Custom Modeling Maximizes Campaign
By breaking down the modeling process into individual steps and simplifying the process, we can see that developing custom models to drive customer retention or recruitment strategies is not as complex as you might think.
During the past five years, the character and shape of direct marketing has changed. As we accept the paradigm of interactive or one-to-one marketing, we no longer can ignore the importance of relevant and highly detailed data in successful marketing campaigns. However, once you have built your database and have gathered masses of transactional or external data on your customers and prospects, the challenge is leveraging this for maximum effect. How do you use data to predict consumer behavior?
The concept of modeling is simple: Define your objective and model the data - internal or external - to predict a specific behavior, such as propensity to respond to a credit card offer. However, the reality can appear more daunting. This complexity is due, in part, to the prolific generation by the academic world of new and exciting modeling methods to predict and understand behavior. From regression to neural nets, the methods have been evolving as fast as the statistical packages that use these techniques.
Not a conference goes by without a respected speaker encouraging us to "mine your data." The very term data mining has come to represent the panacea of marketing. If only you could mine your data, the answer would appear as if by magic. Millions of dollars have been spent on analysis and software. Though most of it is valuable, do you truly understand what you paid for? Do you really understand the performance of the model you purchased?
The challenge is different today. It is a question not just of using modeling but also of wading through the complexities that surround modeling to ensure you realize tangible commercial benefits from your analytical project.
The solution for modelers is simple: Define the modeling process in an easy-to-understand way, but - more importantly - validate results and encourage client participation and feedback. In effect, break down the cryptic barriers around modeling and open the process so marketers understand how simple it can be.
Let's look at how to create a model to drive net acquisition for a successful financial services invitation to apply, or ITA, campaign. There are five distinct steps:
Enhancement. Enrich your financial services customer data with external demographics, psychographics and lifestyle data. This exploratory data analysis identifies the key drivers for the performance metrics. In this case, we are looking at how best to drive response and approval to an ITA offer, for example, on a credit card.
Segmentation. Identify the segments within the consumer population, pinpointing the characteristics and behavioral trends within the segments. For example, determine the characteristics of your best card customers and your most responsive prospects. By comparing response vs. nonresponse data, you can do this easily.
Scoring. Create the scores and rank targets in each segment from best to worst with respect to optimization of performance metrics.
Testing. Design the appropriate test strategy - reducing financial risk while ensuring the best possible performance.
Evaluation. Conduct constant evaluation of results to facilitate review and refinement of the modeling solution.
Building a model is not easy. It takes skill and experience, but the concept of modeling is simple. Analytical service providers should not complicate the process further by using jargon and complex language that continue to mystify marketers about the process. Straightforward communication and customer participation take the mystique out of modeling and help marketers concentrate on the task at hand - optimizing their marketing campaigns.