The larger the island of knowledge, the longer the shoreline of wonder.
—Ralph W. Sockman
Introduction
Now that we’ve had an introduction to the elements of Dynamic Customer Strategy (DCS), the need for data is apparent. Not all data are created equal, however. To implement DCS, you must:
• Recognize data traps that can lie within Big Data
• Develop a data strategy.
Data and DCS
DCS requires data. If you want to test your theory of how something works, you need enough data to know. Newton may have figured out gravity after watching one apple fall, but chances are pretty good he had a few data points before that last one hit the ground.
Today’s organization has more data than ever before. People have been talking a lot about Big Data and the challenges of managing Big Data, but more data is, by itself, just simply more data not more knowledge.
More than a few years ago, a washing machine manufacturer in India noticed that buyers were ordering eight, 10 or even 20 washing machines but no driers. Who needs a lot of washing machines? Laundromats, prisons, university dormitories, hotels, perhaps, but don’t they need driers too? Thinking these buyers were laundromats, the company sent a sales representative to one particularly large customer to offer coin-operated attachments and heavier-duty machines, as well as driers. Imagine the representative’s surprise when it was learned these customers were using the machines to make cheese! Yes, cheese! The agitation of the barrels and the top-loaded washing machines were perfect for making a kind of cottage cheese popular in one region of India.
Data is not knowledge. Information is not wisdom. You may think customers are washing clothes when actually, they are making cheese.
Data and Data Traps
Companies need to answer the following questions: How well do you know your best customers? For that matter, how well do you know any of your customers? These are not idle questions, nor are these all the questions that need to be answered but this is a good start.
Most companies, like the washing machine manufacturer, divide their customers into groups based on some measure of value. Depending on the maturity level of the Customer Relationship Management (CRM) program and the experience of those leading it, that value might be gross sales, an index based on profitability estimates, or some other metric. But like the washing machine manufacturer’s, this approach is a CRM data trap.
A data trap is the illusion of knowing because you have good data—a lot of good data. Decision makers think they know a lot about their customer because they have good data on their customer, and this warm, fuzzy feeling of security wraps around their decision making like a baby’s blanket. The problem is that the data provide only a limited picture; the rest is filled in with assumptions. “Oh, they must be laundromats.” “Oh, all our biggest customers use our product the same way and they all love us.” When a company falls prey to a data trap, it makes decisions that seem appropriate for the data that they can get, filling in the missing pieces with assumptions.
Transactional data such as the size of an order creates a data trap because today’s enterprise data warehousing technology and Big Data make transactional data easier to access and analyze. Transactional data also adds a layer of information that was previously unavailable, giving marketing professionals more to work with than they ever thought possible. The trap, though, is that so much data from only one source creates an illusion of knowing, rather than actual wisdom.
One fault lies in an assumption that all customers are alike if they purchase the same amount. For example, some companies use Recency, Frequency, and Monetary (RFM) scoring as a proxy for customer lifetime value. RFM Value is a score based on how recently they purchased, how often they purchase, and how big is their average purchase. Like the washing machine company assuming that all multiple machine purchases were intended for laundromats, many companies assume that the drivers for high RFM scores (or heavy use or high volume purchasing) are the same across all high volume purchasers. We say we want one to one marketing, but what we are really saying is we want to talk directly to customers one at a time. Yet, what’s difficult to remember, and to operationalize, is that each customer is different, to some degree, and we need to find ways to capture and act on that difference.
Traditional Data Trap
Transactional data is not the only data that creates a data trap. For example, what type of householder would cook all meals in microwaves, owning top-of-the-line units with browning elements? Or have two dishwashers, one for dirty and one for clean dishes? Most people think such a person is likely to be a young professional or a working mother, someone without a lot of time. The first kitchen I saw like that, though, was created by my 80-plus-year-old grandmother—not to save time, but because it was easier on her to not have to unload a dishwasher or slave over a hot stove. But most marketers would assume incorrectly that, because of her age, she was a technophobe rather than an innovator.
