Enterprise data management

The role of data for businesses is changing dramatically today. Many companies that have used data as a specific part of their operations f...

The role of data for businesses is changing dramatically today. Many companies that have used data as a specific part of their operations for years are now discovering a data revolution: data is coming from new sources, being applied to new problems, and becoming a key driver of innovation.

One innovator is The Weather Company (TWC). This media company started in 1980 with a television channel, The Weather Channel. Since then, it has branched out into third-party publishing platforms, websites, and mobile apps, including the one I use every morning to decide whether to pack an umbrella. Like most media companies, TWC is in the business of making content that draws an audience and selling ads that are placed in that content. Data has always been part of that business model: every day vast quantities of weather data need to be captured, analyzed, and turned into the colorful charts, animated graphics, and reliable forecasts that keep audiences tuning in.

But TWC has discovered that its data can be much more than just the raw material it uses to create programming for its viewers. The same data that the firm collects, manages, and analyzes constitutes a key strategic asset and, increasingly, a source of new innovation and value creation.

I learned about this in detail from Vikram Somaya, who was the general manager of WeatherFX (later renamed WSI), a new TWC division focused on thinking differently about weather data. Somaya was an art history major in college and is fond of quoting Shakespeare, but at TWC, he led the teams of data scientists who analyze the company’s data to generate additional value for both business customers and end consumers.

Weather has a powerful impact on a wide range of businesses. By one estimate, up to one-third of the U.S. economy is shaped by variations in weather.1 Walmart has said that local weather is one of the biggest factors in its predictive models for store sales. TWC’s data scientists work with major retailers to identify when they should predict a spike or slump in their sales so they can adjust their advertising spend (to commit more resources or to hold them back) as well as their merchandising.

The company also works directly with brand advertisers in categories like allergy medication, fleece jackets, and snow tires to predict the best time for them to spend on ad placements. Even our snack food purchases on a given day (nacho chips or pretzels?) have been found to be shaped by whether the weather feels bright, sticky, or gloomy. With digital advertisements (inserted on websites or in apps like TWC’s own), brands now have the opportunity to adjust and target their message on the fly, choosing which image to show specific viewers based on the weather where they are standing.

TWC is even using its data to create new products and services for industries like the insurance sector. For instance, it has built an app called Hailzone for insurers like State Farm and Travelers to offer their auto insurance customers. Whenever a hailstorm is about to hit, Hailzone sends out a text message alert to those customers, warning them to move their cars inside. That saves a tremendous headache for the drivers and costly hail damage bills for the insurer.

The company is even collaborating with some of its most avid customers to grow and improve its data asset. Every day TWC crowdsources data from a community of 25,000 self-described “weather junkies” who pay to subscribe to a service called The Weather Underground. These avid hobbyists spend hundreds of dollars to buy their own weather-monitoring equipment, which they set up on their own property. Findings are shared and discussed among the network of fellow enthusiasts. With typical members uploading weather measurements at their own locations every 2.5 seconds, their input helps the company greatly improve the quality of its own data sets.

TWC has evolved from a media company that simply produces data as part of running its core operations to a company that is treating data as a source of innovation, new revenue, and strategic advantage.

Rethinking Enterprise Data Management

 The third domain is data. Growing a business in the digital age requires changing some fundamental assumptions about data’s meaning and importance (see table 4.1). In the past, although data played a role in every business, it was mainly used for measuring and managing business processes and assisting in forecasting and long-term planning. Data was expensive to produce through structured research, surveys, and measurements. It was expensive to store in separate databases that mimicked silos of business operations. And it was used primarily to optimize existing operations.

Today, the role and possibilities for data are seemingly limitless. Generating data is often the easiest part, with great quantities continuously created by sources outside the firm. The greater challenge is harnessing this data and turning it into useful insights. Traditional analytics based on spreadsheets have given way to big data, where unstructured information joins with powerful new computational tools. But for data to become a real source of value, businesses need to change the way they think about data. They need to treat it as a key strategic asset.

Data as Intangible Asset

For many of the digital titans of today’s business world, it seems clear that the data they capture regarding their customers is one of their most valuable assets. Much of Facebook’s market capitalization is rooted in the value of the rich data it collects on users and in its ability to harness that data with innovative tools for advertisers, helping them understand and reach precisely the right audience.

Data is valuable not just for companies like Google and Facebook. For any business today, data like intellectual property, patents, or a brand is a key intangible asset. The relative importance of that asset will vary somewhat based on the nature of the business (just as brands have greater importance to a fashion company than an industrial manufacturer). But data is an important asset to every business today and neglected at our peril.

But other kinds of data can be valuable as well. In building its Maps service, Google has invested heavily for years in developing a best-in-class set of cartographic data. This includes sending camera-equipped cars around the world to measure out every road and capture its photographic Street View (more recently, it has sent cameras by camelback to map the deserts of Arabia). The company is constantly updating and “hand-cleaning” its data with teams of human data wranglers. It tracks up to 400 data points per road segment (the stretch of asphalt between two intersections). Depending on the pace of economic development, that road data needs to be updated with daunting regularity.

On the other hand, we saw Apple’s failure to invest sufficiently in mapping data which led to a famous competitive fumble in 2012. As part of its ongoing rivalry with search giant Google, Apple chose to remove Google Maps as the default mapping app on all iPhones. Instead, it gave iPhone customers its own new Maps app, running on data Apple had purchased from various third parties. True to form, the Cupertino company had designed a stunning user interface for its app. But it had underestimated the quality of Google’s data asset. Millions of iPhone users who were forced to use the new maps flooded Apple with complaints. Cities were misspelled or erased, tourist attractions were misplaced, famous buildings disappeared, and roads literally vanished into thin air. The errors were so bad that they compelled the first letter of apology by an Apple CEO to customers. In it, Tim Cook went so far as to advise customers to download and use competitor apps from the App Store until Apple’s own maps improved. 

