Subscribe to DSC Newsletter

Hi, I have a biweekly campaign data with mailers, responses and R,F,M variables. I would build the model for the response rate forecasting for future mail weeks using historical data.

Could any one suggest me what statistical technique i should apply.

Views: 1218

Reply to This

Replies to This Discussion

Hi Veerendra,

This is one problem that the entire Direct Marketing - Marketing in general - face, trying to predict the response rate for the upcoming campaigns using past campaign data. CRM solutions kind of address this issue, but I have never worked on one to know more.

Ideally, if all you're looking for is to predict response rate using historical data, just fitting time series after data treatment is done should work. This is more of a continuous time series data, and you're only trying to predict inside of a huge campaign.
If you've got campaigns in short intervals of time, and you've got multiple campaigns, it gets more complex to predict the "expected value" of response rate given the past details!

In all, it's interesting to know if someone would be able to give alternative solutions, and more input on this. I'd love to hear, since I've faced such requirements too many times.
Also, if you can elaborate on the requirement, it'll help whoever wishes to address you problem.
Thank you Arun for reply to my query.

Recently I gone through some literature and I can do the Regression with auto correlated errors.

Since, in this case i have to consider the other independent variables, i can apply regression on response rate with independent variables and take the residuals and fit ARIMA model to this residuals and combined these two models and predict the response rate.

I don't know whether it is right way of doing or not,Could you tell me if you have any other thoughts.
why do you want to predict reponse rates? what's the benefit of such prediction?
... can't you just extrapolate 2d graph with week on x-axis, and response rate on y? ... in xls sheet
You may not be able to predict much given the countless variables at play: different offers, creative, message, target markets, economic conditions, seasonality, etc. It would be extremely difficult to include all these variables to build a decent predictive model for response rate.
You might be better off just building a sales forecast model.
Thank you Adelino . what you are saying is correct.
I will do sales forecast instead of response rate forecast.
Veerendra: I am happy I could help.


I am curious as to what your final solution was. I am trying to solve for a similar problem and would love some input. I have historical response data for a campaign at day level, but the campaigns vary in size (mail volume). I have tried just a simple linear regression, but the predictions are far apart for some future programs versus close for others. Should I be leveraging the timeseries data in anyway to improve the forecast? Any input is appreciated. 


On Data Science Central

© 2021   TechTarget, Inc.   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service