There are two main mistakes that people do when they calculate the CLV. The first one is to simply multiply the customer ARPU by the EBIT margin of the company to estimate the customer profitability in a given month. This is quite bad, but the second mistake is a lot more worse: some people forget about customer churn i.e. they assume that the customer will remain a customer until the end of time. Or if they do assume customer churn, they assume that the probability that a customer churn is constant over time (and independent of the customer); in essence, they use a model of customer lifecycle and customer churn that is totally inappropriate for telecom service providers in a competitive environment. If you want to do better than most, then keep reading until the end of this article.

In theory, calculating the CLV is easy. It is a bit like calculating the NPV (Net Present Value): discount expected future free cash flow to the present. Repeat the calculation for each and every customer and this gives you an indicator of which customers are valuable and which are destroying value. In many cases, you will see that 20% of your customers generate a huge amount of value, 60% of customers are breaking even and neither creating nor destroying value, and 20% of customers are destroying value in a pretty massive manner. A big ‘Aha!’ effect for sure. Show that to your CFO and your COO and they will be…. scared? Intrigued? Questioning?

All right, but let us return to the CLV calculation. As for the NPV, the problem is that you need to estimate the following parameters:

- the discount rate
- the expected value of future free cashflows

and this for all customers! This gives us two problems to solve.

So how do you do this? Well, the first simplification is to assume that the discount rate is the same as the discount rate that you use for the NPV calculation, i.e. the weighed average cost of capital (WACC). If you know your company WACC (8%? 10%? 12%?) then you can use the same number in the CLV calculation. Unless you want to treat a customer as a single ‘project’ and wish to calculate a customer WACC. If you are very good at estimating ‘project’ WACC rather than company WACC, then you can do so (most likely using a Monte Carlo simulation).

If you do not have a clue what we are talking about here, then forget about it and simply take the company WACC (good to know that you have colleagues in Finance who can calculate the WACC for you!). If you have multiple business line e.g. Mobile, Fixed Line, Internet, TV, and if you want to calculate the CLV of a customer broken-down in its components: the CLV from Mobile, CLV from fixed etc, then remember that it might be more appropriate to take the WACC for each business (fixed, mobile etc) rather than the overall company WACC.

But let us not confuse you further – in most cases you will be using one single WACC across the board, for all customers. So problem #1 is solved.

The bigger problem is #2: the expected value of future free cashflows. Problem #2 is a collection of three problems: we need the cashflow, in the future, and the expected value.

Now we can hear you saying: where the hell am I going to get the cashflow on a customer basis? OK, replace cashflow with NOPAT = EBIT x (1-Tax Rate). So the problem is now to get the EBIT on a customer basis. This is easy if your company has done a *Service Costing* exercice in the past. In which case the EBIT is, for each customer who has an individual traffic profile (call volume per month, split of on-net, off-net, incoming etc) the difference between Call Tariff and Call Cost summed over all call types generated and received by the customer during the month. For the Call Cost you should take the Fully Allocated Cost (FAC) or the Long Run Incremental Cost (LRIC), which is similar to a marginal cost.

If the EBIT or EBIT margin is not available at the service (call type) level, then you are unlikely to figure out a meaningful EBIT at the customer level. Taking the company EBIT margin and multiplying by the customer ARPU is a very rough approximation, because you are implicitly assuming that all services in your business have the same profit margin (which is very unlikely). If it were the case, then your customers would equally be profitable in terms of EBIT margin, and you would probably not worry about the CLV. You would think that customers with a high ARPU are valuable customers and those with a low ARPU are less valuable, so ARPU would be used as a proxy for CLV. Which is also wrong because in that case all your customers are profitable today (unlikely) and you forgot to think about two other issues: expected future values.

Whether you take NOPAT (yes!) or ARPU x CompanyEbitMarging (if nothing else is available), we still have to figured out what their values will be like in the next months and years. So we need to *forecast NOPAT* for each customer in the future. And to get the expected value, we need to take into account the customer lifetime cycle i.e. what the remaining lifetime of the customer is. Which brings us to *Churn*.

To forecast NOPAT, you should forecast future tariffs – most likely tariffs will decrease in the future? Also ask your colleagues in Finance who provided the *Service costing* results to generate a forecast of costs for you – if their Service costing exercice is good, then it should not only provide a backward view but a forward view as well. Once you are done with the tariff forecast, the cost forecast and the customer usage forecast (call minutes etc) then you have a NOPAT forecast for this customer.

But while doing so, you might have wondered how many years into the future your forecast should go? This brings us back to the customer lifecycle and the issue of churn. You need to calculate the probability that the customer churns some time in the future. To do this, you need a Survival function that reasonably approximates your customer churn behaviour. The Survival function at time t is: S(t) = Pr (T>t) which means that it is the probability that the time of ‘death’ (sorry, ‘churn’) T is higher than t. The Survival function starts at 1 when the customer is activated (t=0) and is a decreasing function of time, converging towards zero in the long term. To simplify matters, we will assume here that there is only one Survival function valid for all customers and its profile remains the same over time. The Survival function can be estimated from historical customer data and let us assume that you have just done that.

Now for the calculation of the CLV_{i} (t_{0}) of customer* i *at time *t _{0}*

_{ }(now), you must multiply the future NOPAT

_{i}(t

_{1}), NOPAT

_{i}(t

_{2}), NOPAT

_{i}(t

_{3}) etc. for customer

*i*with the Probability that this customer has survived (i.e. it has not churned) at time t

_{1}, t

_{2}, t

_{3}etc knowing that at time

*t*the customer was still alive (i.e. it had not churned). So the factors you need to multiple the NOPAT with looks like this:

_{0}1-Pr (t_{1}-1<= T < t_{1} / T> t_{0}): probability that the customer does not churn during year *t _{1}*, knowing that at time

*t*the customer was still alive. This factor can be expressed as a function of S(t) as follows: 1 – ( S(t

_{0}_{1}-1) – S(t

_{1}) ) / S(t

_{0})

To summarise, the CLV_{i} (t_{0}) of customer *i* at time *t _{0}* (now) is the sum of the discounted future NOPAT

_{i}(t), using the WACC as discount rate, and using the factors 1 – ( S(t-1) – S(t) ) / S(t

_{0}) to weigh the NOPAT

_{i}(t). The factor used as weight is the remaining portion of NOPAT

_{i}(t) that is expected to be generated from the customer knowing that the same customer might churn at any time after t

_{0}.

If you have followed us so far, then congratulations! You should better understand how the Customer Lifetime Value can be correctly calculated and which mistakes can be avoided. We will come back on the issue of the Survival function in a future column on Churn. Keep reading!