# CLVM - 客戶終生價值模型

###### tags: `BA` `Project`

## CLVM - Background
Customer Lifetime Value Models are typically used by leading banks to provide a more "quantitatively informed" basis for making strategic decisions.

* **Customer segmentation and proposition development**:
1. **Potential**: beyond segmentation based on typical "current value" metrics of AUM or revenue contribution, lifetime value provides a proxy of potential.
2. **Informed characteristics**: overlay of behavioral/ propensity dimensions provides a richer characterization of customer segments to inform the development of strategic propositions to tailored to acquire target customer segment.

* **Customer acquisition and investment decisions**: Some levels of quantification of customer future/ potential value contribution provides a more informed basis for making investment decisions.
1. Upfront investment decision (Ex: Branding) to acquire target customers
2. Ongoing investments to service customers and deepen relationship
3. Quantitative of key financial metrics e.g. payback period, ROI, NPV

* **Customer service model and resource allocation decisions**: customer lifetime value, when combined with the current level of contribution, facilitates the optimization of service model and resource allocation
1. Enables differentiation of service model (e.g. high touch, pure digital) based on not just value contribution today, but also potentially in future.
2. Applies also to allocation of high cost resources such as RMs/ FAs.


In the above-mentioned case, with the estimation of Customer Lifetime Value, we can acquire "potential young customers" at lower cost and facilitate future value.

By reviewing future potention and current contribution, we can segment customers and apply better resources allocation strategy.

## CLVM: benefits and achievements
1. Transfer the marketing strategy from "Product Push" to "Customer Center" perspective
2. Developed the CLVM model and transferred both methodology and techniques to Cathay Groups
3. Segmented customers and generated business insights, further supporting strategic decisions

## Customer Lifetime Value - Methodology

CLVM is built on three models, revenue potential model, long-term propensity model, and personalized margin, combined to calculate customer's future revenues.


* **Assumptions and limitations of the model**:
1. Use three-year historical data to predict customer's lifetime value (till age 80), entailed uncertainty.
2. Estimate attrition rate and acquisition rate with groups of customers.
3. Potential of customers will be fully realized
4. Conditions within and without Cathay will remain constant or similar
5. CLVM is a long-term predictive model, with characteristics of hard to verify and not suitable to short-term marketing strategy

reference: [https://blog.hubspot.com/service/how-to-calculate-customer-lifetime-value](https://)

1. Revenue Potential Model: creating customer segmentation
* Differentiate CC customers and non-CC customers because of data-richness discrepency
* Grouping customers with similar behavioral characteristics.
* Describing the revenue variations within the different clusters

2. Long-Term propensity Model: Using Partial Least Squares Regression (PLSR) to predict long-term propensity on different products.
* PLSR can handle lots of features
* PLSR will first do dimension deduction and generate several feature-components, and then use those components to build predictive model
* Key difference with PCA: while PCA only considers variation between features; PLSR integrate the variations both in features and target.
3. Personalized Profit Margin: