaining an error rate of less than one half of one percent.
- For the consumer-data company Buxton, big data comes from the 7,500 data points it holds on the average U.S. consumer.
- At Ford, 74 sensors, 70 onboard computers and more than 130 motors in the automaker’s plug-in hybrids generate big data. Some vehicles produce 25 gigabytes of data per hour.
- At Rentrak, a provider of entertainment metrics, big data comes from tracking what people watch on TV in more than 100 million households and on 85,000 movie screens around the world. It works out to more than 7 billion transactions every day.
It’s clear big data is big. But big data is about more than scale. Researchers often define big data as having three Vs: volume, variety and velocity. Volume is the amount of information. Variety refers to the many types of data from videos to social media posts to GPS coordinates. This so-called unstructured data can be more unwieldy than rows on a spreadsheet. Velocity refers to how quickly information from text messages to stock prices is generated and begins to pile up.
Consumers are leading the growth in big data by using more technology. By 2016, there are expected to be 1 billion people using smartphones and tablets and more than 25 billion devices connected to the internet.
Patrick Pichette, SVP and CFO at Google Inc., told a Britt Series audience that enterprises and consumers are only at the beginning of understanding what big data will unleash. “We’re really in early innings.” He noted only about 2 billion people have a cell phone. “There are another 5 billion that will show up with different economics, but there are 5 billion of them and they want a better life and they want better tools and they want amazing things and the real question is somebody is going to give it to them. Who is it going to be?”
The big deal around analytics
The Britt Series speakers agreed that as remarkable as the amount of information is, the more amazing aspect of the big data story is what it can enable. “It’s not the capture of big data. It’s the execution of big data,” said Tom Buxton, chairman of his eponymous company. He and other Britt participants argued big data will not only reshuffle the competitive landscape for enterprises but also allow companies to offer personalized service for customers on a scale not possible in the past.
Ruben Sigala, SVP of enterprise analytics at Caesars Entertainment Corp., said the exponential growth in data can let companies assemble far more meaningful analyses. “There are a number of ways that if leveraged well you could substantially change the way that you run your business,” he said. “You can learn so much more quickly.”
Sigala and other speakers predicted enterprises that fail to pursue the high-definition picture that big data can render are likely to falter. “As we think about winners and losers in the future, the ability to integrate analytics in a meaningful way into the operations, I think, will be a telltale distinction amongst companies,” Sigala said.
John Ginder, manager of systems analytics and environmental sciences at Ford Motor Company, said more information can sharpen decision-making. “For us it’s around the analysis: Are we going to inform better decisions, quicker decisions?” he said. “We’d like to use big data to serve our customers better. That’s really the ultimate goal.”
Part of that improved service can come from generating deeper insights. “It’s about turning data into intelligence,” noted Chris Kelly, former chief privacy officer at Facebook Inc.
One area where big data can flex its muscle is in taking information, both new and historical, to make informed projections. So-called predictive analytics is aimed at helping enterprises, and even consumers, answer difficult questions based on mountains of information. Mike Gualtieri, a principal analyst at Forrester Research Inc., said while predictive analytics is not new, big data is making it possible to generate more accurate predictions. “Big data reinvigorates this because great predictive models depend on great data,” he stated. Such models might take the form of a decision-tree, for example, illustrating the consequences of a particular action. “This is why it’s not just a buzzword,” Gualtieri said. “People are finding the knowledge in that big data. So it’s not just about storing; it’s not just about its massive size. It’s about mining that data to create knowledge that you can use to outcompete.”
Most industries would benefit from faster analysis, Britt Series participants agreed. “You can’t just go in a room and come up with some hypothesis and some predictive model that you can just then deploy for 18 months or even a year. You have to continually retrain that model for all of the changes that are occurring in the marketplace,” Gualtieri said.
Winning big by getting better answers
Examples arose throughout the series of the types of souped-up analytic tools made possible with big data. Simply having the information is not sufficient. “Operationalization of how to use the data is much more important than the fact that it’s out there,” Kelly said.
