Calculating, Predicting, and Acting On Customer Value

Customer Value (CV, or sometimes CLV or LTV) is a cornerstone of customer-centric marketing. In a nutshell, CV is a prediction of the net profit associated with a customer's relationship with a business over various future time periods. Don Peppers, who along with Martha Rogers is generally credited with starting the CRM revolution with the 1993 book The One to One Future, was fond of paraphrasing George Orwell, saying "All customers are equal, but some customers are more equal than others." The goal of CV in the Clario application is to reveal the customer behaviors which signal these value inequalities and empower the marketer to act on them.

Calculating Historical Customer Value

Clario begins by computing CV for multiple time periods subsequent to every purchase using a profit formula provided by our client, a best practices recommendation, or a little of both. This process takes place with every application refresh. For more recent purchases we are able to compute fewer CV time periods, for older purchases we are able to compute more CV time periods.

Predicting Customer Value

With the foundation of historical CV in place, we predict CV for the current cohort of customers.

  1. First, we replicate the query for the time period one year prior. In all likelihood these are different customers, but they exhibit the same query-configured customer, purchase, and marketing characteristics, and do so in a time window exhibiting the same seasonal characteristics. For these historical purchases, we have a complete set of data for each CV time period.
  2. The Clario application then uses a statistical technique known as Bayesian inference to predict CV for the cohort of customers who are the subject of the query. Bayesian inference updates the probability for a hypothesis as more evidence or information becomes available. This is a natural fit for our use case: for each additional time period containing observed (actual) subsequent purchasing behavior, we are more accurately able to predict CV. The prediction is influenced by historical CV observations for a similar cohort and calendar period.
  3. To improve the accuracy of the prediction, Clario further segments the current and historical cohorts based on two factors:
    • The speed of occurrence of a customer's initial purchase in the query window.
    • The speed of occurrence of a customer's first subsequent purchase.

For context, the image below represents the requisite CV computation for a single Clario query.

value prediction

While there are tradeoffs with any predictive modeling methodology, this approach makes a robust prediction of customer value dynamically for any query/cohort meeting a minimum threshold of statistical significance.

Acting on Customer Value

Finally, the predicted Customer Value, Customer Retention, and Revenue are presented in the application user interface for each time period. Again, all of these predicted metrics are for the current cohort of customers, and use both prior and current year data to predict up to 12-month subsequent CV and CV components (retention, revenue).

value tab

By default, metrics on the Value tab are compared to the historical (prior year) period. If a comparison query is selected, metrics on the Value tab are compared to the selected comparison query.

The knowledge of each cohort's predicted CV, coupled with the other insights revealed in the Clario query, empowers the business to: