Go-To-Market Strategy for a Banking Firm to expand their Mortgage Loan Operations: Using clustering algorithms to identify target customers
A mortgage loan is a long-term loan used to purchase real estate, provided by primary lenders such as banks and credit unions. The loan is secured by the property, and borrowers make monthly payments consisting of principal and interest. If the borrower defaults, the lender can take possession and sell the property to pay off the loan. The secondary mortgage market is where loans are bought and sold between lenders and investors, with the government-sponsored enterprise Fannie Mae being a major purchaser.
This project focuses on devising a Go-To-Market strategy for a bank that wants to expand its mortgage operations into Nebraska and has provided a dataset of conventional mortgages for analysis.
“The goal is to maximize the dollar value of mortgages issued while limiting defaulting loans, with additional considerations for changes in interest rates, the economy, and execution strategy.”
About the data
The dataset 270K records of mortgages acquired from 2000 to 2021. The mortgages have varying terms between 5 and 30 years, with around 195K matured or prepaid, 70K active, and some in delinquent status. The dataset also includes property details, original borrowing amount, credit scores, and delinquency statuses. Details of sales due to foreclosures are also included.
Business Objectives and Process Flow:
Market Analysis:
Assessing characteristic of defaulting loans based on trends in DTI, Credit Score and LTV:
- A higher DTI ratio can signify higher chances of default. 35% of the foreclosed loans and 32% of delinquent loans had DTI >=43%
- A credit score of less than 700 can signify higher chances of default. 56% of the foreclosed loans and 36% of delinquent loans had CS <700
- Higher LTV can again be associated to higher chances of default. Loans with LTV >90% contribute to 40% of foreclosed and 44% of delinquent loans.
“Market is extremely sensitive to a few KPIs”
- No. of loans (demand) vary continuously showing changing trends in the housing market
- Rate of Interest adjusts according to the economy and inflation
- In the years building up to the recession period, we can observe lax DTI ratios, and relatively low credit scores.
- In the recovery and expansion phase of the economic cycle we can observe stricter norms for DTI, LTV and Credit Scores.
- The various ups and downs in the time series may reflect varied house price movements over time and changing macro-economic scenarios, as well as stricter regulations post recession
How are competitors handling the Nebraska Mortgage Market?
- Wells Fargo holds the majority of market share and has a moderate risk appetite
- U.S. Bank ranking at 2nd position appears to make a risky play having: low Credit Scores, Extremely high LTV and Highest delinquency rate
- FHLB Chicago is highly risk averse and consequently has the least market share in the lot and least revenue to loaned amount (3.27%) even though it has the least delinquency rate (0.2%)
- Penny Mac and Quicken Loans both share almost equal market share and are moderately risk averse.
Market Segmentation:
“Market Segmentation would help us understand the right set of customers that needs to be targeted”
K-means clustering can be used in market segmentation for a mortgage loans market to group similar customers based on their characteristics and behavior, allowing lenders to tailor their marketing strategies and loan products to specific customer segments for improved customer acquisition and retention.
Using the Kmeans algorithm to cluster mortgage loan applicants based on features such as principle amount, interest rate, DTI, credit score, and others, can lead to market segmentation into 6 distinct groups. This approach can help mortgage lenders better understand their customer base and tailor their offerings to specific segments to improve customer satisfaction and increase revenue.
Creating comparative archetypes :
Creating financial archetypes involves grouping similar borrowers into clusters based on their financial characteristics. Two key factors to consider are the revenue potential and the risk of default. Revenue potential is determined by the principal amount and the interest rate of the loan. Risk of default is a function of several variables such as credit score, delinquency metric, loss metric, and original combined loan-to-value (OCLTV). By comparing the different clusters along these two dimensions, financial institutions can gain insights into the market segmentation of the mortgage loans market and adjust their lending practices accordingly. This approach can help lenders minimize risk while maximizing revenue potential.
RECOMMENDATION: PROFILING SEGMENTS TO UNDERSTAND KEY MARKET SEGMENTS FOR THE BANK TO FOCUS ON WHILE ACCEPTING LOANS
The kmeans clustering algorithm was used to analyze mortgage data and identify six distinct customer segments. The first segment is considered the safest and most profitable market for the bank to enter, with high revenue potential and low risk of default. The second segment is also safe but less profitable, while the third segment is characterized as high risk but high reward, meaning it may have some temporary delinquencies but offers a profitable market for the bank.
The fourth segment is deemed less safe and least rewarding, while the fifth and sixth segments are highly risky and lossy markets to avoid. These segments are characterized by very low credit scores, high DTI, and OCLTV values. They may seem profitable due to high ROI and longer payment terms, but there is a higher probability of default with more delinquencies.
By identifying these six distinct customer segments, the bank can better understand the market and tailor its strategies accordingly. For example, the bank could focus on marketing to the safest and most profitable segment, while also considering the potential risks and rewards of entering the high-risk-high-reward segment. At the same time, the bank can avoid markets that are highly risky and lossy, thus minimizing the likelihood of default and protecting its investments.
CHOICE OF FOCUS-SEGMENT CAN BE DIFFERENT FOR DIFFERENT GEOGRAPHIES IN NEBRASKA
Distribution of Nebraska’s mortgage business across the geography:
- In the above map, revenue is plotted across Nebraska State based on Fannie Mae’s data
- We can clearly see that more than 75% of the business is condensed in just 2–3 zip3 areas
- Bank can focus on this area to have more lucrative plans than other areas
- This area is also a strategic location for bank to start their head office for Nebraska region
Distribution of historic vulnerability of delinquency
- In the above map, delinquency metric* is plotted across Nebraska State based on Fannie Mae’s data
- Urban areas near Lincoln and Fremont have high delinquencies but are also high on revenue, we can try to optimize ‘High-risk-High-reward’ segment here
- The western and central Nebraska region is mostly less populated region with a low revenue but still has high delinquency
- We must make sure we target only safe and high reward segments only and avoid risky segment
Note: Normalized Delinquency Metric =Σ Unpaid Balance in n days ∗n days; This will create a singular delinquency metric where loans wit longer delinquencies will have a higher value according to their unpaid principal balance
The above-mentioned Go-To-Market Strategy would guide about different market segments available at Nebraska for the bank to enter / expand their secondary mortgage loan business. It will help them identify what are the key indicators that suggest high default, what metrics to be considered before purchasing any secondary mortgage loans. The geographic analysis will help them find pain points and strategic locations which needs tailored targeting.