Data is valuable not just for companies like Google and Facebook. For any business today, data like intellectual property, patents, or a brand is a key intangible asset. The relative importance of that asset will vary somewhat based on the nature of the business (just as brands have greater importance to a fashion company than an industrial manufacturer). But data is an important asset to every business today and neglected at our peril.

One of the most common ways that businesses can build an asset out of customer data is through loyalty programs. For years, retailers and airlines have offered loyalty miles, points, rewards, or a tenth sandwich free in hopes of increasing customer retention and total spending over time. But, today, much of the value of loyalty programs is in the accumulated customer data that they generate. When I sign up for your loyalty program, I am explicitly asking you to track my shopping behavior in order to earn rewards. That gives your business much more than an address for direct mail; your data about me grows over time to help you better understand my unique behaviors and interests as a customer.

By designing new customer experiences with data in mind, companies can extend this model of providing customer benefits in return for customer data gained. Take Walt Disney Parks and Resorts and its new MagicBand wristbands. Promoted as a way to bring the convenience of smartphones in to the traditional theme park experience, these colorful rubber bracelets (outfitted with RFID tags) allow guests to enter the park, unlock their hotel room, purchase meals and merchandise, and skip the wait on up to three rides per day. The MagicBand is the heart of a $1 billion initiative to bring digital interactivity to Disney theme parks, and it aims to earn that money back by increasing the “share of wallet” that visitors spend at Disney. But it is also designed to provide Disney with previously inaccessible data on the behaviors of its guests: Where do they go when? Which rides are popular with which types of guests? Which foods might be better moved to different areas of the sprawling park? The MagicBands even allow guests to opt to be identifiable to Disney staff so that a child can be greeted by name by costumed characters or offered a birthday wish by a talking animatronic animal on a ride. These and other types of personalized service experiences will become available as Disney builds more data around its visitors on both the large scale and the individual level. The trick is in crafting the right experience so that, just as with a loyalty program, customers willingly exchange their data for added value from the business.

You don’t have to be a company as large as Disney or Google to start building your data asset. Even small businesses can now use Web-based customer relationship management tools to keep track of who opened which e-mails, tailor follow-up messages, analyze which offers are the best fit for which customers, and more. As we will see in our discussion of big data, the shift to cloud computing is putting ever more powerful data management tools into the hands of small and mid-sized businesses.

Every Business Needs a Data Strategy

 Once you start to treat data as an asset, you need to develop a data strategy in your organization. That includes understanding what data you need as well as how you will apply it.

An explicit data strategy may seem obvious in industries like financial services and telecommunications, which are accustomed to copious amounts of customer data. But smaller firms and those in less data-rich industries must also develop forward-looking strategies for their data.
The following five principles should guide any organization in developing its data strategy.

Gather diverse data types: Every business should look at its data asset holistically and include diverse types of data that serve different purposes (see table 4.2). Business process data such as data on your supply chain, internal billing, and human resources management is used to manage and optimize business operations, reduce risk, and comply with reporting requirements. Product or service data is data that is essential to the core value of your products or services. Examples include weather data for TWC, cartographic data for Google Maps, and the kind of business data that Bloomberg provides to business customers. Customer data ranges widely from transaction data, to customer surveys, to reviews and comments in social media, to customer search behavior and browsing patterns on your website. Companies that do not sell directly to consumers (e.g., packaged goods companies) traditionally could gather customer data only through market research. As we will see later, even these businesses are discovering new opportunities to piece together data to get a much clearer picture of their customers than was possible before.

Use data as a predictive layer in decision making: The worst thing that companies can do with data is gather it and not apply it when making decisions. You need to plan how your organization will utilize its data to make better-informed decisions in all aspects of its business. Operations data can be used in statistical modeling to plan for and optimize the use of your resources. Customer data can be used to predict which changes in your services or communications may yield improved results. With detailed data from its MagicBands, Disney can make better-informed decisions on which merchandise to feature near different rides and how to manage variable demand and foot traffic. Amazon uses your past browsing behavior to determine which products it should show you in your next visit.

Apply data to new product innovation: Data can power your existing products or services, but it can also be used as a springboard for imagining and testing new product innovations. TWC’s Hailzone mobile app is a perfect case of a company using its existing product data (for its TV shows and apps) to build a new service that added value for multiple customers (insurance companies and their insureds). It helped that TWC was able to step outside its normal perspective as a media company and think about different business models based on things like utility and risk management rather than just viewer eyeballs and advertising. Netflix uses its vast amounts of data on viewer preferences for genres, actors, directors, and more to help it craft new television series like House of Cards. This practice lets Netflix circumvent the traditional network TV practice of investing in pilots for numerous new shows in hopes that one or more will pan out. That’s using data to innovate more quickly and cheaply.

Watch what customers do, not what they say: Behavioral data is anything that directly measures actions of your customers. It can include things like transactions, online searches (a powerful measure of your customers’ intentions), clickstream data (which pages they visited, where they clicked, and what they left in their shopping carts), and direct measures of engagement data (which articles in your newsletter they clicked to read). Behavioral data is always the best customer data it is much more valuable than reported opinions or anything customers tell a market researcher in a survey. That is not just because people lie in surveys but also because, as humans, we are extremely fallible at remembering our behavior, predicting our future actions, or considering our motivations. This is why Netflix shifted its recommendation system from customers’ own rankings to behavioral data as soon as it moved customers from DVDs to streaming video, which made it possible to measure what we actually watch rather than the unopened red envelopes on our dresser. Netflix knows that there are big differences between the movies that we give a five-star ranking and those that we actually wind up watching while doing the dishes on a Wednesday night.

Combine data across silos: Traditionally, businesses have allowed their data to be generated and reside in separate divisions or departments. One of the most important aspects of data strategy is to look for ways to combine your previously separate sets of data and see how they relate to each other. A memorable example of the benefits of combining data sets comes from municipal government here in New York City. Scott Stringer, the city’s comptroller (CFO), was seeking to reduce the costs of lawsuits against the city. He launched an initiative to compare the data on lawsuits and damages paid with other city data sets, including the budgets of different departments over time. A surprising correlation was discovered: after the city’s parks budget had been slashed a few years earlier and its seasonal tree pruning reduced, legal claims from citizens injured by falling tree limbs skyrocketed. The cost to the city from a single lawsuit was greater than the entire tree pruning budget for three years! Once this was discovered and the budget funding was restored, lawsuits dropped dramatically. As your business environment becomes increasingly complex, your ability to find, combine, and learn from diverse sources of data will become more important than ever.

