Understanding Customer Lifetime Value for E-Commerce

Young girl helps her grandfather place an online order.

Understand Your Customer Lifetime Value To Allocate Resources Wisely

Customer Lifetime Value (CLV) is a metric that indicates how much money a customer spends with your brand throughout their lifetime. Knowing the lifetime value of an individual customer helps companies optimally allocate personalized budgets as a part of their retention strategies. In the retina.ai blog, its CTO and Co-Founder, Brad Ito, explains how understanding a customer’s lifetime value at the individual level is beneficial to the business[1], both to maintain a careful budget and lasting customer retention strategy. Among several advantages of CLV applications, understanding and forecasting your customer’s lifetime value will help minimize wasting retention ad spend to highly valued customers who are already engaged on their own.

Customer Lifetime Value is an additional metric analyzed in LXRInsights, NetElixir’s proprietary customer analytics platform, that identifies your future profitable customers. LXRInsights segments your customers into high-value, mid-value, and low-value customers, tracking their purchase journeys from the first point of attribution to final purchase to help you better understand the nuances of their unique journey. Our CLV model built into LXRInsights projects which customers are likely to become high-value customers, who spend 3-5x more with your brand than average customers.

Our objective of calculating CLV at the individual level is to identify the value of each customer and anticipate the customers’ spend; from there, we help design a suitable retention strategy to help turn shoppers into loyal, repeat customers.

Background of Customer Lifetime Value Models

Fundamentally, CLV models are different for contractual (e.g. Subscription) and non-contractual (e.g. Retail) business models. Knowing customer churn is one of the most important factors in calculating CLV, but it is unknown in retail business models. Therefore, the BG/NBD method, a probability-based CLV model, proposed by Peter Fader[2] is used to tackle both the purchase and churn probability of each customer in retail transactional data. Each customer’s purchasing and churn behavior is unique in retail; thus, the model calculates CLV at the individual customer level to use the metric at its full potential.

LXRInsights Customer Lifetime Value model

In LXRInsights, we adopt the BG/NBD model proposed by Peter Fader, calculating CLV for the next twelve months. We specify twelve months because it makes more sense from a business model standpoint to know the value of customers in the near future to better adjust ad strategy so each campaign is impactful. Our predicted twelve months CLV, also called truncated CLV, represents the expected present value of the customers for the next twelve months, based on the customer’s past thirteen months of transactional data. The basic information required to calculate CLV is frequency, recency, age, and average order value (AOV) of the customers.

Diversity of Customer Behavior

The BG/NBD model handles the diversity of customer behavior in purchasing and churn at the customer level. Figure-1[3] below shows the diverse online shopping behavior of five different customers. Each bubble represents one order and the bubble size indicates the amount of money spent. The blue and red customers have regular intervals of purchases but do not purchase frequently. The orange customer made several small purchases in a short period and never returned. The green is a recent customer who hasn’t spent a significant amount until recently.

Example of different online shopping behaviors of five customers

Figure 1: Diverse behavior of different customers
Image from https://antonsruberts.github.io/lifetimes-CLV/

At the end of Month Four, predicting who will churn and who will be retained and turn into the most loyal customer is challenging. The model predicts the future worth of such customers with different behaviors and includes an estimate of which customers might churn; in this scenario, the orange customer is the most likely to churn. The model also predicts expected AOV on an individual level that would help you to wisely allocate your budgets for any marketing strategy. For example, spending a lesser amount to retain the green customer, and a higher amount to retain the blue, red, and yellow customers with respect to the predicted AOV would be an optimal retention strategy since the green customer is a recent customer who has spent less compared to the rest. Ensuring the whole shopping experience, from initial search to receiving their package, is a pleasant, seamless, and exciting experience that will help customers remember your brand.

Experimental Results

Our CLV model predicts the CLV, expected AOV, and expected number of orders for the next twelve months. Tables 1, 2, 3, and 4 below show some results of the CLV for different customers. The high age indicates an old customer whose first purchase was a long time ago, while the lower age shows a more recent customer.

Referring to table 1, “user_3” has a high age and recency, which indicates that the customer is an old customer and still recently bought from the business. Such customers have a high CLV, as they are loyal customers.

Customer Lifetime Value Model

Table 1: CLV data of high-value customers

Referring to table 2 below, we see a group of recent customers who are active and thus are assigned a good CLV by the model.

Customer Lifetime Value Model

Table 2: CLV data of recent customers

Table 3 below shows customers with a mix of low and medium CLV. For “user_7” in table 3, there is a large gap between recency and age, indicating that the customer did not return after placing a few orders. Because of this gap, they are assigned a low CLV and predicted as churned by the model.

Customer Lifetime Value Model

Table 3: Data of customers with projected low and medium CLV

The customers in table 4 have similar behavior to each other in regards to frequency, age, and AOV, but they are assigned different CLVs. The difference lies in that “user_9” never returns after placing a few orders, but “user_8” returns every so often for similar purchases.

Customer Lifetime Value Model

Table 4: CLV data of different behavior

Conclusion

Customer Lifetime Value unveils hidden information about your customers. According to past data, customers who are in the high-value segment now need not necessarily be profitable in the future. A customer in the medium-value segment could have a higher CLV for whom you need to run special campaigns to help them become high-value customers. Realizing the expected value of the customers at the individual level will help business leaders better allocate budgets and strategically retarget the right customers at the right time to encourage them to become loyal customers.

Visit LXRInsights to request a demo of our customer analytics platform to understand your customers’ future profitability better.

Further Reading

 

References

[1] Why it is important to calculate CLV at an individual level? https://retina.ai/blog/why-calculate-clv-at-the-individual-level/

[2] Peter S. Fader et al., “Counting Your Customers” the Easy Way: An Alternative to the Pareto/NBD Model, 2004, Marketing Science Journal, Volume 2, Issue 2, 275-284

[3] Figure 1: https://antonsruberts.github.io/lifetimes-CLV/

Request Consultation

Search The Blog

Categories

Why NetElixir?

18+

years experience

8+

years as a UPS Digital Connections Partners

1 of 27

agencies (out of 7,000+ U.S. agencies) in Google's Agency Leadership Circle

We love cookies!

Want results?

Let us show you how we can help.