INTRODUCTION
The tagline for customer analytics has long been the right offer to the right customer at the right time and through the right channel. Marketing investment rides a fine line between two complementary goals: a) creating opportunity by convincing consumers to behave in a certain way; and b) capturing opportunity by being in the right place at the right time when the customer is ready to buy. Customer analytics helps businesses focus their marketing investment on customers who may be persuaded to buy a product they would not have otherwise purchased or customers who are already likely to make the purchase. All too often, however, marketing strategy is based on pre-determined goals of the company rather than filling the needs of the marketplace. In this regard, customer analytics is used to identify customers who are likely to respond to a pre-determined offer, at a time convenient to the business making the offer, through a pre-budgeted channel. The offer, timing and channel are too often determined by the plans of the business rather than the reality of the marketplace and the environment in which the customer behaves.

Companies often limit their analysis to micro-level data and focus on fitting the customer into the box they have drawn. The businesses that stay ahead of the curve examine micro-level data in the context of macro-level data. They understand the macroeconomic trends that create the environment in which customers behave, and bridge the gap between predicting how customers have behaved in the past and how customers will behave in the future. They are proactive in predicting behavior change based on larger social or economic trends (e.g., the popularity of SUVs moving to more fuel-efficient vehicles).

Marketing organizations typically use data mining to develop marketing strategy by focusing on customer-level data, for example customers’ previous product choices, frequency of purchase, age, income, etc. These variables provide insight into the factors that are associated with customers’ past behavior and have some predictive utility in predicting how customers will behave in the future. Macro-level forces, however, have a large impact on how customer behavior changes over time. Variables such as energy prices, mortgage rates, inflation, housing values and unemployment, just to name a few, can add insight into how customer behavior will be different in the present and future than it may have been in the past. These environmental variables can be added to predictive models and segmentation schemes to enhance their predictive utility and make the models more robust. Examples of micro-level and macro-level variables are listed below. 

Customer analytics is, at heart, a comparison of the characteristics associated with different people behaving differently. By selecting the characteristics associated with the desired behavior and not associated with the undesired behavior, marketers can direct their energy and resources towards incenting the behavior they want and winning higher market share. Micro-level variables help to explain why two different people will behave differently in the same environment. Macro-level variables, on the other hand, help explain why the same person may behave differently in different situations. By adding macro-level dimensions to predictive models and the interpretation of predictive models, analysts can better leverage historical data to predict how people will behave in the present or future, based on the combination of micro-level attributes and the current environment.
 

EXAMPLE # 1: THE AUTOMOTIVE INDUSTRY
The automotive industry provides a great example of how individual customer patterns interact with the economic environment to drive behavior. A number of micro-level customer attributes can be combined to segment the population based on different automotive behavioral patterns. Meaningful segments can be identified by examining age, income, marital status, presence and number of children, recency of last purchase, the vehicle purchased most recently, the cost of their most recent vehicle purchase, and other customer-level and vehicle-level attributes. Figure 2 illustrates what four of these segments might look like and examples of their possible differences in vehicle purchasing behavior.

Each of these segments is likely to have different vehicle purchase patterns. They will differ in purchase frequency, vehicle partition choices and the factors that influence the vehicle that they drive home when their purchase is complete. Although the micro-level variables drive the segmentation scheme, the macro-level variables will impact the ultimate behavioral differences between the segments and should help to drive marketing strategy. For example, interest rates can impact the frequency that individuals purchase vehicles. When interest rates were high in the late 1990s, banks and captive finance companies priced vehicles in a way that made auto leasing quite attractive. As interest rates plummeted to historical lows, financing a vehicle purchase became more attractive, particularly to segments that were looking for the best value rather than committed to getting a new car every three years. The marketing organizations that understood various segments’ priorities and understood the impact that interest rate trends would have on consumer choices were able to capture market share from the marketing organizations that did not. 