In putting together a data strategy, it is also important to understand that many of today’s data sets are very different from the spreadsheets and relational databases that drove the best practices of data-intensive industries in the pre-digital era. The entire nature of available data, and how it can be applied and used by business, has undergone a revolution in recent years. That revolution is commonly termed big data.

The Impact of Big Data

The term big data first appeared in the mid-1990s, introduced in tech circles by John Mashey, chief scientist of Silicon Graphics, around the time of the birth of the World Wide Web.5 But the phrase entered the broader business conversation around 2010 as businesses of all kinds began to grapple with the vast supply of data generated by digital technologies. At first, the term seemed a bit faddish, a marketing ploy used by data storage firms to get IT departments to increase their spending on data servers. But the real changes at work have been much more profound than the size of hard drives or server farms.

Make no mistake: the size of data sets is increasing rapidly. Every graph representing the amount of digital data stored worldwide each year shows the skyward leap of an exponential curve. These curves all recede exponentially into the past as well. The sheer amount of recorded data, in other words, has been growing for a long time likely since the origin of computers, maybe since the origin of writing.

So what is new about big data if not the rapidly growing “bigness” of it?

The phenomenon of big data is best understood in terms of two interrelated trends: the rapid growth of new types of unstructured data and the rapid development of new capabilities for managing and making sense of this kind of data for the first time. The impact of these two is shaped by a third trend: the rise of cloud computing infrastructure, which makes the potential of big data increasingly accessible to more and more businesses.

Big Data Is Really Unstructured Data

Traditionally, a firm’s data processes were based on analyzing structured data the kind of data sets that fill a database with neatly organized rows and columns (e.g., with addresses of customers, inventories of products, or expenses and debits of various financial accounts).

But the big-data era has been marked by the profusion of new types of unstructured data information that is recorded but doesn’t fit easily into neat forms. A business may have access to the ungrammatical text posts of social media, the flood of smartphone-generated images, real-time mapping and location signals, or the data from sensors rapidly spreading over our bodies and our entire world; all these types of data are rich in meaning but difficult to parse by familiar tools like spreadsheets.

One of the biggest sources of unstructured data is social media. As over a billion users worldwide participate in networks like Facebook, Twitter, and Weibo, they are constantly producing vast amounts of data in the form of their posts, comments, and updates. This social data is attitudinal (what people are saying can capture their opinions, likes, and dislikes) and can be used to measure affinity (whom they friend, follow, or link to reflects social ties and allows businesses to infer relationships between them and others in their network). And this data is real-time and continuous, allowing businesses to analyze shifts in opinion, sentiment, and conversation with precise longitudinal detail. Because of this, numerous organizations have sought to gain insight from the analysis of social data. Brands monitor their reputation over time based on what customers are saying, the Centers for Disease Control uses social media to help track the spread of flu and influenza, Hollywood predicts the opening weekend performance of new movies based on the social “chatter” after opening night, and economists have even used social media to effectively predict stock market performance.

Another new kind of unstructured data is location data. The data being generated by mobile devices like smartphones comes with geolocation markers, which provide a continuous record of where we are and where we’re going in real time. The inclusion of location data with other kinds of behavioral data adds tremendous additional context. Increasingly, search engine results are shaped not just by the words we are using in our search but also by where we are when we search. (If we Google the word pizza, we are likely to be shown the closest establishments, with links to their phone numbers and addresses, instead of pizza history or recipes.) Research by my colleague Miklos Sarvary has shown that the patterns of where we go at various times of the week (as measured by our phones) reveal a great deal about who we are. By analyzing these “co-location” patterns, Sarvary and his coauthors were able to show that customers with similar location “footprints” were likely to buy similar products and could be effectively targeted for marketing based on that data alone.

The biggest emerging source of unstructured data is the sensors that are becoming embedded in everything around us as we shift to a world of truly ubiquitous networks. By 2020, Cisco expects that over 50 billion devices will be connected and sharing information over the Internet and the vast majority of these devices will not be computers, smartphones, or Web servers. This phenomenon, known as the Internet of Things, encompasses smart automobiles, factories and product supply chains, and lightbulbs and home appliances as well as sensors embedded in the watches and clothing we wear and in the medicines we ingest. Together, all of these applications will soon result in billions of devices transmitting and generating new sets of data that can be put to business use. For example, GE has installed sensors on its jet engines that allow the engines to continuously post updates on their status and operating details. (GE calls the system “Facebook for jet engines.”) This real-time data lets airline mechanics monitor the status of critical aircraft equipment so they can make repairs when they actually are needed rather than on a schedule of estimated need. This makes fleet maintenance more efficient and makes air travel cheaper and more convenient.

New Tools to Wrestle Unstructured Data

The second trend shaping big data is the rise of new technological capabilities for handling and making sense of all this unstructured data. If not for this, big data would be simply a giant haystack in which the needle of business insight might well be invisible. Fortunately, a range of technological developments is expanding our abilities to use the unstructured data that technology is producing.

The continuing exponential growth of computer processing power is a big factor in our improved ability to use data. Moore’s law, coined by Intel cofounder Gordon Moore in 1965, predicts a doubling in the performance of computer chips roughly every eighteen months as transistors become faster and smaller. For fifty years, the prediction has held, and the results have transformed the world. ENIAC, the first modern computer, was built in 1946 and filled a room the size of a small gymnasium. But by 1983, when I first studied computing, my student-grade Texas Instruments pocket calculator had more processing power than ENIAC. Moore’s law tells us that this decade’s supercomputer is the next decade’s pocket device.

