By Tim Nichols
The term “big data” is a somewhat all encompassing umbrella for the morass of distributed, loosely connected, and often unstructured data that modern business operations are blessed with. Blessed, that is, in the same way that someone on an episode of A&E’s Hoarders is blessed with a wealth of old newspapers and magazines. Information without useful access and analytics is just clutter.
What you have is big data. What you need is big wisdom. Somewhere buried in the stacks of hard drives that litter your server room, or floating in the cloud of web service and social media sites, is a trove of data about your customers. Data, which if transmuted to wisdom about your customers, would enable you to provide better, quicker, more efficient service, to the delight of both your customers and the company. If only alchemy were that simple.
According to Microsoft’s “Global Enterprise Big Data Trends: 2013” study, customer care is driving 41% of the demand for big data solutions. Clearly there’s an organizational hunger to know more about the customer, and there is recognition that customers are leaving ever larger and more accessible digital footprints. But what big wisdom can all that big data offer, and how can you go about synthesizing it?
What’s In a Name?
This starts with knowing which data is associated with which customer, facilitated by having a complete collection of known customer handles. Handles may include actual names, email addresses, and phone numbers. But they also include Twitter, Pinterest, or YouTube usernames, or other online aliases like Amazon reviewer profile IDs.
It’s long been a best practice to unify all your contact modalities into your CRM system. In this way, customer interactions can proceed seamlessly between phone, email, chat, or even social media channels. Yet equally important, is the linking of other non-transactional activities by this customer, in which your company may be interested. These activities may include social media posts, product reviews, blog entries, or online rants. While it’s nice to know that BigDogBob86 has just declared your new product to be the best thing since Chuck Norris, it’s better to also know that he called twice last month, is married with kids, and lives in Austin.
Most CRM systems support this sort of cross-modal tracking today, but in practice it’s maddeningly difficult to do it well. People have multiple email addresses, multiple phone numbers, multiple Twitter accounts, and multiple internet connected devices. They feel quite free to use them all, and have little patience for your institutional clairvoyance not being up to the task of knowing they contacted you before.
There’s no magic solution to this issue, but it certainly helps if anyone or any system interacting with the customer makes a point of recording the current modality identifier (e.g. phone number, email address, Twitter handle) into the customer record. Customers will generally have little tolerance for or interest in cleaning up their records, so this opens the opportunity for you to have lots of deprecated, or even incorrect, contact information about any given customer. But storage space is cheap, and incorrect, old, or unused handles will just never be searched. In other cases, like a shared home phone number or a Facebook page with multiple administrators, the customer search may return two or three possible customers—still a better situation than having the search return no one. Further, CRM systems that support the addition of a “verified” or “preferred” flag on each contact handle are quite useful. In this way it’s possible to denote a known-good or preferred number or address at which to contact the customer.
There’s also the opportunity to proactively mine for handles used by your customers through social media. Many people reuse handles across multiple sites. After all, the typical goal of social media is to be found, not to stay hidden. If you know one handle, it’s straightforward to test that handle against other sites to see if it results in a valid page, and if so, assume it’s the same user. This can be easily automated and done without user intervention.
This collection of customer handles can not only be used within your analytics and business intelligence tools to look across disparate data sets to synthesize wisdom about specific customers, but also to create views of different demographic slices to which they belong.
Consumer privacy issues are a delicate balancing act. It’s in your company’s interest to know everything about the consumer, but the consumer is often inclined to be a bit less forthcoming. Pitney Bowes conducted a survey in 2012 of consumers in Europe and the U.S. and their attitudes about the collection of personal information. The survey found that consumers are aware of the value of their data, and they also value their privacy. The survey found one-third of consumers are unwilling to trust any type of organization with their personal data. But for the most part, consumers are ready to part with certain types of data so long as they perceive a benefit in doing so.
Dimitri Maex, Managing Director of OgilvyOne and author of the book, Sexy Little Numbers: How to Grow Your Business Using the Data You Already Have, suggests the value equation is different now; “…it is no longer a question of consumers’ willingness to share data but one of establishing a fair value exchange.” Maex’s study showed that 72% of US consumers are willing to share data, as long as they receive something valuable in return.
Millennials have an even stronger willingness to share personal data for rewards. Aimia’s study “Born This Way” notes, “When asked in an open-ended question what value brands needed to provide in exchange for sharing their personal information, Millennials identified reward incentives as the top factor fueling their trust—and in greater numbers than their older counterparts. They also expect brands to secure their personal data and to use it in an ethical manner.”
The value customers receive for their information sharing does not need to be directly monetary. While discounts and coupons are always welcome, value may take the form of enhanced levels of service or even ego-appealing bonuses such as the opportunity to have their comment or review appear on your website.
Clearly, customer privacy requests must be respected, but the chances a customer will opt-out are significantly lessened if there is some perceived value to remaining in. Further, much of the digital footprint data about customers is now public. This means that while customer opt-out requests may dictate how your agents or marketing programs interact with customers during transaction processing, it doesn’t necessarily limit your back-end analytics.
