The concept of Customer Lifetime Value (CLV) essentially treats customers as long term investments where you invest in them systematically over a period of time in the expectation that you recoup the customer acquisition and servicing costs plus some profits. In a sense they are similar to financial instruments such as bonds. As anyone with a background in finance theory knows, bonds can be priced, and so we can apply a similar logic to “price” the value a given customer brings to the business.

Customer lifetime value is therefore simply the calculation of the present value of future profits from a customer over his or her relationship with the firm. Today there is a lot of interest in quantifying the value of a customer over such an extended period of time. Many companies realize the importance of nurturing these relationships and “investing” in them.

From a business owner’s or executive’s point of view, engaging in such measurements can be of great value because you will be addressing some of these important business questions along the way:

1.  Are we spending too much money to acquire a customer?

2. Assuming that it is cheaper to retain a customer than to acquire a new one, on average how long do we retain customers?

3. Is the above assumption valid, i.e., how does customer retention compare with the costs to acquire a customer in the first place?

Challenges in performing a customer lifetime value assessment

CLV is cohort based. What this means is that we start with a group of customers at a certain time, analyze the retention levels and the resulting profits for that group. However this is easier said than done. There is a lot of necessary data preparation that goes into this type of analysis. Typically this involves sorting and organizing a lot of “flag” variables which are basically binomial or binary in nature: “did they buy?”, “did they spend \$X in the time period” and so on. Each of these variables will essentially contain a “Yes/No” or a “1/0” response. This requires having to write database logic to group customers by customer ID and/or transactions.

Analysts who are comfortable with programming tend to use SQL and run this data prep using the original database. But if you are not a good programmer, you can use a drag and drop tool such as RapidMiner to perform any data transformation. In either case, data preparation will involve first cleaning up “dirty” data. This is usually the biggest challenge.

What data is needed to compute CLV?

Before putting together a laundry list of data needed, it is a good idea to analyze past initiatives to develop some meaningful time frames for analysis and a strategy. For example, a business can revert back to a previous campaign to obtain some key attributes such as retention rates, margins and a discount factor.

Most basic CLV models make the assumptions of a constant retention rate and a constant margin. Additional information will make this clearer, for example if customers are in a contractual agreement such as cellular phone customers, then it is easy to identify a time frame for the computation. If not, then a timeframe for a “lost” customer may need to be defined, for example, no orders in 12 months implies a customer lost for good.

In the end, here is a list of derived and raw data that may be needed for analysis over the chosen time frame: Customer Revenue, Customer Cost to Acquire and to serve, Cost of Goods Sold, Number of Customers (as a cohort): Number of customers at start of period and number of customers at end of the period, in addition to the raw customer transactional information.

Originally posted on Wed, Jan 09, 2013 @ 06:49 AM

Photo by Riccardo Annandale on Unsplash

Categories:

#### No responses yet

This site uses Akismet to reduce spam. Learn how your comment data is processed.