Customer Lifetime Value in Three Dimensions

Series: How Smarter Attribution Leads to Smarter eCommerce Marketing

This post is the eighth in a series about smarter attribution marketing. Need to catch up? Check out the previous entries in the series:

1. All Attribution Models are Not Created Equal
2. Last-Click Attribution – Last Click, Last Choice?
3. Beyond Last-Click Attribution – The Customer Journey to Conversion Model
4. Turning Your Attribution Model into Online Marketing Optimization
5. Taking Attribution to the Next Level: Customer Lifetime Value
6. Supporting Your Attribution Strategy with Real-Time Programmatic Display
7. Is View-Through Attribution a Good Look for Online Marketers?

Going Deeper on the CLV

In post number five in the series, we introduced why marketers should concentrate on “long term” optimization based on the customer lifetime value (CLV) of their users instead of a pure cost-per-order (CPO) optimization calculated for each single purchase. We talked about how concentrating investment on new customers can help to achieve sustainable growth for your customer base. But even with this differentiation, questions remain. Who are the “right” customers (in terms of long-term value), and how should I treat my existing customers?

The goal of this article is to give you some additional insights not only on why, but also how to calculate the CLV. The basic idea behind CLV calculation is to understand the total value in terms of either sales or (in the best case scenario) margin that a customer creates within a given time frame. If you look more closely at the subject, a customer’s purchases can be broken down into three dimensions:

R – Recency
F – Frequency
M – Monetization

Purchase Recency:

For recency, you must be able to calculate how recently a customer has completed a purchase. When examining the recency dimensions of your customers, you can gain additional insights by analyzing the customer groups that purchased certain items to understand which items are most likely the next purchase, and to discover the average time until the next purchase for given products or categories. This kind of analysis is typically the backbone of product recommendation engines or post-purchase retargeting engines.

Purchase Frequency:

Frequency, like recency, it a matter of time — in this case a measure of how often a user is completing a purchase on the website. Even as a stand-alone data point on a per-user basis, purchase frequency can be a very useful dimension when deciding a customer’s potential value (higher frequency = higher potential value for a given time period). It becomes even more valuable if you begin to analyze aggregated user data, using cluster analysis to identify trends and understand which shopping behaviors indicate which CLV. For example, as an online wine reseller, your frequency analysis might reveal that buyers of red wine buy more often and more consistently than their white-wine-buying counterparts. (This example was discussed by Florian Heinemann of Project A Ventures in our dmexco seminar this year.)


The third (and trickiest) dimension of the three is monetization, a hard measure of how much value a user has generated — but it goes beyond pure revenues.  Monetization means taking into account the actual margin generated from the products purchased (not every product generates the same margin), as well as taking product returns into account. Especially for shops with a high return rate, the CLV per customer (and likewise the CRM group) can change dramatically once returns are taken into account.

From a tracking and analytics standpoint, monetization is the toughest dimension to measure. While recency and frequency can be evaluated with an advanced web tracking solution, measuring monetization requires much bigger data muscles, because the data warehouse where the data is stored has to be able to intelligently communicate with the business intelligence (BI) mechanisms. Don’t forget that, depending on the policy of the shop, returns can happen quite a while after the original purchase — to properly measure monetization, you need to be able to hold onto all that data in a usable way!

Advancing Your CLV Strategy

It sure sounds like a lot of work, but if you’re looking to sharpen your strategy, switching from a last-click to a customer journey attribution model as a first step will already dramatically impact how you distribute your marketing spends per channel. If you take it one step further to start working around not the single-transaction CPO but the potential CLV of your customers, the implications for long-term value creation are tremendous. You could break the CLV strategy down to three “levels” of refinement:

1. Splitting between new and existing customers,
2. Splitting based on the CLV of existing customers (CRM groups),
3. Predicting the CLV of new customers based on analysis of historical data of existing customers.

For the third point, advanced marketers with strong BI teams can analyze historical data from existing customers (usually in windows of 30-60 days) to find clusters of behaviors that indicate a certain CLV. In this way, marketers can predict the CLV of new customers either from the user’s shopping behavior or by the channel(s) through which the user reached the site — both dimensions can be a key indication of potential value. Depending on your product range, the category of the first-time-purchased product could also be a strong indicator of CLV.

Beyond the division between new and existing customers, many online shops are also starting to put more strategy into reactivation of existing customers. For example, a given shopping behavior may indicate that the next buy is probably on the horizon within a certain time frame — let’s say buying a laptop might mean you’re likely to buy a hard drive in 60 days. If the user (especially a frequent buyer, or taking seasonality into account) does not purchase within the expected time frame, it could make sense for the marketer to increase the CPO target for the push channels (such as display advertising or email) to convince the shopper to purchase again. This reactivation methodology of course works best with channels in which the marketer can make adjustments at a user level, such as programmatic display.

What’s Next in the Life of the CLV?

In time, we may see — and expect to see — that marketers will no longer have CRM “groups,” but rather that each individual user will comprise his very own dynamically changing CRM group, with a constantly changing CLV and strategy assigned to him, based on the indications of his shopping or browsing behaviors.

In the wonderful world of programmatic display, we can already see the benefits of reaching users on different devices with the CLV approach, which can help the marketer to adjust spendings, reach the user at the right time and track the entire customer journey across channels and devices. You’ll find more on that topic in our next post in the series!

Beyond the Click PreviewFor more insights into the world of attribution modeling, download our white paper Beyond the Click: Intelligent Attribution Modeling for eCommerce Marketers.