What Wisdom Lies Within?
There are two basic approaches to analyzing big data. One is called rich wandering, and is a relatively novel approach for data analysis made possible because of the integration of disparate data sets enabled by big data analytics tools.
In some ways, rich wandering may be thought of as surfing through your data. Consider the times you’ve been surfing the web, only to find that you stumble onto something particularly interesting; something that affords a valuable insight or provides a new perspective you otherwise never would have thought to go looking for. You arrived at that gem by following a series of links, which from the outside would appear unrelated and uninteresting. That’s rich wandering—the art of serendipitous discovery.
In much the same way the web enabled rich wandering through information; big data analytics enable rich wandering through data.
You might start with a search to constrain an initial data set. You might use faceted classifications to filter certain structured data elements to further constrain your set, in much the same way that EBay or Amazon lets you use facets to narrow your shopping selection. Then you might use graphing tools to create various vectored representations of the set along multiple dimensions. Each peek leads you to think about the data set a bit differently. And that leads you to try different searches, different facet filters, and different graphs. The result being that oftentimes you’ll wind up discovering something valuable you never intended to look for.
This is not to encourage spending hours wandering aimlessly about in the data. Much like web surfing, this can consume copious amounts of time. It’s more akin to the process of discovering things en route to looking up other things. A flexible real-time analytical tool will allow you to wander from your set path to satiate your curiosity—and serendipity happens.
A second analytical approach for big data is the more traditional approach of creating and generating periodic reported metrics and showing metric trends over time. This is not fundamentally different than the analytical techniques you are probably using today, but may require the use of tools specially designed to traverse heterogeneous and less structured data repositories to generate the reports.
What’s new here is the ability to generate metrics and trends that were previously unavailable or too difficult to track.
For example, any state-of-the-art consumer affairs operation is already tracking all the interactions customers have directly with the company. Some are also actively monitoring their company Facebook pages or Twitter feeds for posts that require intervention and handling. But do you know what your customers are saying when they are not speaking directly to you? More importantly, do you know who is listening?
Net Promoter Scores (NPS) have been around for a decade now. The premise of NPS is to gauge a customer’s intent to recommend or advocate for your product by classifying them as a Promoter, Passive, or Detractor based on direct survey results. NPS has often been used as a metric of success of the consumer affairs operation, but some companies also use NPS as a service differentiator. You might go the extra mile to appease a Promoter, where you might not do the same for a Passive customer who is likely to switch brands at the drop of a hat anyway.
While being a Promoter might indicate intent to share, it doesn’t usefully assess a customer’s ability to share. That’s where Social Influence Scores (SIS) come in. SIS is a measure of the ability of a customer to influence their social connections.
One factor of someone’s SIS is their raw number of connections, or their social graph density. That’s relatively easy to measure based on the number of Facebook friends, Twitter followers, or blog subscribers someone has. But graph density doesn’t say anything about the actual influence of that person. It merely indicates the potential to influence.
Wisdom comes in understanding a customer’s social relationship strength, centrality, and prestige. These are more qualitative measures, but can be gleaned by analyzing:
- The frequency of posting original content (as opposed to sharing or retweeting)
- The number of references to a specific company or product
- Geographic or demographic diversity of the social graph
- The post activity in terms of total comments, replies, shares, or retweets
- The post activity in terms of number of unique commenters, repliers, sharers, or retweeters
By combining all these factors, a customer’s true influence can be estimated, and service differentiated accordingly. It may well turn out the squeakiest wheels aren’t worth greasing because, while they make a lot of noise, no one is listening. Alternatively, some comparatively quiet voices might be having an outsized impact on your brand. This is big wisdom worthy of the investment in big data tools for customer care.
Flipping the Equation
The Oracle white paper, “Integrate for Insight,” argues that “big data—information gleaned from nontraditional sources such as blogs, social media, email, sensors, photographs, video footage, etc., and therefore typically unstructured and voluminous—holds the promise of giving enterprises deeper insight into their customers, partners, and business.”
Accessing that insight requires investment in tools as well as the labor to wield them. It also may require adjustments to your customer care processes to facilitate the integration of, and connection between, these far flung data sets—many of which you may not control. But the results can be profound.
Historically, customer care operations have struggled with determining what they can discern about their customers’ behavior from the available data. The advent of big data turns this equation around. The world of big data makes it impractical to start from the point of asking what you can do with the data available. It provides the opportunity for operations to seek answers to questions by asking what data will factor into those answers, and then seeking the connections and tools to acquire that data.
Perhaps more importantly, this breadth and depth of data can provide answers to questions operations may not previously have even thought to ask.
Tim Nichols is a former SOCAP International Board Member and currently serves as Chief Strategy Officer for Wilke Global. You can find Tim on mySOCAP or on Twitter @TimAtWilke