Recent technologies have further enabled data processing on a large scale with acceptable costs. In-memory computing can accelerate analytics to the kind of real-time computing that allows digital advertising to select the ad seen by each visitor to a webpage, based on the weather where they are, the sites they have visited recently, or any other critical determinants that can be mined through data. Hadoop is an open-source software framework that enables distributed parallel processing of huge amounts of data across multiple servers in different locations. With Hadoop, even the biggest data sets can be managed affordably.

Other tools focus less on increasing power and more on making sense out of the chaos of unstructured data. New data-mining tools allow programs to sift through the raw stuff of social media and pick out patterns that human managers then can examine to recognize trends and key words.

Perhaps the biggest advances in managing unstructured data have come from new developments in “cognitive” computing. Natural language processing, for example, can interpret normal human language, whether from spoken commands, social media conversations, or books or articles, without adaptation. It is critical to the development of systems that can identify patterns in big-data sets of human language, such as recordings of customer phone calls to call centers. Another key development is machine learning resulting in computing systems that can recognize patterns and improve their own capability over time, based on experience and feedback. As computers are modeled around neural networks, they go beyond just spotting patterns in unstructured data: they receive feedback from their environment or human trainers (indicating which conclusions were wrong and which were correct) and reprogram themselves over time.

Natural language processing and machine learning are combined in a system like IBM’s Watson, which can read vast amounts of written language and develop ever more accurate inferences by using feedback and coaching from human experts. Watson famously debuted on the world stage by playing the quiz show Jeopardy! where it bested the top human champions by combining encyclopedic recall with a human-like ability to have educated “hunches” (e.g., estimating that its best guess to a question had a 42 percent likelihood of being correct). Since then, Watson has moved to the real world. Physicians have trained Watson, using a library of millions of patient case histories, to the point where Watson is more accurate than many doctors in making an initial diagnosis of a new cancer patient. Watson and similar technologies will be at the forefront of the next wave of big-data analytics informing everything from customer service, to fraud detection, to advertising media planning.

Big Data on Tap from the Cloud

 An additional trend is shaping the impact of big data: a revolution in the storage and accessibility of both data and data processing. In the old data paradigm, for a business to manage data, it needed to invest in owned infrastructure to collect and hold all of the data as well as any tools to analyze it. This significant capital requirement led to disparities among companies, with many unable to afford the sophisticated use of data. Today, businesses no longer need to store their own data, and even small businesses are increasingly able to access the leading tools for using unstructured data. The reason is the rise of cloud computing.

Think of voice-recognition systems like Siri or Google Now on our smartphones. There is a reason Siri doesn’t work when our iPhones are offline: the computations required to understand spoken language and respond to it are too intensive to be managed with the processors on a current smartphone. Yet Siri works perfectly fine when able to access the cloud. All our device needs is a steady connection so that it can send our voice to a remote server with all the power necessary to process that unstructured data and respond in real time.

Increasingly, more and more computing applications and services are delivered seamlessly over the Internet, with the real processing power residing in the cloud rather than on our devices and computers. Amazon Web Services (the company’s huge B2B computer services division), Microsoft, Google, and others are all driving a shift to a computing environment where businesses increasingly meet their needs through subscription and SaaS offerings rather than by buying and installing the most powerful computers on their own premises.

Cloud computing has profound implications for scalability and small business. Services like Watson are available “on tap” to businesses, just like cloud-based storage and customer databases are for small businesses. This means that big data is not the exclusive terrain of the world-class companies with huge IT departments. Any business can tap into best-in-class analytics tools today from cloud providers like SAP and IBM paying only for the data and the processing it uses. Big data doesn’t have to have a big price tag.

Where to Find the Data You Need

 As you begin to put together a data strategy, you will start with the data you are generating in your own business processes. However, you will likely identify gaps in the data you need for some of your goals. Finding the right additional sources of data is critical to filling in gaps and building your data asset over time. Important sources of data from outside your organization include customer data exchanges, lead users, supply chain partners, public data sets, and purchase or exchange agreements.
Customer Value Data Exchange

Lead User Participation 

Lead users (a term coined by Eric von Hippel) are your most active, avid, or involved customers. Their greater needs lead them to have greater interest in interacting with your products or business, and they can often be a unique and powerful source of data. We saw one example in The Weather Underground: the volunteer army of meteorological enthusiasts who happily contribute real-time feeds of additional weather data to TWC as part of participating in that community. Other companies use exclusivity to identify and leverage their lead users. Alexandre Choueiri, L’OrĂ©al’s president of international designer collections, explained to me that the cosmetics firm creates and engages confidential customer communities for designer brands such as Viktor & Rolf. The allure of joining a special club (literally called the “secret service”) appeals to consumers, and the exclusivity helps the brand learn more about loyal users not just casual one-time purchasers. “You get fewer people,” Choueiri told me. “But they’re really engaged. We sell this brand through the retailers, so this engagement tool is how we get data.” By engaging lead users, brands can solicit input and feedback from much more selective and important communities.

Supply Chain Partners

 Business partners can be crucial sources of additional data for building your data asset. Companies producing consumer packaged goods now work closely with large retailers and with retail data services like Dunhumby. Power, leverage, and levels of trust can greatly influence who shares data with whom in many industries. In the travel industry, large airlines (such as Delta) can have nearly 100 million customers enrolled in their loyalty programs. But airlines and the online travel agencies (such as Travelocity or Orbitz) share only limited data. As a result, neither the agencies nor the airlines have access to the full picture of customers’ travel behaviors when they want to customize pricing and offers at the point of sale. Increasingly, data partnerships will be a key element of how businesses negotiate terms of working together.

Public Data Sets

 Another important source of new data is publicly accessible data sets. Some of these are in online public forums. The car reviews website Edmunds .com, for example, contains many years’ worth of discussion forums providing huge amounts of unstructured data in customers’ conversations about car models, makes, preferences, and experiences. Many social media platforms, like Twitter, are easily searchable for real-time data. In addition, governments are increasingly providing public access to large data sets in machine-readable format. The U.S. government’s census data, for example, has been in huge demand since being made available. In addition, more and more city governments are opening up APIs to let innovative businesses make use of government data and to spur new business opportunities.

