I have a data set which has weekly spends of four types of marketing campaigns over the course of three years. I did a few quick EDAs and came to the conclusion that a multiplicative model would be appropriate. So I log transformed the variables and ran a linear regression. Now I want to calculate the ROI for each marketing campaign for all three years. So I went ahead and calculated the total spends for each marketing campaign for all the three years. I kept the spends of other marketing campaigns as zero(to calculate the sales coming from one particular campaign) and predicted the sales after which I calculated the ROI. But now I realize the model has been built to predict when the spend values are in the range of a few thousands and the value I used to predict the sales for three years is much much larger than that(around a million and half) and doing so is actually incorrect. So the question is I have built a multiplicative model at a week level and now I want to calculate the sales coming from a particular campaign for three years. How do I go around to do this?

  • $\begingroup$ This is not really that clear especially to people who do not know what an EDA is or ROI. It is also hard to see what time periods are involved in your predictions. $\endgroup$ – mdewey Mar 15 '17 at 12:18
  • $\begingroup$ Do the decomposition week by week and sum over the three years. $\endgroup$ – salient Mar 16 '17 at 0:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.