Demographic data can also create a data trap. CRM professionals fall into a transactional data trap; traditional marketers fall into a demographic data trap. Traditional marketers tend to treat all members of a demographic group as being the same, whereas many CRM professionals treat all members of a RFM decile (that is, a band of 10% of customers based on RFM score) the same. Victims of either trap are not practicing relationship marketing. What they are doing is using the knowledge they have available to create an illusion of knowing, an illusion that makes them feel like they know their market when, in fact, the actions they take are based on incomplete pictures.
Think of it this way. If you reviewed a million pictures of an elephant’s butt, you’d still not have any indication that there was a trunk on the other end. If you have a million transactions to review, you can’t tell who’s washing clothes and who’s making cheese.
Avoiding the Data Trap
Is transactional data not necessary? Does all demographic data lead to the illusion of knowing? Of course not—both transactional and demographic data are needed. Transactional data can help you identify event-based opportunities for dialogue or sales, aid in determining the value of a particular customer or provide insight for other activities. With demographic data, offers can be couched in the right language, and other decisions are supported. But these two forms of data are only a part of the total picture.
The question then becomes how to avoid the data trap—or rather, what is the right blend of transactional, psycho-demographic, and motivational data needed to truly understand the customer? To know the answer to that is to develop a data strategy.
Developing a Data Strategy
A data strategy has four steps:
1. Acquire—find the right data based on the decision to be made or other business need
2. Analyze—develop the right model to inform the decision maker
3. Apply—use the model in making the decision, implementing a strategy, or executing a campaign
4. Assess—review the results to determine if the data and model were worthwhile.
There are two business needs for data. The first is to make a decision now, whereas the second is to simply understand. Clearly, the need to understand will lead to decisions, but I separate the two because understanding is what supports the development of the types of theories that serve the creation of strategy.
First, let’s talk about the types of data you need to collect. Big Data is not just more of the same—it’s new sources of data, too, many of which were not set up to be data. There’s behavioral data, such as web browsing activity; transactional data, such as what was purchased and in what sequence (and academically-speaking, a form of behavioral data); psycho-demographic data which helps us understand lifestyles, age and gender effects, and the like and are heavily used in understanding and promoting brands; and motivational data, or the types of data that help us understand how our buyers consume or use our products. Motivational and psycho-demographic data are very much related; motivational data is specific to the product and use situation but driven by psycho-demographics. Add in descriptive data like blog content, Facebook post content, and the like, and you can see how Big Data is comprised of so many sources of data.
You already know that you need more than just transactional data or you fall into a data trap like the appliance company. Let’s take a look at more examples.
Overstock.com’s Chief Executive Officer, Patrick Byrne, believes in the value of transactional data because it is behavioral data. He says that you can use that behavioral history to transform future behavior, and he’s right. For example, if you can’t get some buyers to come back and make a purchase within a certain window, they are lost forever. For Overstock.com, that window is 45 days after their first purchase. At first, Byrne suggested you make a discounted offer to them on the 44th day. The trouble is that not everyone will reply because they weren’t there for the discount to begin with, so you aren’t offering a relevant message. Another result could be that you are training discount-oriented buyers to wait 44 days, or it could be that you only get price-sensitive buyers to return, losing the higher value customers willing to pay full price.
Every company trains its customers, sometimes unintentionally. For example, Penney trained their buyers to never buy anything at full price by having so many items on sale so often. Customers knew that at least twice a month, many products were drastically reduced, so they waited for the markdowns. In late 2011, JC Penney hired former Apple exec Ron Johnson to turn things around. His first strategy was to wean buyers off of discounts and coupons, trying to give shoppers other reasons to visit Penney. Unfortunately, he was unable to retrain Penney customers and his tenure was relatively short-lived (about nine months). JC Penney is a cautionary tale for those who use discounts as the first marketing weapon of choice.