Purchase or Exchange Agreements

Lastly, there are many opportunities for businesses to purchase or swap legitimate, valuable data with other firms. Businesses should avoid companies that offer shady sets of customer records collected through questionable means. Instead, firms should seek out the many reputable services that enable anonymized data comparisons. Anonymized data lets a company learn things like the conversion rate of offers (the portion of customers accepting the offer sent). The company’s data shows which customers got the offer, the retailer’s data shows who made a purchase, and the third-party service measures the conversion rate without revealing customer identities (which could be a violation of privacy terms).

Sometimes data can be received through an exchange or donation. During the 2014 World Cup, Waze shared anonymous driver data with city governments in Brazil to help them identify and respond more quickly to traffic buildups and road hazards. In Rio de Janeiro alone, up to 110,000 drivers a day were providing traffic data through Waze’s API. Since then, Waze has been developing partnerships with other governments, such as the State of Florida. The company is not asking for payment but rather is seeking an exchange of more data. By receiving real-time data from highway sensors and information on construction projects and city events, Waze is improving its own data asset.

There are many more sources of data available today. The challenge for your business is often simply choosing which ones will best fit your needs. A recent forecast published by the Journal of Advertising Research summarized the changes anticipated in market research: as businesses are faced with a “river” of continuously generated data, the goal of research is not to expensively manufacture data, but to find the right tools to “fish” in that river in order to draw forth the insights and intelligence needed.

Turning Customer Data into Business Value: Four Templates 

As organizations gather more data and develop it into powerful assets, the next challenge is to continuously apply these assets to create new value for themselves.

We’ve seen examples of how product or service data provides value by enabling a business’s core service to customers: think of TWC’s use of weather data and Google’s use of mapping data. We’ve also seen that business process data can yield value by optimizing and improving decision making, even in surprising ways like Stringer’s use of budgetary data.

If we look at customer data, we can find recurring patterns of best practices used to add value across differing industries and organizations. We can think of these practices as four templates for creating value from customer data: insights: revealing the invisible; targeting: narrowing the field; personalization: tailoring to fit; and context: providing a reference frame.

Let’s take a look at each of these four data value templates and see how they are applied in different industries to create new value.

Insights: Revealing the Invisible

 The first template for value creation is insights. By revealing previously invisible relationships, patterns, and influences, customer data can provide immense value to businesses. Data can provide insights into customer psychology (How are my brands or products perceived in the marketplace? What motivates and influences customer decisions? Can I predict and measure customer word of mouth?). Data can reveal patterns in customer behavior (How are buying habits shifting? How are customers using my product? Where is fraud or abuse taking place?). Data can also be used to measure the impact of specific actions on customers’ psychology and behavior (What is the result of my change in messaging, marketing spending, product mix, or distribution channels?).

Today, many businesses have access to large quantities of customer data in the form of online conversations about their products and brands. A good example is automobile manufacturers. My colleague Oded Netzer of Columbia Business School, along with three research coauthors, has dug into the data created by discussion forums to explore what it reveals about the automotive market structure and consumer behavior. Netzer’s team applied a variety of text-mining tools algorithms that are trained on human language and apply formulas to detect patterns in huge quantities of unstructured text from online conversations. One area of their research looked at how customers perceive brands. By examining patterns of statistical “lift,” they could identify which specific attributes are more frequently associated with one auto brand versus its closest competitors. The patterns revealed opportunities in terms of audiences to target, content for messaging, and ideas for product development.

Netzer’s team also used the data to investigate the impact of long-term advertising efforts. They focused on a period when Cadillac had spent millions on brand advertising to shift customers’ perception of Cadillac from “classic American car” (like Lincoln) to “luxury brand” (like Lexus and Mercedes). A textual analysis of the conversations over several years showed that, consistent with the campaign objective, the Cadillac brand was gradually moving in customers’ associative perceptions from the first group (classic American brands) to the second (luxury brands). When the researchers compared this with public data on dealer trade-ins, they confirmed that the shift in perception was also a leading indicator of purchase behaviors. Rather than trading between Lincolns and Cadillacs, more and more customers were exchanging their luxury cars for Cadillacs.

In another case, Gaylord Hotels used insights from customer data to sharpen its referral strategy. The business has a few large hotel properties that are well suited for major events as well as personal stays. With a limited advertising budget, it knew that referrals (word of mouth from happy guests) were the biggest source of new customers. So management set a priority to increase that word of mouth by improving the already good guest experience. The first step was an internal review of operations that identified eighty areas of focus that might help inspire customers not only to be pleased but also to actually mention Gaylord to others. The obvious next challenge was prioritization: Which items on this long list were most important? To help, the company undertook an analysis of social media data, looking at every instance where the hotel’s name was mentioned by customers in public platforms like Twitter. Customer recommendations and praise were examined for any clues as to what had spurred them and at what point in the customer’s stay. The results were illuminating. A short list of just five elements of the guest experience seemed to have the greatest influence in sparking word of mouth, and all of them took place in the first twenty minutes after arrival.

Targeting: Narrowing the Field

 The second template for data value creation is targeting. By narrowing the field of possible audiences and identifying who is most relevant to a business, customer data can help drive greater results from every interaction with customers. In the past, customers were often divided into a few broad segments for targeting based on factors like age, zip code, and product use. Today, advanced segmentation schemes can be based on much more diverse customer data and can produce dozens or even hundreds of micro-categories. How a customer is targeted can change in real time as well, as they are assigned to one segment or another based on behavioral data such as which e-mails they clicked on, rewards they redeemed, or content they shared. Ideally, customer lifetime value should be included as one metric for targeting customers based on their long-term value to the business.

Custora is a data analytics company that helps e-commerce businesses determine the likely customer lifetime value (CLV) of their website visitors that is, not just their likelihood to buy in this visit but their likely profit potential in the future. This is done by analyzing historical customer data and applying both a CLV model and Bayesian probabilistic models. For example, when a new customer makes just one purchase on a website, Custora can predict that they are likely to make six purchases in the upcoming year, totaling $275 and placing them among the top 5 percent of the company’s customers. Other predictions based on historical data include the category the customer’s next purchase will likely come from (e.g., home furnishings vs. lawn care). The model can even provide warning signs such as predicting that if this customer doesn’t place an order for three consecutive months, the business can assume they have only a 10 percent chance of returning.

