With all these powerful contributions to data-driven customer profiles, let me take the next source of big data—product usage. As products become digital, they contribute usage data, which can be analyzed and used for customer modeling. Let me take you through a couple of examples to describe the extent of big data available from product usage and how it can be analyzed and potentially combined with social media and location data.

For a number of decades, television producers relied on a control sample of audience viewing habits to gauge the popularity of their television shows. This data was collected using, first, extensive surveys in the early days of television programming and then, later, special devices placed on a sample of television sets by companies such as Nielsen. With the advancement in the cable set-top box (STB) and the digital network supporting the cable and satellite industries, cable operators can now collect channel-surfing data from all the STBs capable of providing this information. As a result, the size of data collected has grown considerably, providing marketers with finer insights not previously available. This information is valuable because it can be used to correlate channel surfing with a number of micro-segmentation variables.

Television STB data is available at the household level, while mobile device content data provides content viewing by individuals. Consumers are beginning to watch content on both platforms, and sometimes they even use both at the same time for complimentary information access. This data is collected from the devices, cleaned up, and correlated with programming data to ascertain the timing of customer behavior. If I use a two-way STB to watch television, the supplier has instant access to my channel-surfing behavior. Did I change the channel when the advertisement started? Did I turn the volume up or down when the commercial started to play? What if the consumer started to watch the television, but left it on for the day while going to work (maybe while turning off the television, but not the STB). A fair amount of cleanup is needed before this data can be analyzed. STBs are geographically located. If we know the television viewing habits of a community of people, that information can be utilized for beaming specific messages to that community. Aggregation and correlation can be used to analyze STB location data, combined with STB usage data.

As wireless devices get smarter, agents installed on these phones collect device usage information to analyze this data for device or network quality improvement. Sometime ago, my iPhone 4S was showing erratic behavior. It would start heating up all of a sudden and drain its entire battery in a short time, leaving me without a working phone.

I made an appointment with the genius bar at Apple and showed up with the phone. The genius bar representative connected my phone to his laptop and could see my battery temperature history. He concluded that a bad app was causing this erratic behavior and that the best remedy was to remove all the apps, reset the phone, and reinstall a new version of the apps one by one. CarrierIQ offers similar services for Android phones and has been providing device analytics to wireless service providers. By most industry measures, the numbers of smartphones being returned by customers in the first year are running at an unacceptable and unsustainable rate. Perhaps unsurprisingly, the rates appear to vary dramatically between handsets, but appear to be averaging 15 percent-20 percent. What is surprising is that over 40 percent of these devices turn out, upon further investigation, to have nothing wrong with them. This insight was echoed in the recent IWPC Mobile Field Returns Survey (September 2011), in which a selection of US and European operators were asked to report on the volume and type of field returns. Again, no-fault-found ranked in the >30 percent category from most operators.14

Once this data is collected, can it provide any value to sales and marketing? Many consumers live in houses or apartments with poor wireless coverage. We often tell our callers to call back on a landline so that we can reasonably converse without poor wireless quality or dropped calls. Network equipment providers and wireless operators have worked together to provide network devices, which can be attached to a broadband Internet connection to provide wireless signals.15 Since these devices use broadband access, the usage is no longer counted as minutes connected, and the device shows a whopping five-bar network connectivity. How do I identify a micro-segment—miserable subscribers who get poor network coverage near their house? This is a good example of utilizing mobility pattern information to identify residential accommodation for a subscriber and device data to establish poor- quality network coverage. Once joined, the intersection of the two is a target list to market “home connects”!

Regulators have asked most telcos and cable operators to store call detail records and associated usage data. For a 100-million-subscriber telco, the Call Detail Records (CDRs) could easily exceed 5 billion records a day. Telecom operators have been collecting CDRs for a long time. As of 2010, AT&T had 193 trillion CDRs in its database. As phones became more sophisticated and consumers started to use the phone for activities other than calling, the CDRs started to include other forms of communications, and everyone started to use the term xDR, where x is a variable that takes many meanings, depending on whether it represents calls, text, data, or video. Big analytics has lately started to provide telcos with sophisticated capabilities to analyze this data and find useful nuggets of information about their customers.