Assume Overstock.com had two types of buyers—those who were hard-core discount buyers and those who bought because just because shopping at Overstock was fun. How would this knowledge help them? One way it would help is less reliance on deep discounting. Under Byrne’s original thinking, Overstock.com would not necessarily know which type of new customer they were dealing with until the 44th day and a discount was offered. Either the buyer responded to the discount and bought or the discount was ignored. If the buyer does purchase with the discount, perhaps it is a deal-seeking buyer whereas the “fun-seeking” buyer was lost. Only the low-margin customers were retained.
One alternative would be an offer of some fun items or a promotion that appeals to fun-seekers on the 34th day. Ten days later, all non-responders get the discount offer. Now Overstock.com would know who the fun-seekers are and who the hard-core discount buyers are (see Figure 3.1). Future offers could then be tailored to fit the drivers of their behavior. Further, if a large group did not reply to either offer, then perhaps a third group of buyers is appearing. Overstock.com needs to do something to identify what motivates those buyers—and, therefore, needs more data.
Figure 3.1. Approaches for retaining customers.
Think back to Chapter 2—you can see that you have two competing theories operating here for why buyers buy. One theory is that buyers shop at Overstock for the bargains; another theory is that they shop there to find unique things because it is fun. The truth is that both could be right, just not right for the same buyers or for the same product categories or for the same situations.
What Data?
To determine what drives or motivates a particular segment, two key areas of data are also needed in addition to transactional and demographic data. These additional types of data are motivational data and lifestyle data. These two types of data are intertwined at the acquisition, the analysis and the application stages. In some situations, motivational or lifestyle data may fall into the “nice to know” category, but consider my grandmother once more. Why hasn’t anyone sold a side-by-side dual dishwasher that accomplishes the same thing? Would GE or another company create a market if they understood the people living lifestyles that would benefit from such a product?
Basic CRM promises, such as making an offer individualized to a customer, cannot be fulfilled without motivational and lifestyle data. Even some of the most basic CRM foundations, such as determining your customer’s value, are suspect with only transactional or demographic data.
For example, assume you own a fashion retailing firm targeting young women. Reaching them through catalogs, stores, emails, and websites, you have that omnichannel approach down. One question haunts you, though: What is the “life” of your customer? If she is 20 years old, is her customer life another 3 years, 5 years, or 80 years? And can she then be moved into another customer category reached through another division of your company?
Rather than lifetime customer value, perhaps it is better to think about defined customer value. Defined customer value is the value of a customer for a defined product category for a defined period of time. And these definitions also require motivational and lifestyle data.
Earlier, I said that a data trap of any kind is any situation in which you have a lot of data, and it is the volume of data that gives you the illusion of knowing. Some have argued, for example, that the sum of a customer’s total transactions over time, multiplied by some expectation of life span, provides a good picture of that person’s lifetime value.
Assuming no major changes in the market or innovations alter the relative value, and further assuming the buyer only buys from one company, that could be an acceptable level of knowledge. But it ignores the share of wallet, or your share of purchases that the buyer makes out of the budget for that type of purchase. Share of wallet seems easy to calculate, but the key is how big that wallet is.
To really have an idea of a customer’s potential value, you have to know what the total purchases are in that category of the buyer’s budget. Soft drinks, water, sports energy drinks, and beer are all beverages, but the buyer may not consider those equally as substitutes. Although I’ve played soccer with guys who think beer is a sports drink, most of us would not consider beer an alternative to PowerAde.
That means understanding when the buyer considers your product, and against which competitive products is it important to calculate your share of wallet. How the customer calculates the wallet is more important—you may think you compete against energy drinks, for example, but if the customer doesn’t consider your beverage when quenching a need for energy, then energy drinks aren’t your direct competition.
Here’s another aspect to consider: time. How much time does your buyer have, not only for enjoying your product but also enjoying your shopping experience? For example, a topic in golf retailing right now is whether 18 holes divided into two halves of 9 is the right number, or whether it makes sense to have 3 portions of 6 holes so that people with less time can get in a quick game.