Putting the data to work is critical. Real-time analysis can save companies money, noted David Chemerow D’73 T’75, COO and CFO at Rentrak Corp. When Warner Bros. Entertainment Inc. released an installment in the “Harry Potter” franchise, Rentrak data indicated the movie was such a hit that running further advertising was not necessary. “On Friday afternoon as it opened we called up Warner Brothers and said ‘Guys, stop your TV advertising.’ They said ‘Oh my God, what’s wrong? The movie’s a hit, isn’t it?’ And we said ‘It’s so much of a hit, you’re sold out for the next seven days. Kill your TV ad-spend and bring it back next weekend.’ They saved $20 million by doing this,” Chemerow recounted.
Ford’s Ginder explained how the company uses big data to help solve the type of logistical problems that batter any large company with a sprawling operation. Ford buys the molds, stamping dyes and other tooling its suppliers use to make Ford parts. This intellectual property is valued in the billions of dollars. One challenge centers on whether to source parts locally for a brisk-selling vehicle such as the Focus. “Do all the North American Focuses get parts from North America? Do all the European Focuses get parts from Europe?” Ginder asked. “Or, do you instead do some global sourcing?”
Ford set up an analytics tool to tease out the best answers. “We are using this to help make decisions of outsourcing for these global programs and given the immensity of the data input and the types of decisions that are involved in it, it’s a huge help,” Ginder explained. “It wouldn’t be possible to be done by a human being with a spreadsheet.”
Michael Angus T’87, group head of global payment strategy at MasterCard Advisors, the credit card company’s professional services group, said the company uses its enormous storehouse of data to save money. “Done properly you see impacts on fraud rates or impacts on risk default rates where you’re talking multiples,” he said. “You’re talking cutting the fraud rate by a factor of four or cutting the default rate by a factor of two or three with the right application of big data.”
At Caesars, the company can intervene in real-time to enhance or perhaps preserve a good relationship with a customer. “We can draw a pretty direct line between a customer service score and how he or she will transact with us on a given trip as well as over a lifetime,” Sigala said. He described how indicators such as a person’s social media activity can help the company determine the customer’s satisfaction. “It gives us an opportunity to interject within trip and outside of trip to preserve or recover from any problems that they may be having.”
The company would like to tap into more data on the customer experience. It would be helpful to know, for example, how long a guest had to wait in line at a company property or whether a person bypassed a blackjack table because a dealer wasn’t present. “There are technologies now that start to enable that and, for us, the more properties you visit within our footprint, that’s meaningful,” Sigala said. He expects it will become “increasingly important” to monitor lines and the service experience. “I think we’ve got a pretty clear roadmap as to how we’re going to get after it, but it still is early days.”
Sudev Balakrishnan T’07, director of e-commerce and product management at high-end fashion retailer Bluefly.com, said the company is able to use data to make faster decisions that can influence shoppers. “This kind of analysis is going to be required in the industry,” he said. Using traditional methods of data crunching, the company is able to identify “when a customer is in flux about a cycle or two cycles in advance of that behavior becoming permanent. So, that gives you basically one shot at persuading that customer one way or the other. Now, with advances in sort of the granularity of the data that we have and leveraging the networks that we have, we think we can extend that by three or four more cycles,” Balakrishnan said. He underscored that such an improvement could lead to a tremendous change in revenue and therefore for the company’s profitability. “That’s why I think there is a legitimate urgency around getting this right. Because ultimately winners and losers may very well be judged on how they function in this space.”
A range of industries have deployed big data analytics. Oil companies untether huge amounts of data to help detect deposits of natural resources. Other companies are increasing the level of detail they have in supply-chain data. Some manufacturers have used big data tools to monitor moving parts in a factory for vibrations. Such gyrations can signal a part is nearing the end of its lifecycle and could need to be replaced. Swapping out parts too early is costly but so is waiting until a breakdown because such events can idle a plant.
Big data tools are muscling into areas outside the business world. With the backdrop of the 2012 U.S. presidential election, the series explored the role of big data in politics. “What I have seen big data be able to do is tell campaigns when a poll might not be entirely illustrative of the situation that they’re in,” said Nate Murphy, election center manager at NationBuilder. The company offers a software platform for organizing communities and makes available public voting records. Murphy noted the 2008 Hillary Clinton campaign achieved what it believed was a sufficient number of voters to win the Iowa Caucus. He noted the Barack Obama campaign turned to big data technology to reveal a more detailed analysis of the field. “The Obama campaign did a great job with big data to track the independents or Republicans that they were bringing to the polls for the first time.” What followed was a record turnout. “The only campaign that predicted that was the Obama campaign because they were doing great modeling and metrics about who they were actually building relationships with and what sort of contacts they were bringing in. So, I think the real advantage to big data is it allows you to all of a sudden have these new types of metrics and more accurate tracking of what your actual field operation is doing.”
