Modeling
Choice Models and Other Inferential Techniques
Inferential modeling is used for various marketing applications – for instance, to identify the drivers of purchase behavior, customer satisfaction and brand equity and to power customer/prospect database scoring algorithms that assist marketers in predicting the interests and potential behaviors of specific customers or prospects. Various forms of predictive models, including discriminant analysis, regression analysis, structural equation models and derived importance models are frequently used to isolate the key drivers of customer purchase decisions as well as the critical factors that influence brand equity and customer satisfaction. Techniques like factor and correspondence analysis are used to discern similarities and distinctions among product attributes, and to identify the characteristics that are most closely associated with different brands and products. Techniques like CART, CHAID, and Bayesian CART are often used to create database scoring algorithms with which to predict customer orientations, interests, and motivations, and to identify "top prospects" for a particular product or service.
Drivers of Purchase Behavior, Customer Satisfaction, and Brand Equity
Customers are not always able to tell us directly what motivates their behavior or engages their allegiance, and marketers must resort to inferential statistics for depth of insight. A variety of techniques can be used to uncover and isolate the key drivers that influence purchasing, customer satisfaction, and brand equity – sometimes with disappointing results that fail to "lift the veil" or may even mislead. Choosing the right modeling approach and avoiding model misinterpretation requires an intricate knowledge of the market, as well as a thorough familiarity with the strengths and limitations of different analytic approaches available. We make deft and careful use of all the relevant tools, routinely exploring multiple approaches to maximize the clarity of inferences that are drawn. The impact can be enhanced with user-friendly Proprietary Software that enable clients to identify the implications of potential product or service improvements and to more effectively manage marketing resource allocation.
Customer/Prospect Database Scoring
National Analysts Worldwide creates inferential database scoring models (for use in targeted marketing) using a combination of primary research, customer data, and secondary sources. After models have been built, we will score existing prospect/customer databases, or perhaps an integrated database, with predicted values from these models.
Examples of assigned scores may include:
- Predicted customer segment
- Purchase propensities for individual products/services
- Likelihood to respond to promotional offers
- Attrition propensity
- Likelihood-to-switch/upgrade
- Total wallet (value) potential
These types of modeling applications are advantageous because they provide information that is not available in raw customer or prospect databases or from secondary sources. Model-based inferences allow clients to extend the value of detailed custom research, and to inform and execute strategies for each individual customer/prospect in their databases.
The model development process typically includes an evaluation of multiple alternative methods; all modeling methods are tailored to fit the nature of the available data. In particular, models make proper use of sample survey research data which might require sophisticated sampling and analysis strategies. These efforts not only ensure that model predictions are as accurate as possible, but also ensure that benchmark measures of predictive performance that are calculated using the survey data are seen in the overall dataset.
These models and associated insights are typically delivered to clients in one of two ways: (1) National Analysts Worldwide can score client databases, assigning predictive "scores" to each individual customer and prospect record; or (2) National Analysts Worldwide can deliver detailed scoring and assignment models (NA-Link®, NA-QuikLink®), that clients are able to integrate into their existing data management systems. They can also be used either in a stand-alone manner or in conjunction with data mining and customer relationship management (CRM) applications, to help clients efficiently locate the customers and prospects that are most likely to respond to specific offers.
For more information on National Analysts Worldwide, please e-mail us or call (215) 496-6800.