5
$\begingroup$

RFM - is a ranking model when all customers are ranked according to their purchasing F requency, R recency and M monetary value. This indicator is highly used by marketing departments of various organizations to segment customers into groups according to customer value.

The question is following: are there any substantial models based on RFM scoring (or related to) which have solid predictive power?

Update:

  • predicting which customer will most likely spend more
  • who is going to upgrade/renew subscribtion/refund etc

Update2:

  • I understand, this is simple problem with three independent variable and one classifier. My guess and experience say these pure three factors do not predict future customer value. But they can be used together with another data or can be an additional input into some model.
  • Please share which methodologies worked for you personally and are likely to have high predictive ability. What kind of data you used together with RFM indicators and it worked well?
| improve this question | |
$\endgroup$
  • $\begingroup$ Not sure I understand the question completely, but features based on RFM type calculations are almost always some of the most powerful in a predictive model in the database marketing domain. $\endgroup$ – B_Miner Sep 15 '14 at 16:42
  • 1
    $\begingroup$ Your question is too broad. Prediction for what purpose? RFM models have been around for decades. Pre-internet, every direct marketing organization used RFM for promotional spending -- i.e. who to send catalogs and flyers to. $\endgroup$ – MrMeritology Sep 15 '14 at 17:08
  • $\begingroup$ @MrMeritology, thanks, I have made edits in original questions. I'd like to predict customers tendency to continue buying, spending more or the end of customer life cycle. $\endgroup$ – IharS Sep 15 '14 at 20:06
  • $\begingroup$ what is the dependent variable here and how do we calculate it ? $\endgroup$ – user39524 Sep 21 '17 at 14:42
3
$\begingroup$

A Google search leads to many relevant resources that answer your question:

From a data science point of view, there is nothing very special or unique about this problem. You have three independent variables and one dependent variable. Regression, clustering, and classification methods can be applied.

| improve this answer | |
$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.