Another macro-level force that has impacted the way that automotive customers behave is energy prices. The unprecedented spike in gas prices over the last few years altered car buyers’ behavior, placing a premium on fuel efficiency. For example, the typical family vehicles are migrating from SUVs and minivans to more fuel efficient station wagons and crossover vehicles. The entry into the market of hybrid gas/electric automobiles is enabling manufacturers to take advantage of this changing landscape. Similar to adjusting to changes in interest rates, the marketing organizations that push products that are aligned with the changing environment are much better positioned than marketing organizations that orient themselves towards selling for yesterday’s environment.
 

EXAMPLE #2: MORTGAGE LENDING
Mortgage lending provides another industry example of the key interaction of micro and macro-level data. There are volumes of customer-level data that provide predictive insight into a consumer’s behavior. For example the customer’s mortgage balance, interest rate, term, amount of revolving debt and the trajectory of their credit scores all help to predict behavior. However, this insight is most useful in the context of the macro-level variables. Predictive models identifying which mortgage customers are most likely to attrite by refinancing their mortgage with a different lender are likely to include the amount of equity the customer has in their home and the amount of revolving debt they are carrying. Predictive models identifying customers who are likely to respond to a cross-sell home equity offer will have similar attributes. Current interest rates relative to the customer’s interest rate on their current mortgage will play a key role in determining whether refinance or home equity products make the most sense for that customer. The likely length of time a customer will stay in their home (which can be predicted by micro-level variables) will interact with current interest rate trends to indicate whether a fixed-rate or adjustable rate product is most likely to resonate with a customer. As interest rates rise, most customers are better off in home equity products because their current mortgage rate is lower than what they could obtain at the present time. In a downward rate environment, many customers are best-served by an adjustable rate refinance.

Additionally, homeowners’ amount and percent of equity in their home will greatly impact whether and which of these products are the right option. Understanding customers’ equity requires an understanding of housing markets and the likely trajectory of housing prices. Customers in rapidly appreciating housing markets are in a different position to pull cash out of their homes than customers in markets appreciating much slower. Macro-level housing appreciation data can help provide a general understanding of customers’ equity positions without necessarily incurring the data costs of purchasing automated valuation model results for an entire database of customers. Understanding the customers’ individual differences in the context of the broader environment can help marketers make more strategic investments in their customer incentives and marketing programs.
  

STRATEGIC APPLICATION OF MACRO-LEVEL DATA
Macro-level data can function as a lens through which micro-level models can be interpreted and applied. In the automotive industry example above, a micro-level segmentation of auto customers can be applied with particular consideration of the impact that energy prices are likely to have on particular segments’ purchase choices. Higher income segments will likely have a much higher tolerance for high energy prices than value-driven segments. Similarly, particular segments are likely to respond differently to the interest rate environment. In particular, customers who are motivated by value may be attracted to lower lease payments in a higher rate environment and longer term ownership potential in a lower rate environment.

Alternatively, macro-level data can be utilized to create predictor variables. In the mortgage industry example above, the spread between the current prevailing mortgage rates and the interest rate on the customer’s current mortgage is a good predictor of customers’ likelihood to refinance their mortgage. This application can help identify specific spread thresholds at which customers will act. Similarly, the rate of property appreciation in a particular geographic area can be used as a predictor in order to understand the impact of particular real-estate markets on customer behavior.
 

CONCLUSION
These industry examples illustrate the importance of putting customer-level analysis in the context of the broader environment. Data mining is a specialized skill set, and the techniques and methods can be applied without much difference across multiple vertical markets. However, knowledge of the specific vertical market and overall economic trends is essential to ensure that customer analytics include all of the most important measurable factors that impact customer behavior, and that customer-level analysis can be viewed through the appropriate lens and in the proper context. Micro-level variables can be used to understand how different customers will behave differently in the same environment. However, predicting the present and future requires the ability to predict how customers will behave when the environment is different. Understanding the individual through micro-level variables and the environment through macro-level variables enables data mining to bridge the gap between the past and the future.
 

ABOUT THE AUTHOR
Kenneth D. Levin, PhD is Principal of Levin Analytics, a data mining consulting firm serving multiple vertical markets, including retail automotive sales, consumer lending, travel and health care. He can be reached via e-mail at ken.levin@adelphia.net.