The buyer who buys a sport drink to quench a need for energy is motivated by that need. Motivational data, then, is knowledge that identifies what drives a buyer to make a purchase. To understand wallet size, you have to understand how the buyer sees a purchase, and the buyer sees a purchase based on motivation.
Motivational knowledge is important for both B2C and B2B. In the research I’ve conducted over the past two decades on how organizations make buying decisions, I’ve learned that buyers have personal needs as well as organizational needs, and sometimes these personal needs are dominant. To an organizational buyer, the situation may be about showing off decision-making skills or meeting a profit target by cutting expenses, not about choosing benefits from two wonderful products. To be sure, budgets are on paper and more formal. At the same time, there is discretion within those budgets and it helps to understand the motivation underlying customers’ choices.
One challenge is determining the size of the wallet; the other is determining life. Back to the young women who shop at the fashion retailer. There is one event that changes forever how they shop–college graduation. Identify that, and you’ve identified a major change in their shopping habits. Yes, they still buy jeans, but not as many, and not always at the same price since they are now on their own. Nor is the rest of their fashion shopping the same. These shopping changes are driven by their lifestyle changes.
The Opposite Problem
A data trap requires a lot of data to exist. But what if your challenge is just the opposite—a lack of data? Perhaps your organization has a small customer list or perhaps no one kept the data. Does that mean that this DCS approach won’t work for you? What do you do?
First, don’t put the book down or give it away just yet! The principles of a sound data strategy still apply. In some respects, the data strategy is a bit easier when your universe of customers is smaller. How you go about getting data and storing it will be different as a function of scale, or size. The principles, however, still apply.
Where’s the Data?
Part of the allure of transactional data is that it is already there in the enterprise. Motivational and lifestyle data are not; you have to go get them. And that means research (or paying someone for it).
There are two ways to get data: ask or observe. For example, motivational data can be gathered using a sample of several hundred customers. Using Overstock.com again as an example, you can start with focus groups, asking 10 buyers at a time, “What is the biggest benefit you get when you shop at Overstock.com?” If some say, “It’s fun,” then you know that there is the possibility that there is a fun-seeking group of buyers. Follow this up with a survey of several hundred more customers. Ask additional questions so that you can understand what they mean by “fun.” (Note: focus groups are particularly useful if your population of customers is small.) Alternatively, you could analyze the text of tweets, blogs, and posts and recognize a fun segment is out there.
Similarly, if you have sales people and other front-line customer employees, they probably already have a pretty good idea of what the categories of buyers are. Don’t have a lot of money or time to pursue a big marketing research project? Start with the sales force. Simply by asking salespeople what they think, you can identify groups of buyers to start your marketing strategies.
How valuable are these buyers? Hard to say at this point, but that is something you can easily determine with transactional and behavioral data. As suggested earlier, on the 34th day after someone has made a new purchase and using the additional information about what constitutes a “fun” shopping experience on the website, you create and make an offer that should appeal to the fun shopper. You don’t have to ask that shopper if he or she is fun-driven—the shopper will tell you by responding either with a purchase or a “delete message.” Now you’ve gathered data and probably made a profit at the same time.
A similar way to ask and receive is through progressive profiling. Progressive profiling is the process of asking questions to learn more about your customer but asking these questions over time. For example, a customer signs up for a newsletter—ask for interests. Then the customer calls the call center to order a product—ask for household composition (“Oh, do you have children? That’s a wonderful product for your family to enjoy…”). Hold an event to introduce a new product line—ask for usage information (“How often do you use our product when entertaining?”).
As for determining the relative value of the fun shopper versus the hard-core discounter, time will tell you who is more valuable. As you build data on these two groups through both observed behavior and progressive profiling, you’ll be able to determine the defined value of each as it stands—given how you currently market.
You have an impact on that value. When discounts are offered willy-nilly, the value of the customer base is lower. When offers are appropriately targeted, value curves shift, meaning that some customers increase in value because they respond, whereas others may decline in relative value because they respond to lower-value offers.