Engaging in some big experiments
Big data lets companies experiment in ways not possible only a few years ago. Simple examples belie the complexity of the data crunching some enterprises are doing. The grocery chain Safeway has been testing ways to better serve customers. An employee alerted to the presence of a repeat customer entering the store might offer her a cup of coffee based on her purchase history. The employee also might hand her a coupon for flowers because models predict she might be likely to make this type of first-time purchase.
Philip DeGisi T’09 is director of marketing at pet-products retailer Wag.com, a division of Amazon.com’s Quidsi Inc. He underscored the importance of tinkering with promotions to boost conversions in a business selling consumable goods. “Where it’s pretty repeat-oriented, a fraction — a couple basis-points change in our repeat rate can really move the needle in the lifetime value,” he said. “How efficiently we convert the customer becomes that much better. So, there’s huge upside across all the elements of the business.”
Angus said MasterCard Advisors is working to understand how massive amounts of purchase data can be combined with other information. On its own, much of the information is of little value. He noted risk bureaus might track different risk behaviors that can be married with purchase data. “We’re struggling with what data will add value when combined with our data,” he said. “We learned early on that even the huge amount of data we have by ourselves can be a lot more valuable when we marry it up with other stuff.”
Designing appropriate experiments involving consumers can be difficult. “A lot of where we think about testing-control that can get a little sticky is recognizing the inherent volatility of our customer behavior and then designing experiments that contemplate that volatility and in an accurate way. And that’s why this partnership between the analytics and the operations is absolutely essential,” noted Caesars’ Sigala.
Balakrishnan believes it is important to define the objective and parameters of an experiment. “Data is only, I think, as good as the hypothesis that you start with and you try to see that data supports it.
Getting personal in a big way
For consumers, big data might begin to show itself through the level of personalized interaction they are able to have with technology. Gadgets can act more like human assistants by anticipating needs rather than reacting to cumbersome programming.
Gualtieri offered a personal anecdote concerning the purchase of a new smartphone. As with his old phone, he began to carry it with him on the bus he took to work. After two weeks, without prompting, the new phone displayed his bus schedule. The device learned Gualtieri’s schedule and offered up information to assist him. “The company’s relationship with a customer is more that of a butler,” Gualtieri said. “Your relationship is standing by the side, kind of knowing what that customer’s going to need and giving it to them at the right time and the right place.” Big data and mobility makes this type of relationship possible. “You have more access to information on that smartphone than the president of the United States had 15 years ago.”
For enterprises, establishing a butler-style relationship with consumers requires tailoring services to fit a person’s interests and needs. Buxton explained this process can involve “micro-targeting” — understanding what motivates an individual or what one’s propensity to do something might be.
If the “butler” fails in a task, the results can be damaging because the consumer has grown to rely on the enhanced level of utility provided by a good or service. “People get used to good results very fast,” noted Bluefly’s Balakrishnan. “If there is any loss of the … quality of service they complain on their iPhone right there, immediately.”
The push to create more personalized experiences is in part because consumers have “choice and voice.” There are more products and services available to consumers and because of social media and other outlets, consumers have a louder microphone from which to complain about poor service or to commend good service. “This choice and voice is putting more pressure on companies to create better customer experiences,” Gualtieri said. “To provide better customer experiences, they have to make it personal.” This goes beyond the former iteration of personalization in which a consumer might be able to control insignificant factors such as the font size or color in an app. “When we talk about making it personal, we’re really talking about a personalized experience,” Gualtieri explained. He pointed to technology such as the Fitbit Flex wristband that collects information about how many steps a user takes, how many calories he burns and how well he sleeps. “That data can be used to make personal decisions and make you offers,” Gualtieri noted.
The march toward ever more personalized interactions via technology is also changing inner workings of enterprises. Serial entrepreneur Andy Palmer T’94, founder of Koa Lab, a start-up club, remarked the consumer internet is setting expectations for what analytics can and should be. “Many people have this experience where they walk into their … office and go to do some work and they’re hit in the face with these antiquated and completely inadequate information systems relative to what they get at home on the web every single day,” he said.