InterContinental Hotels Group carefully uses data on the 71 million members of its Priority Club loyalty program to understand and target them more effectively. This data includes much more than zip code and hotel room preferences. Up to 4,000 different data attributes such as their income level, their preferred booking channel, their use of rewards points, and whether they tend to stay over weekends are used to assign each member to a customer group. This level of segmentation has allowed the hotel to shift from sending out a dozen varieties of an e-mail marketing message to sending out 1,552 different variations, targeted around past behaviors and special offers such as local events. These new marketing campaigns have generated a conversion rate (the portion of customers accepting the offer sent) that is 35 percent higher than that of less targeted campaigns the year before.

Using data for targeting can even have a powerful impact in a field like nonprofit health care, thanks to a practice known as “hot spotting.” Dr. Jeffrey Brenner, a family physician in Camden, New Jersey, studied medical billing records from hospitals in his hometown and discovered that 1 percent of the town’s population was responsible for 30 percent of its health-care costs. “A small sliver of patients are responsible for much of the costs, but we really ignore them,” said Brenner.17 He used that data, and small grants from philanthropies, to start the Camden Coalition of Healthcare Providers and focus on “spotting” these patients and improving their care. Over three years, the organization was able to reduce emergency room visits by 40 percent among the initial group of the “worst of the worst” patients and to reduce that group’s hospital bills by 56 percent.

Personalization: Tailoring to Fit

 Once businesses are targeting micro-segments of customers, the next opportunity is to treat them each differently, in ways that are most relevant and valuable to them. This is the third template for creating value: personalization. By tailoring their messaging, offers, pricing, services, and products to fit the needs of each customer, businesses can increase the value they deliver.

Kimberly-Clark, which sells some of the biggest brands in diapers (among other personal care products), uses an audience management platform that integrates data from sales and media channels to build an integrated view of the “customer journey” of each customer. For the company’s business, that means tracking a family’s progression through various products from Huggies newborn, to full-size diapers, to transitional pull-ups during toilet training and “Little Swimmers” (for kids just starting out in the pool). Keeping track of each customer allows it to advertise the right product to the right family.

British Airways has launched a service personalization program known internally as Know Me. Its goal is to bring together diverse data to create a “single customer view” that will help airline staff to make a more personal connection with each customer. Know Me started with a two-year project to link data from commercial, operational, and engineering systems and put it at the fingertips of customer service directors. But the program works only because the data analytics are linked to the judgment and “emotional intelligence” of the British Airways service staff. Know Me data is used to deepen staff awareness of fliers’ personal needs and preferences, and staff are empowered to make their own observations and record data that helps personalize future trips. This feedback loop helps the airline deliver more-relevant offers to each customer and provide personalized recognition and service during a trip. That can include recognizing a VIP business traveler even when traveling in coach class with family so that service staff can welcome and thank them and offer a glass of champagne. It could also mean providing discreet assurances to a customer who has previously indicated they have a fear of flying. With urgent updates entered in the system in minutes, one flight crew spotted a passenger’s iPad, forgotten on board, and passed word to the connecting flight crew to notify the passenger. One of the most popular service touches has been that of welcoming customers mid-journey when they have reached Silver Tier status, the first level that offers access to lounges. The airline has seen extremely positive response from customers, both one-on-one and in long-term tracking of their satisfaction and their likelihood of recommending British Airways to others. In addition, Know Me has allowed the airline to broaden its view of customers far beyond its loyalty-program members, with a goal of knowing the needs of all of its 50 million fliers.

One challenge of personalization has been the proliferation of different devices and platforms where customers interact with a business. How does a business know it is communicating with the same individual on a phone, tablet, and PC, let alone through Facebook, its own shopping portal, or a display ad being served up by Google on pages all over the Internet? The good news is that this challenge is diminishing rapidly, allowing for “addressability” of the same customer across numerous platforms. As David Williams, CEO of database powerhouse Merkle, explained, we are quickly becoming able to communicate to individual consumers with “addressability at scale” across Google, Facebook, Amazon, and all the dominant platforms of the Web.

Context: Providing a Reference Frame

 The final template for data value creation is context. By providing a frame of reference and illustrating how one customer’s actions or outcomes stack up against those of a broader population context can create new value for businesses and customers alike.

Putting data in context is at the heart of the “quantified self” movement evidenced by customers’ rising interest in measuring their diet, exercise, heart rates, sleep patterns, and other biological markers. Nike was one of the first companies to tap into this trend with its Nike+ platform, which originally used in-shoe sensors, then the Nike Fuel wristband, and later mobile software apps. At each stage of its development, Nike+ has been designed to let customers capture their data and share it with their online communities. Nike customers who track their running data don’t just want to know how they did today; they also want to know how today’s performance compares to their own performance over the last week or month, to the goals they have set, and to the activity of friends in their social network. Context is king.

Comparing their own data with the data of others can also add value by helping customers understand the probabilities of different outcomes. Naviance is a popular platform for U.S. high school students preparing for the college search and application process. One of its primary services is a tool that lets students upload their transcript data (test scores, class grades, high school attended) and compare it against a huge database of students who have applied to college while using Naviance. Based on the past results of similar applicants, the platform can show students their likely odds for admission to different colleges they are considering. Rather than applying in the dark (as we did in my day), students can use Naviance to find out which college on their list is a long shot, which one is a sure thing, and which schools fall in between.

Sharing and comparing customer data can be a powerful way to identify hazards. BillGuard is a popular financial protection app that tracks its customers’ credit card statements and helps identify both fraudulent billing (e.g., if the card was one of 50 million hacked in the latest cyberscandal) and “grey” charges (hidden fees customers likely didn’t realize a company was charging them). BillGuard’s algorithms are effective precisely because they compare a customer’s bills against the anonymized bills of peers and against whatever charges were flagged as questionable by any other customers in its community.

