This algorithm results in clusters of customers, based on their actual and predicted spending levels, to determine their importance to your company. It also predicts the following for each customer:
- Whether they will still be your customer one year from now.
- How many transactions can be expected from this customer in the next X weeks?
- Is this customer likely to churn?
- Will this customer be among those who will provide the most value to the company going forward?
Description
CLV input data and output
| Input data | Output |
| Transactional data
Customer_id. Date. Order_id. Order_line_id. Amount. Discount. |
Churn probability. Expected amount. Actual spending segment. Future spending segment.
|
Details and output examples
The CLV algorithm is based on a complex statistical model that employs the inter-purchase time and purchase regularity. The model also takes the ‘frequency paradox’ into account, meaning that after each high frequency purchase cycle, there is a period of inactivity.
The model computes:
- Churn probability.
- Expected spending.
- Actual spending segment.
- Future spending segment.
Example of the output
| Customer ID | 11223344 |
| Churn probability |
0.21 |
| Expected amount |
€140.23 |
| Actual spending segment |
Silver |
| Future spending segment |
Silver-bronze |
Tags:
- Statistics-based model.
- Predictive.
- Clustering.
Previous page: Dynamic Customer Engagement | Next page: Introduction