Big questions around health care
Perhaps no area requires as much personalization as health care. The topic of health care arose throughout the Britt Series as participants, students and faculty alike probed how amassing a greater amount of real-time information might remake health care in the U.S. and beyond.
Robert Mead, SVP of marketing, product and communications at Aetna Inc., said the company is developing apps and other tools aimed at helping patients better track their health care data. “You’ve seen people go into doctors’ offices with shoeboxes full of files,” he said. “If they’re a caregiver they go with their elderly parent — they take two boxes full of prescriptions to sort out. What’s she on? What’s she taking? And is this right? And I think anybody who has an elderly parent has been through that where you’ve had to go to their medicine cabinet and you look up and you say ‘Oh my God. Are they really taking all these things? Are they taken at the right time?’
“And it’s really about this thing,” Mead said, holding his smartphone before the audience. “It’s really about … the mobility and convenience of that information and that support and that advice and that help that really gets people engaged.”
Michael Palmer, head of innovation at Aetna, discussed the types of companies the insurer has acquired to allow the company to offer more personalized care for patients. “What we hope to do with this kind of big data analysis with these companies is to allow them to put really personalized intervention programs in for the individuals in the population. And as we are able to aggregate all this data and all these data points we think this is going to drive a higher engagement of the population in their own health,” he said.
Koa Labs’ Palmer sees health care as offering enormous challenges around big data. He said simple questions such as how many patients a facility has or whose conditions are most acute can prove vexing. More intricate questions are all that much more complicated: “How many of the patients in the hospital today have been readmitted? What percentage of those have been readmitted for things that we could have avoided in some way?” Palmer and others noted hospitals are now using big data to run deeper analytic models and determine how to avoid having to readmit patients.
“We need radically better outcomes,” Koa Labs’ Palmer added. “And much, much greater efficiency. We are not going to incrementalize our way in the health care system to the kinds of improvements and changes that we need. We need radical things and I believe one of the only ways to get there is to make these metrics, this information sort of generally accessible that shows people how inefficient and how poor we are at managing our own health and our own healthcare system.” He envisions real-time information-gathering as being central to how to improve health care in the U.S. “You walk into an ER [and] your medical record should show up immediately,” Palmer said. From there, doctors and other health care providers would have a jumpstart on beginning to diagnose and treat problems.
Seeking big results from analytics
Though it’s still evolving, the role of analytics in the enterprise isn’t new. Big data is allowing enterprises to reimagine what’s possible. It’s also nudging corporate cultures to become more centered on data. The shift at Ford has been palpable. “There is a lot of emphasis on data — on making decisions around data and using that data,” Ginder said. “That really has strengthened the role of analytics at Ford.”
Koa Labs’ Palmer sees big data as supercharging longstanding functions around analytics within the enterprise. Big data analytics, with its real-time dashboards of business metrics, is eclipsing older terminology such as business intelligence and, from the 1980s, executive information systems.
Palmer contends it is still difficult for many enterprises to address simple but important questions. He said executives with whom he speaks often cannot answer question such as “Who is your most active customer today?” or “Who is giving you the most orders today?” Getting such an answer can be complex. Palmer recalled the response from one frustrated executive: “If I was going to answer that I’d have to go to all these systems and I’d have to ask all these people.”
Palmer sees it as critical that answers to such questions be automated so an executive can determine minute-by-minute the most important customer or the cost of inputs such as commodities. Analytics within the enterprise should not be about projects that cost tens of millions of dollars or that produce little-used reports from consultants. “It’s about answering really simple but hard and important questions about your business every day.”
A big play on mobile
Mobility is one of the forces that is not only creating mountains of big data but also increasing what data can do for consumers and enterprises. The pace of growth reflects improving technology for handling all of this information. Cloud computing, for example, makes it possible to gather, store and recall large amounts of information on demand.
Angus offered examples of how MasterCard Advisors is looking to mobile as one of the forces powering new possibilities, in particular regarding location-based services. “Big data is at the front end. Using mobile is at the very front end and so we think about how it’s connected to payments and, therefore, how it’s connected to offering deals,” he said. “The technologies exist [to identify] the SKU you’re standing in front of … or the store you’re walking by or what’s going on. That’s going to provide a huge amount of data that we can marry in.”
