Customer & Loan Portfolio Analysis

Context

SHBFinance focuses on providing unsecured consumer loans to low-and middle-income customers.

Despite building a large customer base with millions of borrowers and a nationwide workforce, the company was significantly affected by COVID-19, which reshaped customer demographics and altered loan portfolio characteristics.

In this shifting market environment, analyzing customer behavior, loan features, and year-over-year performance is critical to support sustainable business growth.

Request

Identify SHBFinance’s customer profiles

Define the characteristics of loans disbursed

Explore the correlation between customer profiles and loan products

Compare lending performance across years

Provide strategic recommendations to enhance loan origination and sales

1. Data Cleaning:

Almost all data is encoded, and most of it is represented as numbers, regardless of the meaning of the column.

2. Define SHBFinance's customer profile:

Demographics

Strong presence among young to middle-aged adults (70% from 25-44), the majority is marital status 2 (>50%)

Strong skew towards the lower-middle-income group (15.8M VND on average, 4.7% below 5M VND)

Education: More than 50% of customers have a level 1 education

Working experience: A significant portion are in the early stage of their careers (mostly 0-5 years of experience)

Behavior

Although the customers are diverse, most of them still have quite similar work backgrounds

• 99.9% of income comes from resource 1 and 5, while 90% of labor contracts are type 4
• 2/10 different job types of customers (type 6 and 9) accounted for more than 60%

Most borrowers are not burdened, > 1/3 of borrowers may have a wife and children or elderly parents (1-2 dependants)

Business situation

There are still some provinces that do not have many loan contracts (13/63 provinces under 1000, 6 of which have under 5 contracts)

Based on the information that customers provided company information, we can see:

• Most of customer does not want to disclose company information, current residence or does not have a job
• Two areas adjacent to SHB's key branches (HCMC, Binh Duong) and major cities (Da Nang) had high number of contracts (>15K)
• Beside rural or underdeveloped areas, Hai Phong presents potential for growth due to its urban setting

Assumption: company_address_province is encoded from the first ID number of a province. (to synchronize with permanent column information)

3. Identify the characteristics of the loans borrowed:

Loan behavior

Customers mainly borrow for shopping purposes, buying cars or paying for accommodation

They only borrow a small amount for various consumption purposes (<6M VND)

Customers mainly prefer medium and long term loans rather than short term loans

Loans for education, medical or travel are short-term. Besides, accommodation and vehicles are long-term loans (~2 years)

Product

Among 19 loan product categories, categories 0, 1, and 2 account for the majority (20.54% product 0, 33.54% product 1 and 20.45% product 2)

These can be products that serve quick daily payments: electricity, water, medicine, car repair, clothes, buy a bus ticket,... (<400K VND)

In particular, these products only appear in Business Line 1 - a business line where the Loan Amount is mostly < 1M VND

-> Loans are mostly unsecured, with low limits that vary between purposes

Other features

Monthly Interest: Ranging from 0.1-3.2%, Healthcare's average interest rate is the highest.

Has Insurance: Only 0.01% of contracts are uninsured (assume null = Unknown), interest rates for insurance range from 0.5-3.75%. Those without insurance only borrow at an interest rate of 0.2-0.3%.

Disbursement Rate: Almost achieve at 85%, loan disbursed through channel 3 achieve a 100% rate, but with a small number of contracts

Liquidity: Evenly distributed across products, products type 16,17,18 exhibit lower liquidity compare to others.

4. Correlation between customer profiles and loans:

Utilised debt risk metrics for K-Means clustering

Using some metrics that will be become features for K-Means

Results from K-Means suggest 3 distinct customer segments

3 debt risk groups named according to their characteristics

Customer demographics after using KMeans

(Some aspects is average value)

High Risk Group

% of users: 1.9%

Age: 30

Income: 3M

Dependants: > 1

Working Year: 2.8

DTI: 5.8%

Average Risk Group

% of users: 52.4%

Average age: 35

Income: 15M

Dependants: 0-2

Working Year: 5.4

DTI: 1.06%

Low Risk Group

% of users: 45.7%

Age: 37

Income: 17M

Dependants: 0-1

Working Year: 6.1

DTI: 0.87%

Hypothesis

High Risk Group: Low-income borrowers with high debt burdens and many dependents are likely to have unstable employment, indicating significant financial strain and an increased risk of default.

Average Risk Group: Moderate-income borrowers with moderate debt burdens and few dependents indicate potential financial responsibilities, suggesting a need for careful financial planning.

Low Risk Group: Older, higher-income borrowers with lower debt burdens, fewer dependents, and many years of senior experience in the industry suggest greater financial stability and a lower risk of default.

=> When the amount of money customers need to pay in a month is a large percentage (%) of what the amount they need to spend on their living, they bring more risk to SHBFinance

-> The high-risk group has VERY LOW INCOME but often needs to support more than one person in their family, and may spend 15% of their income on debt. To reduce risk, offering longer loan terms with smaller payments and advice on better benefits would help.

5. Compare loan effectiveness over time:

The lending data covers two years (2022–2023). Comparing SHBFinance’s loan disbursements, we observe:

In 2023, the number of loans at SHBFinance increased by nearly 15% compared to 2022

Seasonal pattern in customer borrowing behavior: at the beginning of the year (especially around Tet), customers tend to avoid taking loans – reflecting the traditional mindset of Vietnamese people to avoid debt at the start of the year

Loan volumes steadily increased month by month, peaking towards the end of the year

In September–October 2023, thanks to the partnership with Zalopay, loan disbursements reached the highest level of the year – demonstrating the effectiveness of such collaborations

In addition, the average customer age in 2023 is likely older compared to 2022, especially in two loan purposes: "Healthcare" and "Bills"

The high risk group appears in the older age group (> 40) with a very low percentage (0.5%), while the low risk group makes up the majority (70%). Because the average income of this group is 12,264,000

Hypothesis: Despite Vietnam recovering from COVID-19, many older people with pre-existing health conditions are still dealing with long-term health issues. This has increased their medical expenses and forced them to borrow money.

6. Recommendation:

Slide proposal: Require 1 & 2

Slide proposal: Require 3