Churn Model Development, Monitoring, and Adjustment

Churn Model Development, Monitoring, and Adjustment

DOI: 10.4018/978-1-4666-6288-9.ch010
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

This chapter is based on the fact that the finalization of the model building stage is the beginning of the periodic monitoring and redesigning stage. The churn solution should be adopted by market changes, internal company policy changes, portfolio structure changes, and other factors. The chapter gives answers about monitoring frequency and techniques with which the company could realize when to change into the existing churn solution. Another important topic covered in this chapter is “what if” analysis techniques, how to make scenarios for future churn trends regarding planned changes while taking in consideration the current state of the existing portfolio. The chapter ends with business strategy creation based on revealed knowledge from the churn solution and explains the importance of cooperation between business sectors and analysts in all stages of churn solution development from planning and realization to usage.
Chapter Preview
Top

10.1 Churn Model Development

Developed churn solution is a result of current portfolio state. It is unrealistic to expect that developed solution will be usable, predictive or reliable without periodical redesign. Changes in market conditions, implies changes on developed churn solution. To achieve this aim it is prescribed to make periodic validation of existing solution, and in case that it shows poor performance it should be recalibrated or redesigned. Due to different potential changes in the market, competitor’s movement, macroeconomic changes or other factors, developed churn solution could become inappropriate for function for which it was made. To be aware is developed churn solution is still valid or it needs recalibration or redevelopment, it should be periodically validated. There is no rule how often validation should be done, but there is simple rule that turbulent markets needs more frequent checks on developed models than steady markets. This trend became obvious in situation when new competitor enters into market, and churn solution is one of the instruments for having control what is going on into existing portfolio due to amended market condition, where changes could be very often and invisible without deeper data analysis.

Each type of the developed data mining models, which makes final churn solution, could be stricken by market changes. Predictive models could lose their predictive power due to fact that some other type of customers and their behavior became more risky than at the previous period. It simply could be explained with the fact that competitors could target different market segments from company portfolio as a result of changing its market strategy. Changes in competitor’s strategy will not cause loss of predictive power in predictive models immediately; it will be relatively long process. It is important to realize those trends in early stage, which could give an opportunity to company which is threatened with those strategy to make right decisions on time and to calibrate and redesign existing churn models. If model validation is performed very often (e.g. monthly), these trends will not be so explicit, than if model validation is performed less often. Frequent validation performing assures early warning systems recognition. Changing in competitors strategy could also affect on developed segmentation models, prospective customer values and all other developed models, which are integral part of, churn solution. Regarding all that facts it is important to perform periodic validation process, which assures solution, which is applicable for current market situation.

Frequent validation process is welcomed, but problem is that within short time spans there could be too few sample elements for analysis, which will capture new trends into portfolio. One of the technique is using part of period on which model was developed (most recent one) to achieve reliable data sample. This technique has a weakness, because it does not captures real recent sample only, it combines it with part of the sample on which mode was developed. Mitigating circumstance is that it takes most recent on periods and it shows real picture in current portfolio state. As mentioned before, some of the tools which analyst has in performing validation process (mostly for predictive models) are:

  • ROC curve

  • Kolmogorov- Smirnov test

  • Stability index

Complete Chapter List

Search this Book:
Reset