MasterCard Advisors’ enterprise clients see mobile and big data as “really powerful ways to get very specific value propositions in front of consumers based on where they are and what they’re doing.”
Mobile data is critical in nearly any market. It is accompanying growth of big data in the developing world. “Everything that I see in payments and in data in developing economies somehow equals mobile at some point or other. It’s just the only thing there and everybody’s got one,” Angus said. “It’s the one thing that we see that enables payments and enables data-acquisition in developing economies. It’s really disruptive.”
Koa Labs’ Palmer noted mobile data in health care can help fill in important gaps in a patient’s electronic medical record. “The administrators in these [hospital] systems need to recognize the fact that most of the interesting information related to a patient does not exist in their EMR system, it’s going to exist on their cellphone.”
Big data and mobile can make a potent mix for enterprises seeking access to well-off consumers. “Mobile, by far, is the data source that has the most potential for changing the way we think about consumers or measure or look at consumer data,” said Alexis Hoopes T’06, director of online merchandising at Nordstrom Inc. “Now every consumer is interacting with a device that is … sending off and receiving data at all times both with you and your store, outside of your store, interacting on different websites. I think there is huge, huge potential there.” Hoopes sees using such information — provided privacy concerns are addressed — as a way to bridge the online experience and the in-store experience.
Big questions about privacy
The forces of personalization and mobility also give rise to important discussions around privacy. Forrester’s Gualtieri offered the example of the U.S. Department of Homeland Security, which is examining video of people’s faces to determine their mood or likelihood of certain behaviors. “That’s really big data, because that’s video data,” he said. “There are just enormous ways of using data here to make it personal.”
There are, for now, limits to how far enterprises will go to engage consumers on a personal level. Series participants saw shifting lines on privacy as one big obstacle.
Hoopes stated Nordstrom doesn’t use video surveillance to monitor customer behavior. However, she sees “huge applications” for such technology. Using video surveillance in stores could allow for some of the experimentation and testing of product offerings that are possible online. Applications might include heat maps or mobile-phone trackers that would reveal how a customer navigates a store, for example. “Having that in-store customer behavior piece and getting the big data elements of that to feed in … could really change some decision-making,” Hoopes said. She cautioned there would be big concerns around privacy so the company would have to ensure those were addressed before pursuing any sort of monitoring beyond what it does to deter fraud and theft and to maintain security.
There are other risks to offering customized, personalized services enabled by big data analytics. Hoopes pointed out the better services get, the more customer expectations increase. Even a now-commonplace service like a recommendation engine has ratcheted up consumer expectations. “You’re searching for something and it is a restaurant and you have no idea how Google just filled in your predictive search,” she said. “That is amazing and exciting the first time it happens. Then you come to believe it and expect that over and over. Now you come to Nordstrom.com and you start typing in our search bar [and] your expectation has changed.”
“Even though we’re all kind of at the beginning [of using big data] there are external factors and things that we’re interacting with daily that we need to be watching and saying ‘Well, OK, what is the next bar?’ I mean at some point one question is ‘What is the incremental value?’ and then pretty quickly it becomes table stakes to just having that good customer experience,” Hoopes said.
Caesars’ Sigala noted big data allows enterprises to combine varying types of information to develop a more complete picture of a customer but that doing so also can raise privacy concerns. “There are opportunities to fill in gaps and leverage third-party information to give us a fuller view of our customer. And that … is where you start to get into ‘Are we crossing the creepy line?’ and so that part is something that for us in a heavily regulated industry, we are very careful about. But there are opportunities and I know there are organizations that do this quite well.”
Koa Labs’ Palmer contends some of the efforts around privacy should be redirected from trying to cordon off information and should instead focus on governance. “What you really want to do is you want to control what people actually do with your information — whether you’re being denied healthcare, whether you’re being denied life insurance, whether you’re being denied job opportunities based on this information.” Pursuing legal remedies for a misuse of information would be more efficient than trying to wall everything off, Palmer offered.
“I’m not saying we shouldn’t try and protect information,” he said. “Security is important. There are a lot of bad actors out there in the world, but … the intuitive and the more subtle thing that matters way, way more is putting in place the regulations that are necessary in order to ensure that people use information the way that it is permissioned by the people who generated that information.”
