Lifetime Value Models?Has anyone used JMP to develop Lifetime Value Models? I have looked online & don't see much on the JMP side, only SAS.
Re: Lifetime Value Models?In the past when I was asked about CLV I told them that JMP has the statistical capabilities of building a CLV model (e.g., standard least squares for predicting values, survival models for tenure and logistic regression for acquisition), but JMP doesn’t have a button solution to that...
As Malcolm and Martin point out we need to ask the customer carefully about what kind of data they have gathered, how they define a customer value and what variables they consider to use in these models to predict different components that calculates CLV. It is not a small task to develop CLV and deploy them; however if the customer is a trained modeler, and has the “right” data and would like to consider JMP to replace SPSS or STATA or enhance their arsenal, then we have a shot at.
A good read but it requires a fee to download “Modeling Customer Lifetime Value” http://journals.sagepub.com/doi/pdf/10.1177/1094670506293810
Re: Lifetime Value Models?I agree on what Malcolm said. just few more things to consider:
I had few conversations in the past with departments in existing customers with similar requests. It is not that easy to get in there, as many would like not to create the model on themselves, but get it right away fom the data in a visual BI dashboard. If there is someone who would create that for them, it's a nice way of spreading JMP into other departments. Otherwise I've heard the BI dashboard argument knocked out JMP.
Regarding the example, the model has not changed, but some use additional variables, which they add to the model with weighting.
CLV need to be treated carefully, as it has often high prediction variance. On the one hand it's hard to precisely estimate the duration of the customer relationship, on the other hand it is even more difficult to have relevant In/Output streaming data which could act as a solid base for prediction. Therefore most of the time one use values of experience. Young companies without that infomation use common rules of thump values. So the CLV model theoretically has good properties but is difficult to be used practically unless there is good transactional related data (buying history).
More information you need to take care of you'll find in the wikipedia link: https://en.wikipedia.org/wiki/Customer_lifetime_value
Re: Lifetime Value Models?Here is an example I created 10 years ago when exploring opportunities for selling JMP into financial services. This wasn't very successful as it is outside of our core market of science and engineering:
1) Run the survival script in the table Lifetime Value Analysis(2)
2) Right mouse click over the Estimates of Survival outline in the survival report and make into data table. this gives the table called survival probabilities after deleting a few columns that are not needed.
3) Join the resulting survival probabilities table with monthly margin to create the table called expected monthly margin and derive the final column in this table by multiplying the probability by margin.
4) Run the script called Lifetime Value (36 months) in the resulting table to get the tabulation summary of lifetime value by market segment.
This was a fairly standard process for calculating lifetime value 10 years ago and an internet search may help you determine if this is still the preferred method.
If this is a new sales opportunity, I would proceed cautiously. It is outside of our core market and could be a big draw of your time with a low probability of pay-back.
Best,
Malcolm
