While visiting a banking client recently, I joined a discussion between the customer relationship management (CRM) team and branch managers. One manager described a customer who used several services: credit cards, checking and loans. This customer carefully balanced his checking account and was never late with a payment on his auto loan. His credit card was another matter entirely, with consistently late payments. "It was as if a different person was managing the card," the frustrated manager observed. In fact, the contrast was enough for the bank to investigate potential fraud. They found that the customer was carefully avoiding bouncing checks or repossessions but had a casual disdain for credit card debt. The manager described him as a "Jekyll and Hyde" customer. Other managers had similar tales of customers with highly divergent behavior when dealing with the bank in different ways.

Master data technologies such as customer data integration (CDI) typically focus on the need to identify the checking, loan and card customer as one person. Operational systems should use the same names, dates of birth, Social Security numbers and any other shared data. Yet in this case, the branch manager was also right. In a sense, a different customer was using the card. This credit card customer had different preferences, different habits and posed different risks and advantages to the bank compared to the checking customer. Any business addressing, targeting or classifying him would do well to consider these differences.

The issue at hand, then, is how to establish individual records correctly, as needed for CDI or master data management (MDM), without losing sight of their differing behaviors. I often recommend that master data teams create models with a finer level of detail. The additional facets involve modeling not only customers, but personas of each individual too.

A persona in this sense is an entity related to an individual record that captures a distinctive behavior or personality of that individual as it relates to your business. Customers, in such a model, can indeed have multiple personalities. You will find value in recognizing personas and attending to them in your data models. You can improve your dialog with customers, identify potential churn more precisely and avoid some data quality problems.

Business-Driven and Customer-Driven Personas

Banks and other financial service providers can have diverse interactions with their customers simply because they offer products for varied circumstances. In such cases, taking an aggregate view of a client's behavior is misleading. In the case of our Jekyll and Hyde banking customer, if precampaign analysis for a savings product aggregated his diverse behavior into a single score pulled down by his credit card usage, the bank may have overlooked a potentially profitable savings customer. Just as undesirable, if good savings behavior tips the balance, the bank may offer a credit card product to a customer the bank would not otherwise encourage.

These examples of personas are business driven. Dissimilar products encourage, or at least support, dissimilar behavior; in other cases, different business processes between divisions of a company can all but force unlike behavior on customers. Figure 1 shows these relationships between a customer and bank diagrammatically.

Figure 1: Multiple Personas When Dealing with a Diverse Enterprise

In contrast to financial services, some businesses have uniform connections with their customers. Even here, you may identify personas and find them useful in CRM and CDI initiatives. An online travel agency may be so homogeneous as to have a single portal for bookings. Yet, customers can still exhibit different behaviors. When traveling for work, I may book business class, stay alone in four-star downtown hotels and have inflexible travel dates; when buying personal travel, I may fly coach, stay with my family in motels or resort hotels and be flexible about my arrangements. These examples of personas are customer driven and arise from their differing circumstances, needs and preferences. Figure 2 shows how a single customer - easily identifiable as such, possibly using just one logon ID - may have different behaviors in different circumstances.

Figure 2: A Customer's Multiple Behaviors with a Uniform Enterprise

Other emerging factors may be significant, notably for e-commerce. The researcher Danah Boyd has done fascinating work on understanding "impression management" in social networking.1 The persona you consciously present to the world on a professional networking site, such as LinkedIn, may deliberately differ from that which you present on informal sites, such as MSN Spaces. Moreover, Boyd has found that for teenagers especially, these personas can be ephemeral. They are easily discarded and recreated.