Palmer suggested health care and other industries with sensitive information around individuals would do well to adopt the model used by credit rating agencies. “Somebody wants your credit report, you’re going to find out about it, right? You’ll get a note and it says, ‘Hey, you know, this person wants your credit report, are you going to allow them to do that?’ It’s a great model for what we need to do in terms of privacy going forward in this country.”
Getting a big boost from transparency
Another element of allaying some privacy concerns is to give consumers direct and accessible answers, according to Kelly. The former Facebook executive said a marketing message that might have been cobbled together by combining sets of data about a consumer can unnerve or annoy the recipient. To avoid this, enterprises should take pains to ensure a message is on-target with the recipient. Just as important is letting consumers “very easily” have questions answered about how they received a message. “That’s all good for the industry in the long run. That’s one of the reasons why we built a privacy infrastructure to allow people to be more secretive with some things if they didn’t want them to be publicly shared but also to make it easy for them to publicly share them if that’s what they wanted to do.”
“The primary obligation that big data collectors have and should have … is transparency, is clarity about the collection processes, the operational processes that are used on that data and … a knowledge about what people have,” Kelly said. He contended that some of these requirements should be promoted through regulation.
Bluefly’s Balakrishnan said consumers can be understanding when they see how they came to receive even a highly tailored ad. “Customers like some information on how you come up with decision making.” He noted even simple declarations such as “you viewed this” or “your friends viewed this” help consumers understand why they are seeing a message. “You want to give a brief rationale to why you did it and that kind of resonates with people who see the outcome.” Consumers also respond when an ad, for example, was the result of an action the consumer took, such as signing up for an email.
“The causal connection overcomes any privacy hurdles,” Balakrishnan contended. “If it is very abstract it kind of gets marketed as ‘We are monitoring you to do X, Y, Z’ then there is a lot of suspicion about it. So, privacy really is the question about marketing.” He still prescribes a cautious approach. “You cannot be in the e-commerce space without having a very high level of trust with your consumer. Because once you drop the ball you’re toast.”
Buxton sees crafting the right message as key and believes some privacy concerns are overblown. “I don’t even think it’s an issue. If you target the right message to the right person, they’re excited to see the message. If you target the wrong message to somebody then they don’t like that information.”
“I get marketing pieces that I enjoy all the time. I actually look forward to some of those things,” Buxton said. He noted the company does business with a couple thousand retailers and doesn’t see privacy as a big worry. “I never hear the word creepy. I never hear the word invasive. I just don’t hear it. People like to be talked to about something they care about.”
“By looking at the lifestyle characteristics, looking at how you act, what you look like, you can target any specific group that you can imagine,” Buxton noted.
In some areas, a lack of understanding about what is public information can lead to surprises. NationBuilder’s Murphy said in areas such as politics, voters can be receptive to well-tailored messages though are often surprised by them. “I don’t think many consumers or voters understand both what is in the political sphere public record already and don’t understand how big data works so it’s sort of the scare of how targeted these messages are becoming.”
Cobbling together information about voters is now big business, Murphy said. “That data, whether they like it or not, is public record and it’s being bought and sold by companies. It’s being bought and sold by political consultants.” Campaigns at all levels will continue to rely on more detailed information about voters. “We’re seeing city council candidates use big data. We’re seeing someone running for the school board using big data,” Murphy noted. “In small elections where there are going to be thin margins or you only have to go out and contact 3,000 or 4,000 voters, big data can allow you to run no-money campaigns.”
Reaching voters in an effective way can serve democracy, Murphy said. “I definitely don’t think … in the political world having your message better tailored is creepy. I think it’s only good for democracy and only good for campaigns to actually get their message across.” The flyers that once blanketed entire districts are now far more focused thanks to the analytics tools campaigns are deploying. This is more effective than spending money on broad media buys. “Being able to target in a more efficient way and get your messages that people actually want to hear is in fact a good thing even if there’s going to be sticker shock for a while.”
Nordstrom’s Hoopes believes privacy standards are an ever-shifting line. That is in part because mores around still-emerging areas like social media and mobile technology are evolving. “What peoples’ privacy concerns are in five years are going to be completely different.” This makes it difficult for enterprises to determine the types of big data and analytics systems in which to invest.
