iSixSigma

Marketing Sample Size DOE

Six Sigma – iSixSigma Forums Old Forums General Marketing Sample Size DOE

Viewing 6 posts - 1 through 6 (of 6 total)
  • Author
    Posts
  • #32109

    Lomax
    Participant

    I have two statistical questions concerning sample size and DOE for an insurance marketing firm.  The company produces a direct mail marketing package to sell an insurance product. The key measure is response rate, that is, the number of people who return the mailer requesting to purchase the insurance product divided by the total number of mailers sent out. 
    1.  For discussion purposes, let’s say a bank’s customers are the target “population.”  The customers are ranked in deciles according to a model that predicts which customer’s are more likely to respond positively to the particular insurance product.  Typically, the top two or three deciles receive the mailer to minimize printing costs, etc.  As a sanity check, a small “random” sample from all ten deciles receive the mailer to validate the model. This validation is not used to measure the response rate statistics though.  My question is this:  Does the “targeting” of  the bank’s customers nullify any statistical analysis on the response rate because the “sample” is not random?
    2.  The company would like to conduct some DOEs by changing different levels of factors in the direct mailer.  Factors might include insurance amounts, free gifts, etc.  Obviously, the only data gathered from the DOE will be from those mailers that are returned which is typically 1.5%.  In terms of a DOE sample size, if I use Minitab’s DOE sample size calculation, I will need to divide the calculated sample size by 0.015 to get the number of mailers we need to send out assuming a 1.5% response rate…correct?  Let’s say when I do this, I get 16,000 for the sample size.  For a 3 factor, 2 level full factorial DOE, will I then evenly spread the 16,000 across all eight treatments? That is, I will have 2,000 mailers for each treatment combination?
    I’ve tried to explain this as best as I can…if you need more info to help me answer the questions, please let me know.  Thanks in advance for any guidance on these issues.

    0
    #85379

    Pam Hunter
    Participant

    You can calculate 2 response rates, one response rate from the random sample of the 10 deciles and one response rate from the first two deciles.  The response rate from the 2 deciles cannot or should not be generalized to all bank customers, only the customers who meet your criteria for being most likely to return the mailer. 
     
     
     

    0
    #85394

    Lomax
    Participant

    Thank you Pam!

    0
    #85432

    Dr Antony
    Participant

    Neil,
    Very interesting application of DOE in marketing. Why don’t you carry out DOE for a particular cluster of customers in the first place?? Can you define and formulate the nature of the problem at hand and Why do you think ” an experimentation is the solution ” for this problem? Please also tell me th eobjectives of your experiment and what you want to get out of the experiment. I have done a lot of work on DOE, Taguchi and Shainin’s approach. Moreover, the application of DOE in marketing comes under a different name called ” Conjoint Analysis”. Anyway, you need to understand the reasons for experimentation before you design the first experiment.
     
    Dr Antony

    0
    #85434

    Adriano Barbosa
    Participant

    Dear Dr. Antony.
    Thank you for the remembering. I was just trying to rememnber the name for it in Marketing application.
    Adriano Barbosa, Brazil.
    [email protected]

    0
    #86129

    ali
    Participant

    hi
    i  am conufused about what to choose in a six sigma project relating to marketing . if any one have some suggestions kindly mail me before ,
    23 may ,2003 thanks
    ali

    0
Viewing 6 posts - 1 through 6 (of 6 total)

The forum ‘General’ is closed to new topics and replies.