Further, acquisition strategies may then change. As relative value of each group becomes known, motivational profiles should be continuously tested. In other words, the motivational (and lifestyle) profile of the group is continuously fleshed out by testing offers and responses, which can then be used to more effectively acquire similar customers.
Case Study: Gallery Furniture
If you live in Houston, you know Mattress Mack and that he will “absolutely save you money!” He’s justifiably a legend in Houston, not just because of his crazy commercials over the past 30 years but also because he’s civic-minded and believes in sharing the wealth of his success.
Mack (nee’ James MacEngvale) realized, though, that furniture retailing is changing, just like all consumer purchasing. Showrooming, or the practice by consumers of going to a retailer to touch and feel a product before ordering it cheaper online, was beginning to start to become a problem even in furniture. Mack fought it with “Get it TODAY!” delivery, with aggressive pricing, and with clever inventory management and sourcing that makes showrooming next to impossible. If Mack is the only one with a particular product or color, showrooming won’t work so he also began to strengthen his custom product line.
What happened wasn’t that Mack just defended his market share. He actually grew his sales and his profits.
Even so, Mack wasn’t content. He challenged long-held assumptions in retailing, one of which was that customers who leave won’t come back to buy. With a strong follow-up program, he’s learned just how wrong that assumption is. I can’t share all of the details on that program, but what he will let me tell you is this:
• A furniture customer today is likely to change over one full room, and possibly more, in the next 18 months—even if all that was purchased was a wing chair or a mattress.
• Some types of buyers will change over their public areas every three years.
• Find those buyers, capture their hearts, and you have first shot at all of their business.
Now you can draw those first two conclusions from transactional data but if you stop there, you’ll fall into a data trap. You’ve got to add other data, as well as draw and test inferences, to win their affection.
Starting simple, Mack began with a newsletter. The newsletter contained stories that appealed to certain segments. By simply tracking who read what, a clearer picture of who the reader is and what that person is interested in began to appear.
Further, depending on the segment, there are certain events each year that can trigger purchases. Back to school might mean Junior needs a desk, but it might also mean Dad wants a big screen for football (and if he’s a Houston Texans fan, custom leather furniture with the Texans logo—can’t get that anywhere but through Mack). Yes, you could advertise in the Houston Chronicle, but how much more powerful is it to send those offers directly to the households that might be interested? And, make them a low-cost high value offer (like meet Arian Foster, star running back of the Texans, who is at Gallery Furniture every Tuesday for a radio show during the season) that they can’t get anywhere else.
Sounds like just good retailing. But there’s a data strategy behind it that includes continuous improvement because success can be tracked back to the offer. That’s one problem with advertising—as one famous advertiser said, “I’m sure half of my advertising works. I just don’t know which half.” With this approach of marketing directly, using data, new data are captured and better offers crafted and presented.
The Rest of a Data Strategy
Data are the foundation of a Dynamic Customer Strategy. Good data is required to understand how variables relate in driving customer behavior, and getting good data means having a sound data strategy. Acquisition of data can include using transactional datasets from your financial data, but it can also include “asking” customers by giving them an offer and seeing if they respond or by asking questions over time.
Summary
Creating a data strategy involves more than simply identifying what data you want or cataloging the data you have. Many decision makers fall prey to data traps but with Big Data, the most common is thinking that a lot of data equates to knowledge. A good data strategy includes acquiring the right data for the decisions to be made, analyzing appropriately, applying the knowledge operationally, and assessing the data and the process for value.
Discussion Questions
1. How can Big Data increase the likelihood of being trapped by data? What’s the solution?
2. Are you guilty of showrooming? If so, describe the situation. How close does the in-store price have to be to the online price to get you to buy right then and there? What long term effects do you think showrooming will have on how products are bought?
3. Explain how the quote at the start of the chapter relates to the chapter’s topics, as well as how it relates in general to DCS.