Customer Lifetime Value Modeling

Following my previous post on dropping unprofitable customers and some posts on blogs that I follow (Kevin) (David) (Jim) (Ron), it's time I touch on customer lifetime value.

Getting past the definition that customer lifetime models measure the present value of future cash flows that are expected from customers, here are some characteristics of a workable model:

• measure lifetime value at the customer level - unlike most approaches, I believe that a workable model will not be built at an aggregate level, (for instance as an average of present value measured across all clients) but rather a model that is built for each customer
• the future is more of the same - although you'll be projecting cash flows, our best bet is to bese them on known, historical ones instead of making up some ideal ones
• include costs and revenues - the calculation for prospective value should address margin, thus you'll need to account for renue as well as costs... namely, marketing costs.
• no magical numbers - although CLV is a single number, you'll need to be able to understand and group purchase patterns, so that you understand how there may be several paths to profitability; therefore, also score your customers in terms of frequency, magnitude and tenure

Then, divide your customers into three groups:

• profitable - the name says it all
• marginal - either marginally profitable or not, or margin-neutral
• unprofitable - essentially costs to the enterprise

Once you've built your model you will be able to better design your acquisition and retention efforts. In a nutshell, you want to:

• acquire new customers that look the most like your profitable ones - a clone or response model can be based on an augmented data set that contains these customers
• retain and cross sell to your profitable customers - see what insight you can gain from grouping (clustering, segmenting, whatever you want to call it) profitable customers per pattern if possible or relevant, this should give you an idea of how different purchase patterns are and how you can position products to foster up-selling and cross selling
• increase the profitability of marginal or break-even customers - again, you want to foster purchase patterns that mimic those of the profitable customers. This is where great brand planning and advertising comes in.
• drop unprofitable customers - have them pay for support, charge re-stocking fees, insulate them from discounts and special offers, etc.

All in all, targeting will always follow a targeting process that is oriented towards identifying profitable customers prior to designing strategies or tactics, for instance, if these customers are highly responsive to DM in-home one week after airing a TV30 spot, do not send them a catalog (unless this proposition has been tested and proved positive), also, if you're developing a responder model, use a data set that is composed of profitable customers rather than just "customers".

Comments

[Jim Novo]

Thanks for adding this to the conversation! I think the more ways we explain the idea, the more likely it is people will extract what is applicable to them and use it. I’ll add a couple of comments over on my blog.