Let's talk a little bit about customer retention or customer attrition, when we're talking about the subscription context, it's actually a little bit easier for us. This is where the retention behavior or the attrition behavior is directly observable. It's revealed based on someone's actions. In order to cancel service with your cable provider, you have to call them, you notify them that you want service cancelled. There's not a chance in the near term that you're going back to that company once you end that relationship. So this is pretty common when we're talking about contractual relationships, whether it's telecom, magazines, gym memberships, you must inform the company that you are ending the relationship. That's when you stop being billed. We're going to look at a couple of examples and when we're looking at the exercise later on in this session, we're going to be focused on that particular type of behavior. The type of attrition that's a little bit more difficult to identify is latent or unobserved attrition, and this happens many times for retailers. If we think about your shopping behavior at a website or at a local retailer, you never actually have to tell them, if you're not going to come back, you just decide to start shopping elsewhere. Well these retailers when their making the estimate of how much a customer is worth, they have to take into account some measure of customer retention. But it's more difficult because they don't know with certainty that a customer has ended the relationship. And so what's often done in practice is to say, if you have not conducted business with us let's say within the last quarter, within the last six months, within the last year. We are going to assume that you've lapsed as a customer and we're going to assume that future revenues from you are zero. And while it's not technically correct, in practice, that's going to generally work reasonably well as far as modeling the customer value, right? So we're going to be focusing in a lot of our examples on observable attrition where we observed the retention decisions. So for example, think about Netflix, every month that you were auto billed, you've made the decision to renew service. If you decide to cancel service that's the end of your relationship, the billing stops at that point in time. So we know exactly when you churned as a customer. All right, one thing that's often talked about is the idea of wanting to increase our retention. We want to have as high a customer retention as possible. Ideally, we'd like to have 100% customer retention. And that's something that's going to be hard to achieve. And in fact, probably nearly impossible for a lot of categories. We've done some research in this space, and working with a telecom provider, what we found was most of the churn that's happening actually comes about for reasons that the company can't control. If we think about the reasons that customers churn. It may be because they're moving out of the service area, maybe because they're experiencing financial hardship. Neither of these are things that the company can do any about. What a telecom provider can do something about is the quality of service is responding to competitive efforts. And so ss long as you have reasons for churn that are uncontrollable, you're never going to hit this idea of 100% customer retention. And even if we could say, spend an amount in order to get to that 100% customer retention. Is it worth it, they're just some customers who are inherently less loyal than others. Well you can keep on throwing money at those customers, and we can use an approach whether it's a duration model. Or something like logistic regression to relate how much we spend on retention after that customers decision to stay with their current provider. But think how much your necessarily going to have to spend on those customers who are inherently less loyal than others. And is that the best place to be spending those resources? Might the resources be better spent trying to acquire new prospects? And that's what our exercise later in this section is going to take a look at. So if we think a little bit about this role of customer loyalty. This is a schematic that was taken from work by Eva Ascarza and Bruce Hardie. Taking the approach that loyalty is common driver of both customer retention as well as usage behavior. The more loyal I am to a particular provider, whether it's a telecom provider or a financial services, provider. Or let's say a performing arts center not only am I going to use more of the services that they provide. But I'm also going to stick around longer as a customer, all right? So we're hitting both on the retention aspect as well as the relationship development aspect, the breadth and depth of the relationship. So how do we connect customer attention to customer value? And this is a graphic that has been taken out of work by Gupta, Lehmann and Stuart from 2004 that relates the customer retention decision. And what they try to do in this research is tie the customer acquisition, customer retention behavior to stock price movements. So let me walk you through what you see in this schematic right now. So a given cohort, these are customers who were acquired at the same point in time. So let's focus for a second on Cohort 0. We begin with a number of customers n sub 0, and that's in the first period. Well in the second period, we're going to make the assumption that we lose some of those customers. So we retain a fraction r, of those customers. So after one period, how many customers remain? What's the fraction r multiplied by how many customers we started out with n0 of those customers, of the n0 times our customers we have after one period. Those that remain, if the next period, if we assume the retention rate are as constant. Then it 's going to be n0 times r squared, that will retain one more period. N0 times R cube, that will retain a period after that, n0 times r raised to the teeth power. So gradually, we're losing customers over time. But we're retaining our fraction of the previous customers, all right? So that for customers who started in Cohort 0. What about the customers who started in time 1, in period 1? Well same story, they started out in period 1 with n1 customers. After one period, if our retention rate is constant. They have n1 times r customers, n1 times r squared customers 2 periods after they began, n1 times r cubed 3 periods after they began, all right. And so that's looking cohort by cohort, but what if we were to say, let's look at time 0, time 1, how many customers do we actually have? Well if we look at time 0, we started out with n0 customers, that first cohort, all right. What about at time 1? Well at time 1, we've got n0 times r customers. So those from the initial Cohort 0 that are still with us, and then we have n1 customers that came in with the new cohort. What about at time 2? In time 2, we're going to have n0 times r squared, people who've been around for 2 periods, n1 times r. Customers from Cohort 1 who stuck around at least for one period and then new customers coming in Cohort 2, n2. All right, so the schematic looks both at the retention behavior within a given cohort as well as at the acquisition behavior. So if we're looking at n0, n1, n2, this is looking at the customer acquisition process, right? So the value for this particular organization, if we take these three cohorts. We've got three cohorts of acquired customers, and within each of those cohorts, do we retain the customers? And we're making the assumption that each of these cohorts provide us with margins potentially based on how long they've been around as customers. So we can use this approach to say, all right, well, suppose we observe a company's behavior or their customer's behavior in terms of the number of customers that they have around. If we have an estimate for that retention rate, we can identify the rate at which customers are being acquired. And we can come up with projections for, not just the value a particular cohort, but an estimate value of the total customer base. And that's what was done in this particular research. So the authors had focused on five different companies. So Capital One, Ameritrade and ETrade focused on financial services, Amazon and eBay, online marketplaces. The authors had come up with an estimate for the value of the customers, as well as looking at the market value of these particular companies. So we have an estimate for the value of the customers, estimate for the values of the companies, and in some cases such as Amazon. Trying to draw parallels here, we see that there's a little bit of a disconnect. But if we focus on the numbers from the financial services companies, you know these numbers, not as far off. So if we look at the numbers let's say for Ameritrade, to some extent for Capital One. And then for ETrade, the inferred value of the customer base lines up pretty well with the market value of the company. So the customer base can be thought of as one of, if not the most valuable assets that companies have. It's an asset that we need to put marketing efforts into acquiring these customers, we need to put marketing efforts into retaining these customers. So what the authors had also conducted in this analysis was looking at what would be the best place for us to expend resources? Is it to increase the retention rate? Is it to increase the acquisition rate? Would it be to increase our margin or to reduce the discount rate by enlarge with the authors had found here? Retention is the place where you get the biggest bank for your book, right? So we can use this framework to try to quantify the impact of investing in retention investing in acquisition. So the different levers that we can potentially pull, what's going to have the biggest impact for the company?
