← Back to portfolio Interactive Dashboard · Credit Risk Analytics

Credit Portfolio Risk Dashboard

An interactive analysis of 2.26 million consumer loans originated between 2007 and 2018, $34B in funded principal, cut by grade, vintage, geography, and borrower profile to show where credit risk actually concentrates.

Same Lending Club universe that backs the Credit Default Risk Model, here rolled up locally in Python and rendered as a live, fully interactive dashboard. Hover any element; drag the time slider; read the maps.

2,260,668 loans $34.0B funded 2007 to 2018 51 states Python · pandas Apache ECharts
The dashboard data could not load. If you are viewing this from a file:// path, serve the folder over HTTP (the page fetches a JSON roll-up from the same origin).
$0
Total funded
0
Loans originated
0%
Portfolio default rate
0%
Weighted avg rate
0
Avg borrower DTI
$0
Charged-off principal
01 · The risk gradient

Risk rises cleanly with loan grade

Lending Club's letter grade is how it prices risk. Default rate climbs monotonically from grade A to G, and the interest rate charged tracks it, the core risk-return relationship a lender is paid to get right.

Default rate and interest rate by grade

Bars: realized default rate. Line: weighted average interest rate.

The sub-grade risk curve

All 35 sub-grades, A1 to G5. A smooth, accelerating climb in default rate.

02 · Risk and return

Is the portfolio paid for the risk it takes?

Each sub-grade plotted by the rate it charges against the default rate it realizes, sized by funded principal. The treemap shows where the capital actually sits, colored by default rate.

Risk-return frontier

x: avg interest rate · y: default rate · bubble size: funded · color: grade.

Where the capital sits

Funded principal by grade and sub-grade, shaded by default rate.

03 · Vintage

Originations and credit performance over time

Monthly origination volume against the default rate of each vintage. Drag the slider to focus a window. The platform scaled fast; performance moved with the credit cycle and underwriting changes.

Monthly originations and default rate, 2007 to 2018

Area: originations ($M). Line: default rate of loans issued that month.

04 · Geography

The geography of credit risk

Two reads on the same map: a choropleth of default rate by state, and a bubble map sizing each state by origination volume while coloring it by risk. Volume concentrates in different states than risk does.

Default rate by state

Choropleth. Darker red is higher realized default.

Volume and risk by state

Bubble size: funded principal. Color: default rate.

05 · Concentration

Where losses concentrate

Default rate as a heatmap across two cuts: grade against loan purpose, and grade against origination year. The darkest cells show which combinations a lender should price hardest or avoid.

Default rate: grade × purpose

Cells with fewer than 50 loans are left blank.

Vintage heatmap: grade × year

How each grade performed by year of origination.

06 · Borrower factors

What the borrower profile tells you

Default rate sliced by loan purpose, leverage, income, housing, term, and income verification. Leverage and term carry real signal; some intuitive cuts carry less than you'd expect.

Default rate by loan purpose

Ranked. Small-business and renewable-energy lending sit at the risky end.

By leverage (DTI)

By income band

By home ownership

By term

By income verification

Methodology & data

Source: Lending Club public loan data, 2,260,668 loans originated 2007 to 2018, via Kaggle. The same loan universe that backs the Credit Default Risk Model.

Definition: a loan is counted as defaulted when its status is Charged Off or Default. Default rate is defaulted loans over total loans in the group. Weighted average rate weights interest rate by funded principal.

Pipeline: the raw 1.2GB file is cleaned and rolled up locally in Python and pandas into a compact JSON of aggregates and cross-tabs (grade, sub-grade, purpose, state, vintage, DTI, income, housing, term, verification). The page renders that roll-up client-side with Apache ECharts. No row-level data leaves the browser; everything is self-hosted under a strict content-security policy.

PythonpandasApache ECharts Choropleth and geo-bubble mapsHeatmapsVintage analysisRisk-return