Angus, from MasterCard Advisors, said the company stands well back from what it ventures is the threshold of consumers’ privacy tolerance. “We walk very far from that line for now until it gets clearer to us and everybody else where a good place to be is, where consumers are comfortable, where regulators are comfortable, where other groups are comfortable,” Angus said.
“We don’t ever get to individual tagging. All the data we look at is anonymized and aggregated,” he said. “There’s a lot of value we probably can’t create. But we’re very worried about walking too close to that privacy boundary and we don’t know where it is yet. We don’t know what people are comfortable trading for giving up some of their privacy either implicitly or explicitly. And so right now we’re probably overly cautious and we found that we can generate a huge number of insights at the aggregate level.”
Concerns about making big mistakes
Working with big data can uncover obstacles beyond privacy. There are limits to what even huge data sets can do. “Doesn’t it sound great? You just get all this data, and you create this model, and you magically create all this knowledge. But it does have its limits,” Forrester’s Gualtieri cautioned.
He highlighted the example of algorithmic trading on Wall Street. Complex models help determine when to buy and sell all manner of investments. Yet for all the knowledge built into such models, it is still often unclear what pushes investments up or down. The algorithms can react to price movements but not necessarily predict them with reasonable success. Gualtieri noted other systems with many fluctuating data points — such as the weather — are difficult to nail down.
Predictive analytics also can hit roadblocks when data sets are insufficient in scale. As there have only been 57 U.S. presidential elections, there aren’t enough data points to draw sound conclusions. “You don’t have a lot of experience data to figure out what a predictive model is, not to mention what the causative factors are,” Gualtieri said.
Nordstrom’s Hoopes underscored the dangers of a small sample size. “One of the hardest challenges we have is we have the most information about our best customers, so that’s a hard sample to really test and draw widespread conclusions from — the people who are your most loyal and most engaged. So, that can lead you to a lot of false assumptions for the rest of the customers that you’re serving.”
In addition, it can be hard to know all the ways customers are interacting with Nordstrom or any retailer with both a bricks and mortar and online presence. “Our data challenge is in identifying the customer and all the touchpoints,” she said. “If we were online-only there’s a huge advantage that you have in that … You create a unique identifier that you tend to stick with.”
“A lot of our data is based on our physical stores that we’ve had for years and years. So, how do you start connecting all those points and get data down to the individual?” Still, enterprises don’t need to nail down every data point on a customer to be at least somewhat effective, Hoopes posited. “We don’t have to get to 100 percent,” she said. “The incremental value of getting really granular for my business is probably not there. I need to get you in the ballpark.”
In recommending products, for example, Hoopes suggests getting close enough can work. “I need to get it right enough that I don’t turn you off,” she said. “I have a couple of opportunities to say I know you. The first time I get it really, really wrong, you’re out.”
Other obstacles for using big data involve not managing human interactions but human nature. Series speakers noted executives often want to rely on instinct as they have in the past. Gualtieri, from Forrester, argued even executives who push for data-driven decisions through processes like Six Sigma can disregard what does not align with instinct. “Then when a big decision came up, they’d throw that process out and … they’d make [decisions] based upon their instincts or their gut or whatever happens on the golf course.”
Ford’s Ginder said simply managing big data can be daunting. “Can you capture and store all this data that is coming at you? This really is maybe the differentiation of big data in our minds.” Ginder noted Ford’s factories and the cars themselves still produce huge quantities of information every day that are not yet stored or analyzed for insights. The company is changing that though challenges remain. “Can you access these data later? Can you retrieve them, search them, integrate them and, especially, visualize them?”
“We’re always looking for better methods to attack these volume, velocity, variety, challenges,” Ginder said. “How can we store more and access it quicker? How can we merge data more automatically from these different sources, especially at Ford where we have a lot of legacy sources of data?” The next difficulty is in running real-time analytics or near-real-time analytics. “How can we transition these big data kind of opportunities into our algorithms that we’ve built over the years for machine learning, artificial intelligence, and so on? That’s a big challenge for us.”
Ginder sees visualization as an instrument for tapping into the beneficial aspects of intuition. “How do you convey the right, the most informative messages to the consumers of that data?” he said. “With the right visualization you can develop intuition that can aid your decision making in the future. So I think with visualization that’s the key to avoiding confusion with big data.”
