question archive There is an excellent case study for assessing marketing campaigns and modality effectiveness in your supplementary text that starts at the top of page 192 with the words "Once upon a time
There is an excellent case study for assessing marketing campaigns and modality effectiveness in your supplementary text that starts at the top of page 192 with the words "Once upon a time
Subject:BusinessPrice: Bought3
There is an excellent case study for assessing marketing campaigns and modality effectiveness in your supplementary text that starts at the top of page 192 with the words "Once upon a time..." Read it and be sure to understand Figure 5-13. Then answer the following questions (over the course of the week, NOT all in one post):
· At the top of page 193, the authors state, "When a marketing campaign includes at least three of these groups, then you can measure effectiveness about the campaign and message." Do you agree with this statement? Why or why not?
· Select one of the groups shown in Figure 5-13, and describe how you could still measure the effectiveness of the campaign and message without using the data for this group.
· Even if the campaign can be evaluated using only three of the four groups, what additional information do you gain if all four groups are included?
Figure 5-13: When you deploy a campaign, four different treatment groups exist. Comparisons between the groups yield different insights.
When a marketing campaign includes at least three of these groups, then you can measure effectiveness of both the campaign and message.
This example also shows the advantage of incremental response modeling, mentioned earlier in this chapter. Chapter 6 , on decision trees, has an example of incremental response modeling in practice.
Step 11: Begin Again
Every data mining project raises more questions than it answers. This is a good thing. It means that new relationships are now visible that were not visible before. The newly discovered relationships suggest new hypotheses to test and the data mining process begins all over again.
Lessons Learned
Directed data mining is searching through historical records to find patterns that explain a particular outcome. The two types of directed data mining models are profiling models and prediction models. These flavors use the same techniques and methodology; they differ only in how the model set is structured.
The solution to a directed data mining problem may involve multiple models that are chained together. So, a cross-sell model may incorporate separate prediction models for each product and a decision rule for choosing the optimal outcome. A response model may be optimized for profitability, so it is really calculating an expected value of the response rather than just the likelihood of response. An even more sophisticated approach is to use incremental response models, where the target is the increase in response rate due to the marketing effort, rather than just the response rate itself.
The primary lesson of this chapter is that data mining is full of traps for the unwary, and following a methodology based on experience can help avoid them. The first hurdle is translating the business problem into a data mining problem. The next challenge is to locate appropriate data that can be transformed into actionable information. After the data has been located, it should be thoroughly explored. The exploration process is likely to reveal problems with the data. It will also help build up the data miner's intuitive understanding of the data. The next step is to create a model set and partition it into training, validation, and test sets.
Data transformations are necessary for two purposes: to fix problems with the data such as missing values and categorical variables that take on too many values, and to bring information to the surface by creating new variables to represent trends and other ratios and combinations. In fact, data transformations are so important that Chapter 19 is devoted to discussing them in greater detail.
After the data has been prepared, building models is a relatively easy process. Each type of model has its own metrics by which it can be assessed, but assessment methods independent of the types of model are also available. Some of the most important of these are the lift and ROC charts, which show how the model has increased the concentration of the desired value of the target variable and the confusion matrix that shows that misclassification error rate for binary response models and the score distribution chart for numeric targets. The next six chapters build on the methodology and dive into directed data mining techniques.