Other examples of businesses using data for context include Glassdoor, which lets job seekers compare their salaries with averages for others in their industry and role, and Pricing Engine, which helps small businesses improve their digital advertising spending (on platforms like Google AdWords) by comparing their own success rates with those of their peers.

Tool: The Data Value Generator

 We’ve looked now at the different types of data being used in business today. We’ve examined the sources where businesses can find more data to fill in their own gaps. And we’ve seen four templates for generating new value using customer data. Let’s look now at how to apply these concepts to generate new strategic options for data initiatives in your own organization. That is the focus of our next tool, the Data Value Generator.

The tool follows a five-step process for generating new strategic ideas for data (see figure 4.1). Let’s look at each of the steps in detail.

Step 1: Area of Impact and Key Performance Indicators

 The first step is to define the area of your business you are seeking to impact or improve through a new data initiative. You might define it as a specific business unit (e.g., product line), a division (e.g., marketing), or a new venture. You might decide that you are looking to apply data to improve customer service at a resort, to develop better product recommendations, to improve outbound communications to existing customers, to improve the customer call center, or to develop a new app to drive customer engagement.

Once you have defined the area of impact, you should identify your primary business objectives in that area. What goals are you hoping to support? In addition to broad goals, what are your established key performance indicators (KPIs) that are being used to measure performance? Because this is a data-driven project, you will want to think about highly measurable outcomes, those where you may be able to clearly measure impact. It is alright if you identify multiple objectives and KPIs at this step; you may end up seeking to influence one or more as you generate your strategic ideas.

Step 2: Value Template Selection

Now that you know the domain you are focused on, look back at the four templates for value creation, and identify one or more that may be most relevant to your objectives:

  • Insight: Understanding customers’ psychology, their behaviors, and the impact of business actions
  • Targeting: Narrowing your audience, knowing who to reach, and using advanced segmentation
  • Personalization: Treating different customers differently to increase relevance and results
  • Context: Relating one customer’s data to the data of a larger population

Which template is most relevant to your business domain? To the KPIs you are focusing on? Which may affect those goals more indirectly? (For example, insights into customer brand perceptions could help influence a goal of market penetration if you can identify the right opportunity to reposition your product.)

You could choose to pursue one template or a combination. Note that targeting and personalization often work together. Whereas targeting efforts are sometimes focused only on identifying the right audience, effective personalization requires that you have some system of targeted segmenting in place. You may already have one template or another more developed (e.g., you are strong on segmentation but weak on consumer insights). The question is, What area of value creation is the next focus for your data strategy?

Step 3: Concept Generation

Now that you have selected a value template (or more than one), you will want to use it to ideate specific ways that data could deliver more value to your customers and your business.

For example, if you select context, how can you best use contextual information to influence desired behaviors? Behavioral economics has revealed that seeing our data in context can be an extremely powerful motivator. Voters are more likely to be persuaded to make it to the polls when reminded of their own past voting history and that of their neighbors. Using this insight, Opower has developed a data-driven service to influence home power consumption. The company, which works with local utilities, shows consumers data on how their own energy usage compares with that of their neighbors. The result: consumers are much more likely to reduce their energy consumption when shown comparative data.

Concept generation should aim for this level of concrete application so you can really define the possible data strategy. For a personalization strategy, what are the specific moments of customer interaction that you are trying to personalize? For example, hotel and casino company Caesar’s Entertainment has pursued a strategy similar to that of British Airways using data for the personalization of service, starting from a loyalty program and aiming to increase repeat business. But Caesar’s focuses on a different set of moments. For example, Caesar’s can determine when a repeat visitor is having a bad night on the gambling floor and will send service staff to offer an unexpected gift a steak dinner, tickets to a show so the customer won’t leave feeling they had “bad luck” at Caesar’s and should try another casino.

At the concept generation stage, you want to produce specific ideas for putting the data to work in your business.

Step 4: Data Audit

Now that you have a strategy in mind, you need to assemble the data that it will require. That starts with surveying what data you already have that could be used to enable or power your strategy. You may have a large, established data set based on your core product or service (like TWC). You may be starting with a data set on website visitors, or you may have access to loyalty-program data. For some businesses, the only data may be an incomplete list of customer e-mail addresses.

Next you should identify what data you still need. For the purpose of the strategy you have sketched out, what data is still lacking? What will it take to provide the full view of the customer needed by your new initiative? You may need to increase your data in terms of

  • more records or rows (e.g., expanding from a limited sample of your customers to a much broader list),
  • more types of data (e.g., adding preference data and transaction data to your customer contact data), or
  • more historical data (e.g., going back many months in time in order to develop an effective analytics tool that can model and predict future outcomes).

Lastly, now that you’ve identified the gaps, you need to determine ways to fill them. This is where you can apply the options discussed earlier: customer value exchange, lead users, supply chain partners, public data sets, and purchase or exchange agreements.

Step 5: Execution Plan

For your data strategy to be effective, you must do more than assemble the right bits of data (the zeroes and ones). You must put that strategy to use in the work of your organization. The last step is to plan for the execution of the key pieces of your data plan.
What technical issues need to be worked out? This may include data warehousing, latency, or how quickly the data needs to be updated. Your IT people will need to weigh in here.

What business processes will need to change? Most data initiatives assume employees of your firm will make different decisions and take different actions based on your data. You will need to identify those changes in advance of rolling out any technical solution.

How can you test out your strategy and build internal support? One of the best ways is to integrate the new data strategy into an existing initiative at your company. Jo Boswell, the program lead for Know Me at British Airways, knew that it would be difficult to enlist in-flight service staff if her initiative was seen as one more competing priority in their work. Instead, she integrated Know Me with their existing customer service program, showing how its data would help staff to deliver on the same four “customer service hallmarks” that anchored all their training.22 Data-driven strategies should be in line with everything your business is doing and help people to do their jobs better.

