The correct way to think about CLV
You are now holding more data on your customer than ever, and as a consequence have the capability to truly assess the worth of each and every customer. This has put CLV (Customer Lifetime Value) into center stage, but can it be trusted?
Common Misconceptions about CLV Calculations
Before clearly defining CLV let’s take a look at what it isn’t:
- Historic Customer Profitability: You will often see the sum of a customer’s past transactions described as CLV. This is a very short-sighted approach, since if all of your customer value lies in the past, you are in fact out of business. Accurate methods needs to be forward-thinking.
- Oversimplified New Cohort Analysis: Sometimes, in an attempt to describe value captured, analysts will describe a lift in AOV for new customers as CLV. Capturing CLV for new found cohorts is an extremely difficult concept and anything that does not involve cross-cohort analysis and time varying covariates is probably something different altogether.
- 12-CLV/24-CLV/36-CLV: Expected Spends over limited time periods (12 Months, 24 Months, etc.) are also often described as CLV. This is useful information when it comes to estimating sales for the upcoming years, but true CLV is open-ended. If that isn’t the case a business might end up overvaluing customers at the end of their life-cycle and undervaluing newer ones.
The one true way to CLV
CLV is the present value of future (net) cash-flow associated with a given customer.
This means that it is a rigorous, forward-looking methodology capturing the full life-cycle of a customer. In order to achieve an accurate view of CLV a business needs to use historical data to derive a complex set of parameters which will let you infer future purchase frequency and AOV for individual customers based on their RFM (Recency, Frequency, Monetary Value) values. The answer you get for this parameters is heavily dependent on what data is used, how that data is prepared and what statistical modeling is applied to it. Standardized methodologies like Pareto/NBD and Beta-Geometric/NBD allow for great accuracy, but when implementing CLV for business use-cases these should just be the baseline.
The truth is that there actually isn’t a one true way for CLV. (Sorry.)
The right way to calculate it is to tailor it to your business and specific use-cases. That might be non-intuitive since the lifetime value of a customer shouldn’t constantly be changing, but what is right for Retail might be wrong for B2B. Ecommerce and Store Sales can give different results, and sometimes customers can be converted from a low-tier to a high-tier. CLV calculations are difficult and need to be done at a regular basis if their results are to be applied in a business context.