Replies to This Discussion

Permalink Reply by Idielle Walters on November 3, 2010 at 1:04am

Hi

I don't think you are going to find a better practical guide on survival analysis using SAS than Paul Allison's book. I did a lot of searching myself. There are good chapters on survival analysis in 'Modern Applied Statistics with S' by Venables and Ripley as well as 'Data Mining Techniques' by Berry and Linoff.

• ▶ Reply

Permalink Reply by Kumud Joseph Kujur on November 3, 2010 at 1:24am

Hi Bashshar,

Sometimes back I was researchering on calculating CLV. I read some of the web resources and came up with the following formula:-
Customer Lifetime Value (CLV) = Customer Present Value (CPV) + Customer Future Value (CFV) – Cost

Where,

CPV is present value in terms revenue contribution till date.

CFV is potential revenue contribution in future.

Cost includes administrative, retention and miscellaneous cost (At this moment I can’t define it)

Calculating CPV is fairly simple as revenue contribution till date is mostly known. Calculating CFV is a tricky part. Estimating a customer’s future purchase and probability of his staying loyal to company is difficult. Hence, we should be able to estimate 2 things,

1) How long a customer would be staying with the company? In other words calculating his retention probability with time duration

2) If a customer is staying with the company for predicted duration of time t from the current date, then what would be his total revenue contribution during this period?

For 1), you would be running survival model. For 2), you might run repurchase model/propensity to buy or manual calculations etc.

I used proc phreg in sas for running survival analysis. I agree that getting a good text from internet on survival analysis is hard, although I haven't read Paul Allison's book.

Regards,
Kumud

• ▶ Reply

Permalink Reply by Idielle Walters on November 3, 2010 at 2:52am

Hi

You can always start by calculating the profitability value (revenue - costs) for each customer and then take the NPV (net present value) of the customers profit. This will give you an indication of the value the customer will add to the company over a period of time. However, if you want to calculate the CLV of the customer you need to use survival analysis techniques to quantify the instantaneous risk that churn will happen at time t, given that the customer already survived till time t. The following paper is from one of the SAS SUGI proceedings and explains how to do this in SAS.

Modeling Customer Lifetime Value Using Survival Analysis
− An Application in the Telecommunications Industry
Junxiang Lu, Ph.D.
Overland Park, Kansas

• ▶ Reply

Permalink Reply by Yi-Chun Tsai on November 8, 2010 at 8:51am

Hi,
I was thinking about using survival analysis to calculate life time value for non-subscription based business. In non-subscription business, you don't have the clear cut definition of a customer CHURN and turn to yout competitor. Do you have any suggestion as to first of all, a good way forward? Thanks.

• ▶ Reply

Permalink Reply by Idielle Walters on November 10, 2010 at 8:04am

Hi

I would suggest you create a hazard curve for the hazard probabilities for non-subscription based customers.
The horizontal axis would be the tenure of customers measured in days/months, while the vertical axis would be the probability that customers stop at a particular tenure point. The hazard would normally increase up to a particular tenure point after which it would decrease and it will give you a lot of insight into the customer life cycle of non subscription customers. Furthermore, it would also be good to classify your customers as new customers and long tenure customers since the churn behaviour of these two groups will be different.

• ▶ Reply