Re-evaluating CRM cliches No 3: Customer Lifetime Valueby
Somewhere in the chorus of every consultant's hymn in praise of Customer Relationship Management is an expression along the lines: "To manage customer relationships properly, we need to understand Customer Lifetime Value."
This week I want to take that phrase, try to examine what it means, and see how we can use it - or similar concepts - in a practical CRM project.
Let's start at the beginning with defining Customer Lifetime Value. A good source of Brownian motion, says the Hitchhiker's Guide to the Galaxy, is a hot cup of tea. A good source of CRM definitions is the Peppers and Rogers web-site.
Their glossary defines Lifetime Value as being the same as Actual Value, which they define as: "the net present value of the future profit stream from a customer". Now let me say immediately that I think this is an excellent definition of Customer Lifetime Value, nearly all consultants would agree with it, so the following is in no way meant to reflect criticism specifically on Peppers and Rogers, at least not more than it does on all CRM consultants who use the expression Customer Lifetime Value.
For the financially illiterate amongst us, let's also try and outline what is meant by "net present value". As I understand this, we expect to make profits from a customer over a period of time (years, months). Let's calculate the profit we'll make in each time period (year, month), and add them all together, to get the total value we're going to earn from the customer. Of course, because a dollar (or pound) in my pocket today is worth more than a dollar tomorrow, we should discount future dollars to reflect that loss of value, and discount more the further away in time the future profit occurs. Typically, you might discount future profits by, say 15% a year. So the net present value is the future profit stream, with each element discounted appropriately.
Now, I'm sure the accountants amongst you are rolling in agony at this definition, and if so, feel free to add a response to the editorial with a better definition (in Plain English) of net present value.
So now we've got a definition of Customer Lifetime Value which says it's the sum of the future profits we get from the customer, where those profits are discounted to take account of how far they occur in the future.
Your CRM consultant will then tell you that you should make your decisions on how you treat your customers based on their Life-Time Value (LTV). Let me give you a practical example of how LTV can be simply used in practical life.
In the UK retail banking industry, university students are treated particularly favourably. Despite the fact that most of them are financially irresponsible, liking nothing better than to spend any money they can get hold of on parties, alcohol, and other entertainments; despite the fact that they have very little money and no idea of how to manage it, the banks fall over themselves to offer students loans, and help them financially through their university educations. Why? Because they perceive that these students will turn into the graduates of tomorrow, and then into the high-earners of the day after, so becoming highly profitable customers of the future. In other words, because of their perceived Customer Lifetime Value, the banks are willing to be nice to them today. Of course the banks are relying on these student customers being loyal (or on customer inertia) to realise these future values.
So this idea of Customer LTV seems to make sense. What possible issues could we have in making it a core part of CRM environment?
Well, let's start with a definitional issue. LTV covers the future profit-stream coming from the customer. Does the past profit stream from the customer have no value? Is the commercial relationship between companies and individuals so cynical that 50 years of past profit to a bank is worth nothing in evaluating how we'll treat a customer now and in the future? I suppose for many companies it really is only about the present and the future, and nothing about the past, but if so, customers' attitudes towards their suppliers should be similarly hard-hearted. We're certainly not dealing with a corner-shop here.
Are there other issues that worry us about customer LTV? Well yes. Let's start with a "reductio ad absurdum" argument. If we don't distinguish between future profits and current profits other than by discounting, we can, in theory, go bust by maximising our net present value. If we identify a set of customers who are unprofitable to begin with, but have such high profits out in time that they outweigh, even after discounting, the immediate losses, the sum of the LTVs could encourage us to go after that set of customers and find that though we're going to make lots of money in the future, we run out of cash (which you do only once) before the predicted profits arrive. Is it to naïve to so suggest that there are a lot of dotcoms who have just suffered the consequences of similar problems?
A far more likely scenario is that before we enjoy the profits from those long-term valuable customers, a competitor, grown fat on the profits from immediately profitable customers (though less attractive in LTV terms), devises a strategy to steal all our customers we've been nursing through their lean times. So tactics can destroy a long-term strategy.
Another key issue is how long is the customer's life? Do we really think that we can predict a customer's behaviour, say, 50 years into the future? This has to be madness. The student of today is going to have perhaps 50, perhaps 70, perhaps 100 years of economic activity in a fast-evolving economic environment. Do we really think we have any idea how we might begin to predict the value of that relationship? The predictive techniques used are also built from historic data, so we are trying to predict the economic value of customers over the next 50 years from customers who have been active in the last 50 years (i.e. from 1950). This is patently nonsense. In practice, companies I've seen trying to implement Customer LTV measures limit themselves to a 3-5 year forecast, but this is hardly lifetime value.
