Customer Lifetime Value is a fundamental acknowledgment that not all customers are equally valuable to your organization. It's important to understand the effects of recency, frequency, and monetary value have on the CLV of each customer in your organization. Learn more about how to calculate CLV and predict a customer's future value in this video in our Evidence Based Marketing series.
This is the third video in our series on Evidence-Based Marketing and in this video, we're going to dive into the concept of Customer Lifetime Value.
As you can see in the evidence-based marketing framework, in the first video, we talked about the integration and enrichment of event in customer data. Since that concept is very foundational to the modeling of customer lifetime value, if you haven't watched that video, now would be a good time to go back and watch it before proceeding with this video.
To get started, it's important to remember that CLV is a fundamental acknowledgment that not all customers are equally valuable to your organization. In fact, it follows very closely to the concept of the Pareto Principle. The Pareto Principle states that for many events, roughly 80% of the effects come from 20% of the causes. Unfortunately, the inverse is also true. Twenty percent of the effects result from 80% of the causes. To make this more concrete, let's substitute causes with customers, behaviors, and ad spend. Let's substitute effects with revenue and profit and assume they take place within some predetermined period of time in the future. Let's say one year in this case. With these substitutions made, the three primary use cases of customer lifetime value modeling reveal themselves.
First, we want to use it to reveal the characteristics of the 20% of customers that will deliver 80% of organizational revenue and profit over the next year. We want to understand how those customers differ from the 80% that will only deliver 20%. For clarity, let's call the 80% lower-value customers and the 20% higher-value customers.
Second, we want to understand the specific behaviors that are different for customers who are lower-value in the future period versus those who are higher-value in the future period. Third, we want to focus our ad spend on developing and acquiring more of the higher-value customers on encouraging and incenting behaviors that result in higher-value customers and on discouraging and avoiding behaviors which result in lower-value customers.
At this point, if I've done my job well, you're saying to yourself, "That's great, but how do I predict which customers are going to drive the 80% and which are going to drive the 20?" To descend, let's bring in the purchasing history for four customers.
Let's label the past, present, and future time periods. As a quick aside, customer lifetime value is sometimes referred to as CLV, CLTV, or simply LTV for lifetime value. In this video, we're using all of them interchangeably. Additionally, the term lifetime and customer lifetime value is more frequently thought of as a payback period than a true lifetime value. A payback period is essentially a way of saying over what future time period do I want to measure the benefit from a current investment.
Our goal here is to assign each of these customers a value in the present based on what they've done in the past and what we anticipate they will do in the future regardless of whether we are looking at their past behavior or predicting their future behavior. Three important marketing concepts are expressed in the graphic on the left-hand side. Those three concepts are recency, frequency, and monetary value of the customer's past purchases. This is a big reason you need the clean data foundation we described in part one of this video series. You need the recency, frequency, monetary value, and many other characteristics to perform descriptive and predictive analytics calculated for every customer throughout the history of their interactions with your organization.
In the diagram on the left-hand side, we visualized recency as the distance between the present time and their most recent purchase represented by the red circles, you can see their frequency represented by the number of red circles on their individual timeline, and you can see their monetary value of each purchase based on the size of the circle on the timeline.
Technically speaking, CLV is the present value of the future cash flows generated by a customer over the payback period, which is in the future. And while that's the technical definition, many organizations start with a simple calculation of lifetime value looking backwards at the customer's transaction history. Unfortunately, this is a very inaccurate and distorted view of the customer's potential future value. The heart of the problem with these analyses is that they don't incorporate recency, frequency, and monetary characteristics. For example, an analysis that only incorporates frequency would equate customers C and D, even though D is far more likely to be gone to our organization based on the recency. If it's based solely on monetary value, it might equate customer B and C if the value of customer B's first purchase was sufficiently large.
Instead of looking backward, we want to look forward and predict their future value. While there are many methods of predicting customer lifetime value, what I'd like to focus on for this video is a simple methodology championed by Professor Peter Fader from Wharton.
