Bank Rate Monitor

Bankrate was founded in 1976 as Bank Rate Monitor, a print publisher for the
banking industry. Our experience has fueled our reputation as the premier
financial authority. When you visit, the reviews, guides and
educational content have been developed by leading personal finance
experts. Bankrate’s product comparison tools, calculators and educational
content help over 100 million consumers make smarter financial decisions
each year. No matter where you are in your financial journey, Bankrate can
help you reach your goals.
One area Bankrate specializes in is personal loans. Bankrate’s goal is to help
match the best lending partner for each individual customer. To do that, we
collect information about the customer to understand who would be the best
fit. We want to maximize the likelihood someone will be approved for the loan
at the best “deal” possible.
In this case, let’s imagine we have three lending partners, “A”, “B”, and “C”. In
this scenario, let’s imagine that each lender is providing the same offer.
However, for each approval, lender A provides us $250, lender B pays us
$350, and lender C pays us $150.
For 100K of our customers, we have collected a set of data in hopes to
understand what variables are most important in determining approvability.
We believe if we collect the right information, we will be able to show the
appropriate lender to each customer to maximize approval rate and revenue.
The team knows there are many possible data points that we could collect to
help us better understand the likelihood of someone approving for the loan,
but they started with this list for now.
This data has been recently collected from the website and the leadership
team has asked you to help make sense of it. They want to know if they can
increase revenue per application by matching certain groups of customers
to specific lenders.
There are three broad categories that they would like you to investigate.
These questions should be used as thought starters – please be creative with
how you analyze and explore each question, leveraging relevant statistics
and visualizations where applicable.
If the team reviews your assessment and decides to move forward, the final
step in the interview process would include the opportunity to talk through
your assessment live with an interviewer for 45 minutes. Please be prepared
at that time to showcase your results using the tool of your choice and to
discuss your findings during the interview session.
1. Explore the variables relationship with approvability:
a. Possible things to consider: Which variables are the most helpful
in understanding if a customer is going to be approved or denied
for a loan? Are there certain variables that are not useful to
collect? Are there any feature modifications or transformations
that would improve the predictive power of a variable?
2. Tell us about the lenders approval rates:
a. Possible things to consider: What is each Lender’s average
approval rate? Are there any clear differences between the three
different lenders on what type of customers they approve? Are
there variables that reliably predict a user’s approval likelihood for
a particular lender?
3. Calculate how much incremental revenue Bankrate could make if they
more effectively matched lenders and customers
a. Possible things to consider: What levers do you have to increase
revenue per application? Are there groups of customers that
would provide better economic outcomes for Bankrate if they
applied to a different lender than the one they applied to? What
considerations should we have in mind if we planned to match
customers with lenders in real time? Do we create more or less
approvals by the proposed changes?
You will have 24 hours to complete this assignment – but note that most
submissions are completed in 4 hours. Feel free to use any tool to answer
these questions. Please submit both your answers and your work used to get
to those answers for the three categories in an exported file type that our
assessment reviewers can easily access (PDF, HTML, Excel, Screenshots,
Powerpoint). Make sure that your work is legible and able to be referenced for
your answers.
Data Dict:
User ID – unique ID to represent the customer who submitted the application
Application – count of application – each row is 1 application submitted (int)
Reason – the purpose for requesting a loan (string)
Loan Amount – the loan value size (int)
FICO Score – score used to make credit risk decisions (int)
FICO Score Group – categorical grouping of FICO scores in the five common
buckets (string)
Employment Status – current employment designation for the customer
applying for the loan (string)
Employment Sector – categorical representation of the industry the customer
works in (string)
Monthly Gross Income – the amount of money the user makes pre-taxes and
deductions (int)
Monthly Housing Payment – how much a month do they pay in housing
costs (int)
Ever Bankrupt or Foreclose – has the customer ever had to file for
bankruptcy or had a foreclosure on their house (bool)
Loaner – the lending partner (string)
Approved – did the customer get approved for the loan or not (bool)
Bounty – how much did we get paid for the application – only occurs on an
approval (int)
Please feel free to create new columns or variable groupings to aid in your


Personal_Loans_-_Pre-Super_Day_Case_Prompt (1)