Increase the profitability of acquisitions offer campaign using behavioural segmentation

Customers with good risk scores are extremely difficult to get as they have multiple offers from various loan providers and customers with bad risk scores bring the risk of NCL with them. Striking a balance between acquiring the mix of customers requires intelligent data-driven decisions and sound financial projections that are based on data.

In the end we want customers to pay on time without defaulting, and expand their credit needs with the organization over time as well. Identifying these customers requires intelligent, data-driven targeting, which cannot be achieved through mass campaigns. Given the high volume of data, decisions must be based on data science and advanced analytics to be optimal.

Failing to do so results in acquiring the wrong set of customers who may switch to other loan providers or default on their payments. This can lead to customers gaming the system for short-term acquisition benefits without developing brand loyalty. Over time, businesses increase their costs without achieving optimal returns.

This solution helps predict customer behavior post-acquisition, providing insights for a 2-5 year span. It optimizes acquisition costs by targeting the right segments, improving customer satisfaction, and ultimately enhancing long-term revenue.

Utilize bureau information of prospects to map “like” customers within your current customer base. Segregate your customer base into targeted personas based on their revenue potential and engagement with the bank, and use this information to predict the behavior of potential customers (prospects).

The categories can be broadly based on:

1. Customer profitability – 2-year/5-year revenue
2. Future potential – Off us balance/loan requirements, debt-to-income ratio
3. Customer purchase behavior – repayment patterns, risk scores
4. Customer contact behavior – digital propensity, contact frequency, customer sentiment

The segmentation approach adopts the principles of maximum separation between each category and high similarity within each category. This makes it easier for businesses to target customers in clusters while also allowing for broad-level customization. Depending on the requirements, the categories can be refined to increase customization.

This solution is designed for business leaders or teams who want to

1. Move from mass targeting to decision engine-based targeting
2. Want to refine their existing targeting framework by introducing enhanced customization

Increase conversion of offers: Customers are more inclined to respond to offers that feel personalized rather than generic. Detailed knowledge of customer preferences results in the creation of the right product features.

Indirectly impact customer loyalty: Tailored offers and services foster customer loyalty, giving the company a competitive edge over others.

Enhance business planning using data strategy: Utilize data intelligence to make informed decisions for customer acquisition, thereby reducing future losses.

Revenue: Boost short-term revenue with higher conversion rates for offers and increase long-term revenue by attracting customers who align well with the product.

Cost: Lower acquisition costs for campaigns through targeted efforts and reduce expected future losses based on P&L projections.

This solution can be constructed with both fundamental and sophisticated data sets.

Essential data requirements are:

1. Customer-level Profit and Loss (PnL) data
2. Bureau data

Customer segmentation can be achieved using various objective and non-objective algorithms. An objective approach segments customers based on a specific metric, whereas a non-objective approach uses a combination of metrics.

The choice of approach depends on the customer’s requirements. For identifying populations based on multiple metrics, a non-objective segmentation is appropriate. If a particular metric is of greater importance, an objective approach should be taken.

The approach will be tailored according to business needs and data complexity.

The most effective algorithms used are:

1. K-means clustering
2. Decision trees
3. Random forests

1. On-prem ( on customer systems or on DeepQ-AI Environment)
2. On hosted cloud space ( Customer or DeepQ-AI Environment)
   ( deployment is subject to data availability in the same environment, or feasibility of seamless data transfers within secured environments)

1. The final outputs may consist of lead lists with customer segment tagging and definitions.
2. Batch files, including Excel files or flat files.
3. Outputs can be delivered via API requests. By sending an API request with inputs, you can receive a response. This can be integrated into an existing application or a separate web application can be developed.