Organizational Challenges of Data

When Mike Weaver was brought in as director of data strategy for the Coca-Cola Company, his mission was clear. “We must understand consumers’ passions, preferences, and behaviors so we can market to them as individuals,” he told me. As an expert in the area of applied analytics, Weaver saw that this required building a data asset in an industry that is not traditionally rich in consumer data. By combining its MyCokeRewards loyalty program with a variety of other data sets observed behaviors on its websites, social log-ins via Facebook, cookie stitching, and data from various partners the company was able to advance rapidly toward its goal of becoming a more data-driven marketer.

But the biggest challenges, Weaver told me, were organizational, not technical. He compared the process of shifting business practices at “the world’s greatest brand/mass media company” to turning an aircraft carrier at sea. He knew that the right data models could be used to develop advanced segmentation schemes for Coca-Cola’s customers, to understand customers’ different needs and wants, and to allow the firm to better serve and communicate with them. But before installing all the data centers and analytics models that would allow for real-time targeting of customers, the company first had to plan out the changes to its business processes. Before a brand can take advantage of its ability to differentiate customer segments in real time and deliver targeted messaging to them, it first needs to learn how to create messages in a very different way. This kind of targeting doesn’t require Coke to create a single, blockbuster Super Bowl ad; rather, it has to create dozens of versions of the same message and test them to see which ones drive response among different customer segments. The first step of the journey, Weaver reiterated, is to plan the changes in your business process before you start buying all the latest hardware or cloud services.

In my speaking, and work with a wide range of companies, I’ve observed a number of common organizational challenges that businesses face as they shift to a more data-driven strategy. Each of them is worth considering when developing a data strategy.

Embedding Data Skill Sets

The first challenge in the transition to a more data-driven organization is finding people with the right skill sets.

This starts with data scientists the folks who can do the technical work of data analysis, be it hand-cleaning the raw data, programming algorithms to apply real-time data in an automated fashion, or designing and running rigorous data experiments. Depending on the organization, it may be using an outside partner for analytics, hiring a single analyst, or building an entire team. Good data scientists have strong statistical and programming skills and often come from an academic or scientific background. They also serve as truth-tellers within the organization. These are the folks who know that data can lie very easily, and they will keep a company honest about things like sample size, significance testing, and data quality (the old “garbage in/garbage out” rule).

But the data experts cannot be the only people in an organization who understand or think about data. In order to truly build data into a strategic asset, everyone in the business has to adopt a mindset that includes using data, and the questions they pose to it, as a part of their daily process. Part of this is educating the workforce about the ways data can be applied in their business. Another part must be developing a company culture that embraces data and analytical thinking. For a consumer goods company like Coke or Frito-Lay, that involves a shift from thinking about marketing as an art to thinking about it as a discipline that includes both art and science.

Lastly, the company may need someone who can bridge two worlds: the world of quantitative analysts and that of business decision makers. This person will be the one who can connect the work of data science with that of the senior managers or the creative types in the marketing department. Think of Somaya, the former art history major who learned to speak the language of both the data scientists at TWC and the advertisers and brand managers who were his clients.

Bridging Silos

Sometimes the biggest challenges to sharing data are within the organization. At Coca-Cola, Weaver found that website analytics data was sitting in one database while data on consumer purchase behavior from loyalty programs was being kept somewhere else entirely. In order to create a complete picture of the customer, he first had to bring all the data together in a unified way.

In many organizations, these divisions are reinforced by departmental silos and each department’s desire for “ownership” of its data (sales data vs. marketing data, etc.). In a research study that I coauthored with my colleague Don Sexton, we spoke with hundreds of senior marketers at businesses across a wide range of B2B and B2C industries. The most commonly cited obstacle to using data effectively was internal sharing, with 51 percent of respondents reporting that “the lack of sharing data across our organization is an obstacle to measuring the ROI value matrix of our marketing.”

In large organizations operating in different locations, another important question is whether or not to centralize data analytics. This is partly a matter of where the data is warehoused but also where the data scientists are. Should each business unit have its own analytics team so it is closer to local decision making? Or should one central analytics unit service the key data needs of every part of the business? As large organizations mature in terms of their data capability, they seem to be centralizing analytics while striving to raise the data savvy of managers in each business unit.

Sharing Data With Partners

Data sharing is critical not only within an organization; it is becoming a key element of negotiations with business partners. Contracts and deals of all kinds are no longer just about who pays what to whom but what data will be shared as well. This sharing is particularly important for businesses that don’t own the ultimate point of sale for their products.

Industrial equipment manufacturer Caterpillar now requires its 189 dealers to enter into data-sharing agreements; in return, it provides them with benchmarks and tools to improve their own sales efficiency and with customer leads generated from Caterpillar’s Web analytics.

Ann Mukherjee, chief marketing officer of Frito-Lay, is able to measure the impact of all kinds of innovative digital marketing for popular brands like Doritos and Lay’s, but this measurement is possible only due to partnerships with key retailers. “Retailers are unbelievable sources of analytical understanding,” and the ability to partner with them around data and measurement is critical to building store traffic and product sales.

As data becomes more essential to business strategy, data sharing will become a key element of every important business partnership with suppliers, distributors, media channels, and more.



Digital Leadership Marketing Strategy
Thought Leadership Zen: Enterprise data management
Enterprise data management
Thought Leadership Zen
Not found any posts VIEW ALL Readmore Reply Cancel reply Delete By Home PAGES POSTS View All RECOMMENDED FOR YOU LABEL ARCHIVE SEARCH ALL POSTS Not found any post match with your request Back Home Sunday Monday Tuesday Wednesday Thursday Friday Saturday Sun Mon Tue Wed Thu Fri Sat January February March April May June July August September October November December Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec just now 1 minute ago $$1$$ minutes ago 1 hour ago $$1$$ hours ago Yesterday $$1$$ days ago $$1$$ weeks ago more than 5 weeks ago Followers Follow THIS CONTENT IS PREMIUM Please share to unlock Copy All Code Select All Code All codes were copied to your clipboard Can not copy the codes / texts, please press [CTRL]+[C] (or CMD+C with Mac) to copy