Deliberately presenting multiple personas to the world perhaps sounds sinister. Still, consumers are learning the benefits of adopting personas when dealing with businesses. They know that e-commerce sites customize user experiences. They understand that even the offers and options available reflect their profile as understood by the business. Perceptive consumers are adapting to this market environment. Just as businesses present different offers to different customer profiles, customers, in their turn, may present favored personas to obtain advantages. The consumer customizes his or her own behavior for different vendors. This is not only for the online world. People have caught on to the ploy of asking for a cheeseburger without relish. They may actually like relish; yet, they adopt a relish-hating persona to  force the fast-food outlet to make a fresh burger. Joseph Turow has outlined some social effects of such developments on either side of the commercial fence in his recent book Niche Envy: Marketing Discrimination in the Digital Age.2

Personas, then, can model customer behavior more precisely. Still, it is important to be wary of false precision or overfitting your customer models. In other words, the extra precision of personas will only be useful if better processes result. Consider the case where one persona of a customer is liable to churn, but another is likely to remain a loyal customer. If the churning persona exhibits the most profitable behavior, failing to recognize the distinction can be costly. In fact, personas can help to detect potential churn, where traditional models may not.

In our banking example, a customer may use several financial products and have an overall healthy relationship with the bank. Even so, if he decides that the savings rate is too low or the loan rate too high, he may switch to a product at another bank. Before the switch, it is common to see prechurn behavior: more irregular deposits or slow payments. However, this behavior could only be apparent in one persona; the aggregate picture may suggest a continuing healthy relationship and mask this pattern.

When personas align with products in this way, the costs can be easy to identify. However, in customer-driven scenarios, such as the online travel agency, both the personas and the costs may be difficult to discover. Nevertheless, they can be significant. A colleague represents a good example. Company policy determines that on business she travel with a preferred airline. She travels a lot and appears to be a good customer. However, after some bad experiences, she has stopped using her business carrier for personal travel. The airline has either not noticed or does not care that she no longer travels with them around holidays or to vacation destinations. The airline loses considerable business by overlooking this fact.

You can also gain advantages in data quality by attending to personas. A common difficulty when deduplicating lists is that different people with the same name may share an address. Addressing Robert Smith Junior, the online gamer, rather than Robert Smith Senior, the fly fisherman, is a wasted opportunity. Worse, in many data warehouses, once you discover such a mistake, it is difficult to rekey the deduplicated data correctly. In contrast, if you find that two personas of one customer should really be two customers, it is relatively easy to add a new individual record and key one of the personas to that.

The simplest way to discover personas in action within your customer relationships is to consider if products or processes are likely to attract or encourage divergent behaviors for a single customer. There are numerous examples. The products of auto dealers are not only cars, but also finance, accessories or services. Each of these may attract different customers, but individual customers may also deal with them divergently. In such cases, it may be sufficient to model each point of interaction between a customer and the business as a persona of that customer.

Knowing your business, you may have an intuition about which attributes determine a customer persona. The online travel agency should understand the differences between business and personal use and explicitly ask the customer for the purpose of their trip. The customer self-identifies their persona of the moment. Understanding such differences both empowers the business to deal with consumers more appropriately and also helps to prevent building erroneous models. Knowing that some products are purchased as gifts, a catalog store may avoid alienating an occasional customer with too frequent and unwelcome offers.

If it is neither appropriate nor possible to elicit self-identification of a persona,  you can still model the problem successfully. Data mining models, notably clustering algorithms, are excellent tools for finding those transactions and customers that fall into useful patterns for the business.

Remember the need for a single view of the customer. CDI and other MDM technologies are still vital to consolidate your customer records. Patterns of behavior, such as fraudulent use of cards and loan accounts, require understanding the aggregate behavior of a customer.

Nevertheless, attending to personas can be a great benefit. Personas will be of most value to you when used for CRM and other marketing-related scenarios. Look for opportunities where customer preferences and habits are highly variable, seasonal or where significant changes in personal circumstances can drive changes in the customer's relationship with your business. After all, both Doctor Jekyll and Mister Hyde had money to spend! 


  1. Danah Boyd. "Ephemeral Profiles. (Cuz Losing Passwords is Common Amongst Teens)" apophenia, personal blog. 1 January 2007.
  2. Joseph Turow. Niche Envy: Marketing Discrimination in the Digital Age. Boston: MIT Press, 2006.

Register or login for access to this item and much more

All Information Management content is archived after seven days.

Community members receive:
  • All recent and archived articles
  • Conference offers and updates
  • A full menu of enewsletter options
  • Web seminars, white papers, ebooks

Don't have an account? Register for Free Unlimited Access