Finally, let's add a reality check. Nearly every company I've worked with has had enormous problems defining historic customer profitability (How do we allocate the fixed costs? etc.), and this has frequently precluded implementing historic measures of customer profitability. If we can't get agreement to how to measure the fixed past, do we really think we're going to get agreement on how to measure the indeterminate future?
Summarising so far, I think we have to say great theory, shame about the reality. But I don't quite want to give up on the idea yet. One other area of difficulty I have with LTV is that it mixes current value/opportunity with future value/opportunity. There is a tactical opportunity to make money now (or better, in a service industry we have an ongoing revenue/profit stream which will continue given that it doesn't change) and a lot of longer-term profit opportunities. LTV adds the discounted values of those together and gives us one variable. Is that sensible? Perhaps we would gain some value if we treated those two figures differently. Perhaps we could call one of those "current value" and the other "future value". Perhaps we could also make a couple of other changes. Given the difficulties with measuring profitability (usually due to the allocation of fixed costs), maybe we could try and measure the contribution which the customer makes (the revenue minus the costs of doing the transaction).
So if we had a measure of customer contribution (immediate value) and customer potential (longer-term value) would we gain anything? Well, we'd probably have customers in a number of different categories:
- customer not contributing today, and with limited potential (manage up or out)
- customer not contributing today, but with significant potential (maximise customer value)
- customer contributing today, but with limited future potential (customer retention)
- customer contributing today, and with significant potential (favoured customer)
Now this is beginning to look interesting. Not only do we have customers in different categories, but there seems to be an obvious customer strategy for those different categories. If we can measure current contribution, and put in place a measure of customer potential, then we can segment our customers based on those measures, and derive segments that should have different customer strategies applied to them.
Now I can feel a few "yes, buts" coming on. The two most obvious ones are:
If it's so difficult to measure future customer profitability, how come you think you can measure customer potential?
The answer to that is complicated and too difficult to go into in detail in this editorial. In principle it involves identifying the major profit opportunities within a company's business model, building propensity models which measure the probability and value of each customer exercising that profit opportunity, and building a matrix of the resulting indexed scores. The sum of those scores can be seen to be a measure of customer potential. The methodology is outlined in a presentation on this site called: Understanding your customers better. I know a number of people who have gone a long way down the road to implementing such a matrix, and those who have been most successful in using it would much prefer if I didn't keep going on about how useful it is!
If I can't get agreement on how to implement customer profitability measures, how are we going to get agreement on measures of customer contribution and customer potential?
The answer here is to make sure that the algorithm you implement is accurate but imprecise. In other words it's the granularity of the answer that causes the difficulty. If you want to measure customer profitability to the nearest cent (penny) then you need a complex algorithm with plenty of room for debate. If you want to rank customers from high contributors to very poor negative contributors on a 5-point scale you need a much simpler algorithm. If we can rank customers on a 5-point scale for both contribution and potential, then we have a 25-cell matrix from which to derive our segmentation - more than enough to get started with - and derive value, provided we are fairly accurate (significantly better than random) in how we allocate customers to the matrix.
Please don't get me wrong in these editorials. These slightly ironic comments on LTV are meant to encourage a little of that painful activity called thinking. Similarly, I don't think that the implementation of a customer contribution/potential matrix is the be-all and end-all of Customer Relationship Management.
What I do think is that the competition between consultants on the theory of CRM has got so far ahead of where most client companies actually are, that there is real danger that we expend huge amounts of energy (and money) implementing a theoretical infra-structure which doesn't deliver any value for many reasons.
Many companies are trying to 'boot-strap' themselves into CRM. What they need, we believe, are a number of pragmatic methodologies which deliver value and represent a stepping-stone towards the CRM world. A number of people have said in response to my first editorial in this series that Database Marketing is one of those. I believe the customer potential / opportunity matrix is another.
Last week we talked about the need to provide the customer with a 360 degree view of the organisation. We hope to talk about other practical ways of making money out of CRM in future editorials in this series.
As always, the best measure of the success of these editorials is the comments you make, so please add a comment to this editorial, or write to me directly, at: [email protected]
The independent resource centre for CRM
To subscribe to the CRM-Forum weekly email newsletter, click here to update your profile.