Clario uses this and a variety of other methods including Bayesian inference in our platform. The model uses very simple transactional data inputs to predict three things in a non-contractual setting. First, the probability the customer will return and purchase again in the payback period. Second, the number of transactions that customer will have during the payback period. And third, the monetary value of those transactions in the payback period. Now, I don't want the name of this model to scare you away, this was explicitly designed to be easy to calculate. I'll put several links in the notes below for an Excel version of this or other resources that can help you understand the mathematics of it. But the important takeaway is that this is a surprisingly accurate way to predict the individual customer's probability of return, their number of transactions, and the future monetary value when they have an established transaction history.
One final methodological comment. Although I've been talking about revenue and/or profit, you really want your customer lifetime value measure to be a profit measure. If we're using a profit measure, we can be sure we've incorporated all of the costs associated with satisfying that customer's demand and we are getting views that are distorted by promotional buying or extremely high return rates or shipping costs.
N ow equipped with a firm grasp of the fundamentals of customer lifetime value, let's look at one of the most common places where an organization begins using CLV. And that's why assessing the customer lifetime value of a specific behavior, typically, customer acquisition. One of the first things you might notice is that not everyone purchases at the same time, so an important data cleansing step, reflecting again on the first video, is to line up these behaviors so they fall within the same analysis periods. When we're analyzing customer acquisition, we can't use the same methodology we used for a customer with an established purchase history. For example, at the moment they purchased, these customers all had the same recency and frequency. They did have different monetary values, and that could be a predictive factor, but we also want to bring in some of those other characteristics I mentioned earlier. These should include the marketing channels that influence the purchase, the product mix in the shopping cart, including the quantity and product categories, the order channel of the purchase, and the promotional characteristics of the purchase. When we incorporate each of these factors into the model, a single number is produced which represents the value of this customer over the payback period. There are many reasons these values vary other than the monetary value of the purchase itself.
Just some examples include the marketing channel of influence, such as a PLA acquisition being of lower value, the product margin levels, the order channel. So a market place order, for example, might have a lower lifetime value or the promotional nature of it a customer required via a deep discount will typically be a discount buyer in the go-forward basis and therefore have a lower CLV as well. For the other side of this equation, we want to understand the cost of acquiring this customer. Instead of thinking of CAC, or customer acquisition cost, in the traditional sense, at Clario we like to think of contribution per purchase, or CPP. CPP basically says let's add up all the costs of the attributed marketing events that led to this customer's acquisition purchase, but then we're going to net these out of the margin generated on their first purchase. Hopefully, this is also a good reminder of the importance of a clean and rich data foundation and good, reliable marketing attribution. Both of those things talked about in part one and part two of the series are critical because they influence the prediction of customer lifetime value and the marketing cost allocation against the customer's first purchase.
I can very quickly interpret the results of this customer acquisition analysis. On the first customer, I lost $25 in contribution net of marketing cost on their first purchase, but they're worth $25 to me over the next year. Generally, this would be viewed very positively. I didn't lose any money and I'm going to make back everything I invested to acquire them over the first year of my relationship with them. I'm a little underwater on customers B and D, and so, I might want to look at those programs and see if there's something about the product margin levels, the marketing channels that I'm using to acquire those customers, potentially dial it back a little.
Customer C is worth investigation. We're $10 positive over the next 12 months. This could be an interesting opportunity to say increase budgets or bids on a digital marketing program, and acquire more customers or investigate the other characteristics, the product margin, the order channel, the promotional characteristics, and understand the root cause of why we're doing so well in that program and maybe double down on it going forward.
Hopefully, you can also see how this approach can be generalized to other high-value behaviors, for example, customer reactivation, second purchase, and other up-sell and cross-sell opportunities. A similar methodology can be used to analyze high-risk behaviors, taking two cohorts of customers, one who engaged in a high-risk behavior and one who didn't. High-risk behaviors might include opting out of your email communications, lapsing to an older recency bucket or quitting the membership or loyalty program. We can then assess the CLV of those customers and decide what the opportunity or tangible cost of that high-risk behavior is to the organization, and how much would be appropriate to spend to avoid its occurrence.
So I hope this video has done a good job introducing you to the concepts of customer lifetime value and how CLV builds on the enriched clean data foundation we presented in the first video, and the attribution foundation we presented in the second video. We're going to be building on the concepts presented here in the next video when we talk about how to build predictive models which will determine which customers are likely to engage in these high-risk or high-reward behaviors, and how you can segment them and create audiences for activation. And as always, I really appreciate you taking this evidence-based marketing journey with us. Thanks.