Bluefly’s Balakrishnan noted big data renders answers that are not always clear-cut. “It’s always important to recognize that big data doesn’t give black and white answers,” he said. Often, enterprises then have to evaluate how confident they are in the data producing the decisions. “You also have to be cognizant of the fact that the results might change over time. So, just because you got a result today doesn’t mean that result is going to hold in two weeks or three weeks. So, you have to keep adapting.”
Big challenges around scrubbing data
Even once information is captured, executives outlined difficulties around kneading the data into a form that is useful for combining with other sets of information. “Can you process it? Can you cleanse it, enrich it, and analyze these data just as we did before with not-so-big data as well?” Ginder asked.
MasterCard Advisors’ Angus said it often is difficult to collect and cleanse data to make it useful. “In risk and fraud we’ve been doing it for 20 years and we’re just starting to get good at it,” he said. “To us it’s that intersection point between the business and data that’s so hard.”
Sigala explained Caesars employs a logistics group that attempts to distill data sets into the cleanest state possible. He noted this requires a set of skills that are “dramatically different” than much of the rest of the organization.
Forrester’s Gualtieri sees a particular challenge for IT because the appetite for ever more data can clash with constructs around data governance. “IT wants to govern the data and control it and they’re used to creating these well-thought-out models. And data science doesn’t work that way. Data science says ‘Give me everything you got and I’m going to run one of these algorithms against it.’”
Sigala sketched out challenges. Integrating the new information into operations can be vexing, particularly for companies that aren’t primarily online businesses. In the internet world, it is customary to rely on metrics for many aspects of operations. “In more traditional bricks and mortar type environments it’s not. You shouldn’t take for granted that the analytics is going to have an equal seat at the table versus all the other elements of the business,” he said. “The decision-making process, regardless of how compelling an analysis may be, isn’t a given.”
Koa Labs’ Palmer said the continued rise of big data will unearth opportunities around working existing data, not just the new sources of information presented by an ever more connected world. “There are really huge opportunities in data integration and I think you’re going to see the next wave of new innovative companies and startups coming out not to store data but how to take the data that already exist, integrate the semantics about that data and actually sort of surface real interesting data to people and as real time as possible.”
A future with big promise
The sources of big data will continue to mushroom. Rentrak’s Chemerow described the company’s recent move into China. The company can now measure TVs, movies and other video entertainment in the world’s most populous nation. The result? Another form of big data: “Four years ago they had zero digital set-top boxes. Today they have 155 million digital set-top boxes.”
Ford’s Ginder explained the range of devices being connected to the internet — the so-called internet of things — feeds the growth of big data and the possibilities around it. “We see huge opportunities in big data,” he said. “Yes, the techniques for extracting value from data analysis have been around for a long time but now this front end, this handling of these large volumes and speeds of data is a challenge that we’re tackling. Big data, again, is going to touch everything. … The real opportunity is to help optimize operation of all kinds of systems. We’re focusing, and I’m sure everyone is, around value and how we get value from these data.”
Buxton noted the buzz around big data is giving way to action. “We’re still in an environment and a world that is not using big data appropriately to make great decisions,” he observed. “That’s changing.”
Increased personalization of goods and services for consumers will continue, Britt Series participants predicted. “Firms that now make things personal, those are the firms that are going to thrive in the future, and the other ones are going to kind of drop off,” Forrester’s Gualtieri offered.
He believes the consumer experience is what makes the marriage of big data and predictive analytics so exciting. Gualtieri cautioned against expecting too much too soon, however. “Those predictive models have limits. They’re not going to predict the weather right now. They’re not going to predict presidential elections. They’re not going to predict the stock market. But it’s just one tool that companies can use to create competitive advantage.”
Google’s Pichette sees big data and the types of insights that can be generated as the type of leap forward that can help achieve goals that appear impossible. “Already enlightenment has really delivered amazing riches to humanity. If you think about it, it’s built on this absolute fundamental optimism of the human capability to understand the world in which we live in, to shift from dogmas — all these beliefs that are either set on myths or traditions or interpretations — but in fact trusting facts. And if you have trusting facts as a premise, is there no better world than the world of big data to do it in?”