How would your marketing, sales, product strategies, and budget allocations change if you could predict which audiences will remain loyal to your brand for years to come? What if you knew which customers would visit your site and make just one highly-discounted purchase before falling off the radar completely? And what if you knew that some customers you are already planning to spend lots of your budget to retain actually aren’t likely to bring lots of value to your organization?
Customer Lifetime Value (CLV) demonstrates who your best customers are. Using this metric effectively revolutionizes the approach to both acquisition and retention marketing.
What is CLV?
Simply put, CLV is a metric that estimates how much value (usually revenue or profit margin) any given customer will bring to your business over the course of the total time they interact with your brand—past, present, and future. It consists of three different factors: average order value, purchase frequency, and lifespan as a customer.
It’s not difficult to approximate total cohort-level CLV from basic orders data. Here is an intuitive and widely-known approach: use customer-level recency (R), frequency (F), and monetization (M), in order to identify customers as active or churned and to simulate orders behavior into the future. Models that build on this type of behavior are known by the acronym RFM.
RFM models are versatile and can be used to calculate all variations of CLV. To add customer-level resolution, assume active customers have identical remaining CLV (and non-identical past value). To add time-level resolution, assume active customers will reorder a fixed number of times at regular intervals since their most recent purchase.
How does CLV vary?
CLV has come to be known by many different names. Here are a few:
- Customer lifetime value (CLV)
- Lifetime value (LTV)
- Predicted lifetime value (pLTV)
- Lifetime revenue (LTR)
- Customer-level lifetime value (CLTV)
These names represent variations of the same idea: modeling customer behavior over time.
However, the definition of “value” in each term may vary. It’s common to use currency (either revenue or profit), and currency may or may not be discounted by time (as done in some financial models). Value may also indicate product demand, referring to direct product purchases or indirect website visits, app log-ins, or product page views. You’ll even see value used to indicate social and brand engagement, in terms of direct referrals, indirect likes, clicks, followers, and reposts.
Finally, CLV can refer to customer value that’s been aggregated over customers, time, both, or neither. CLV may refer to the total value from a cohort of customers, which is a back-of-the-envelope calculation. Customer-level CLV can always be aggregated into cohort-level CLV. In addition, CLV may refer to total value over a period of time, such as a customer’s lifetime. This can be measured in increments of years, quarters, or months (common in retail), or increments of weeks, days, or hours (common in gaming).
When time periods extend from present to future, we typically say “future” or “predicted” CLV; meanwhile, time periods that include the past are referred to as “total” CLV.
In this Academy, we’ll tell a simple but common story. Given orders data: events are transactions, value is undiscounted revenue, and scope is weekly and at the customer-level. Rather than treat CLV as a single number, we see CLV as indexed by time (and customer) and then aggregate on demand. We have the option to generate separate CLV models to handle different types of customer value.
Why is CLV important?
Rather than taking the average of these three variables for all of your customers in aggregate, it’s important to calculate CLV at the individual level. That’s because your customers are individuals who are different from one another, so you need an individual CLV prediction for each one if you’re going to use this metric to its fullest potential.
Aggregate CLV simply does not allow you to adjust who you’re targeting with re-engagement or acquisition efforts to maximize the benefits of each campaign. A good CLV model assesses the commonalities of all the customers unique to your business, then combines that information with per-customer behavior in order to predict future purchases and future dollars spent with your brand.
But developing a CLV model provides more than just one useful metric. In the process, you can discover the most important features that indicate higher or lower customer value. Features are personal attributes that can be based on behavioral or demographic information, like whether or not someone opened an email on a Thursday, or what IP address they typically purchase from when shopping online.
What do you need to calculate CLV?
Calculating CLV at the individual level requires complex information, like the amount spent by each customer and the frequency and duration of their purchases. How much data you have available matters, too—models cannot be trained accurately without a large enough sample size to understand the variation in individual customers that interact with your brand.
To calculate individual-level CLV, you’ll need:
- A clear definition of what it should represent for your business
- A data scientist trained in machine learning
- Sophisticated models developed by researchers
- Lots of customers
- Customers with over a year of recorded purchasing activities
- Comprehensive, valid data about your customers and their behavior
- Lots of computation power
The following lessons and courses in the academy will cover how to calculate customer lifetime value and applications across marketing, customer service, sales, product, and more. Dive into the topics most applicable to your role, and check back for more lessons on CLV!