Social groups can be inferred from any type of communication- emails, SMS texts, calls, Facebook friendships, and so on. It is interesting to see strong statistics associated with leaders’ influence on their social groups. One such analysis involves discovering group dynamics. Communication across individuals can give insights into formal and informal groups. In some situations, these groups have formed among coworkers, and could be very formal with organizational hierarchies and matrices. In other situations, the communications may be due to informal social groups formed via families and friends. In any group, there are leaders who keep these groups together and followers who are influenced by those leaders. There may be ambassadors who may belong to one group but represent them in another, where they have loose ties. There are many ways to discover these groups by using big data. Telcos are rapidly discovering that they have a gold mine of big data in the form of xDRs with social/work group information, which can be used for marketing purposes.

Social group leaders typically have a set of social group followers. If these groups are communicating with each other, it is a possibility that the brand choices made by the leaders will influence the subsequent brand switching among the followers. Let me use an example to illustrate this behavior. I am a member of an investment club, which regularly meets to discuss investment decisions. We have pooled together a small fund, and we make decisions about buying and selling stocks using the pool of funds. In addition, each of us has our individual investments. Our group decisions often influence our individual decisions, especially in dealing with the investment brokers and tools. Collectively, we use five brokers for individual investments across the group. As the group leaders start switching from one set of brokers to another, others start to follow. The group communicates regularly with each other using cell phones, with a significant number of calls and texts. Can we analyze the xDR data to predict brand switching for investment brokers? Once a leader switches a brand, it increases the likelihood for the social group members to churn as well. Who are these leaders? Can we identify them? How can we direct our marketing to these leaders?

In any communication, the leaders are always the center of the hub (see figure 3.1). They are often connected to a larger number of “followers" some of whom could also be leaders. In the figure, the leaders have many more communication arrows either originating from or terminating at them compared with the others.

Leaders in a communications network

Figure 3.1 Leaders in a communications network

How do we identify the leaders? IBM Research conducted a series of experiments with telcos. CDRs, which carry information about person A calling person B, were analyzed. By synthesizing call information and abstracting communications networks, we discovered webs of communications across individuals. We also used the customer churn information to correlate churn among leaders to subsequent churn among followers. Here are some of the highlights from one of the experiments that I helped conduct:16

  • • Leaders were 1.2 times more likely to churn compared with nonleaders.
  • • There were two types of leaders: disseminating leaders who were connected to their group through outgoing calls, and authority leaders who were connected through a larger proportion of incoming calls.
  • • When a disseminating leader churned, additional churns were 28.5 times more likely. When an authority leader churned, additional churns were 19.9 times more likely.
  • • Typically, there was a very limited time between leaders’ churn and the followers’ churn.

Social group data is increasingly available from a variety of sources. This data is becoming an important source of data for the social media products, in addition to the telcos. Each of these sources represents different types of groups. The overall social dynamics of a group is a combination of observations across a number of sources. A set of Facebook friends may not be communicating via phone, but could be actively sharing product evaluations using social media sites.

There are many ways to utilize social networks to influence purchase and reuse:

  • Studying consumer experience—A fair amount of this data is unstructured. By analyzing the text for sentiments, intensity, readership, related blogs, referrals, and other information, we can organize the data into positive and negative influences and their impact on the customer base.
  • Organizing customer experience—We can provide reviews to a prospective buyer, so he/she can gauge how others evaluated the product.
  • Influencing social networks—We can provide marketing material, product changes, company directions, and celebrity endorsements to social networks so that social media may influence and enhance the buzz.
  • Feedback to products, operations, or marketing—By using information generated by social media, we can rapidly make changes in the product mix and marketing to improve the offering to customers.
< Prev   CONTENTS   Source   Next >