🛡️ DriftBreaker

Credit Risk Analytics Platform

📋 Portfolio Summary

2,257,919

Total Loans

$33.99B

Total Exposure

$4.43B

Total Default Volume

12.56%

Default Rate (Count)

13.05%

Default Rate ($)

By Segment

Segment	Loans	Exposure ($M)	Default Rate (count)	Default Rate ($)
high_risk	188,630	$3,405.0M	30.38%	31.53%
low_risk	1,096,132	$15,724.9M	6.47%	6.19%
medium_risk	973,157	$14,861.4M	15.96%	16.07%

By Term

Term	Loans	Exposure ($M)	Default Rate
36 months	1,607,316	$20,495.7M	10.70%
60 months	650,603	$13,495.6M	17.15%

Timing Metrics

PD12: 5.13%
PD24: 9.78%
Peak Hazard Month: Month 12
Median Months to Default: 14.0

Date Coverage

First Vintage Year: 2007
Last Vintage Year: 2018
Year with Lowest Default Rate: 2018 (2.72%)
Year with Highest Default Rate: 2015 (18.25%)

Key Observation: Highest risk segment: high_risk (30.38% default rate)

📈 Originations vs Defaults

Annual Metrics

Year	Loans	Volume ($M)	Defaults	Default $ ($M)	Default Rate	Default Rate ($)
2007	251.0	$2.2M	45.0	$0.5M	17.93%	22.01%
2008	1,562.0	$14.4M	247.0	$2.7M	15.81%	19.04%
2009	4,716.0	$46.4M	594.0	$6.3M	12.60%	13.47%
2010	11,536.0	$122.1M	1,487.0	$16.0M	12.89%	13.13%
2011	21,721.0	$261.7M	3,297.0	$43.4M	15.18%	16.57%
2012	53,367.0	$718.4M	8,644.0	$127.2M	16.20%	17.71%
2013	134,814.0	$1,982.8M	21,030.0	$330.0M	15.60%	16.64%
2014	235,629.0	$3,503.8M	41,408.0	$645.7M	17.57%	18.43%
2015	421,095.0	$6,417.6M	76,851.0	$1,212.9M	18.25%	18.90%
2016	434,407.0	$6,400.6M	71,666.0	$1,105.9M	16.50%	17.28%
2017	443,579.0	$6,585.0M	44,854.0	$712.9M	10.11%	10.83%
2018	495,242.0	$7,936.3M	13,460.0	$231.6M	2.72%	2.92%

Cumulative Metrics

Year	Cum. Loans	Cum. Volume ($B)	Cum. Defaults	Cum. Default $ ($B)	Cum. Default Rate	Cum. Default Rate ($)
2007	251.0	$0.00B	45.0	$0.00B	17.93%	22.01%
2008	1,813.0	$0.02B	292.0	$0.00B	16.11%	19.43%
2009	6,529.0	$0.06B	886.0	$0.01B	13.57%	15.04%
2010	18,065.0	$0.19B	2,373.0	$0.03B	13.14%	13.78%
2011	39,786.0	$0.45B	5,670.0	$0.07B	14.25%	15.41%
2012	93,153.0	$1.17B	14,314.0	$0.20B	15.37%	16.83%
2013	227,967.0	$3.15B	35,344.0	$0.53B	15.50%	16.71%
2014	463,596.0	$6.65B	76,752.0	$1.17B	16.56%	17.62%
2015	884,691.0	$13.07B	153,603.0	$2.38B	17.36%	18.25%
2016	1,319,098.0	$19.47B	225,269.0	$3.49B	17.08%	17.93%
2017	1,762,677.0	$26.06B	270,123.0	$4.20B	15.32%	16.13%
2018	2,257,919.0	$33.99B	283,583.0	$4.43B	12.56%	13.05%

                Summary: Total Book: $33.99B | Total Defaults: $4.43B | Overall Default Rate: 12.56%
            

⚡ Early Defaults Analysis

Early defaults are defined as loans that defaulted within 12 months of origination.

Early Defaults by Vintage Year

Vintage	Total Loans	Total Defaults	Early Defaults	Default Rate	Early Default Rate
2007	251	45	11	17.93%	4.38%
2008	1,562	247	91	15.81%	5.83%
2009	4,716	594	251	12.60%	5.32%
2010	11,536	1,487	488	12.89%	4.23%
2011	21,721	3,297	959	15.18%	4.42%
2012	53,367	8,644	2,886	16.20%	5.41%
2013	134,814	21,030	6,358	15.60%	4.72%
2014	235,629	41,408	12,262	17.57%	5.20%
2015	421,095	76,851	22,869	18.25%	5.43%
2016	434,407	71,666	28,997	16.50%	6.68%
2017	443,579	44,854	27,829	10.11%	6.27%
2018	495,242	13,460	12,732	2.72%	2.57%

Early Defaults by Segment

Segment	Total Loans	Total Defaults	Early Defaults	Default Rate	Early Default Rate
high_risk	188,630	57,315	25,722	30.38%	13.64%
low_risk	1,096,132	70,923	25,826	6.47%	2.36%
medium_risk	973,157	155,345	64,185	15.96%	6.60%

📊 Book Composition Analysis

Annual Composition by Count

Year	Low Risk %	Medium Risk %	High Risk %
2007	47.01%	44.62%	8.37%
2008	51.34%	42.25%	6.40%
2009	53.92%	39.23%	6.85%
2010	54.94%	35.33%	9.73%
2011	56.71%	31.02%	12.26%
2012	55.11%	35.97%	8.92%
2013	45.84%	43.54%	10.62%
2014	41.61%	46.50%	11.90%
2015	45.34%	43.51%	11.15%
2016	47.27%	44.05%	8.68%
2017	47.78%	45.49%	6.73%
2018	55.84%	39.56%	4.60%

Annual Composition by Volume

Year	Low Risk %	Medium Risk %	High Risk %
2007	40.01%	46.76%	13.23%
2008	46.61%	43.67%	9.72%
2009	50.50%	40.70%	8.80%
2010	52.00%	35.34%	12.67%
2011	48.27%	32.64%	19.09%
2012	47.49%	37.20%	15.31%
2013	43.71%	42.91%	13.38%
2014	39.03%	47.18%	13.79%
2015	42.91%	43.27%	13.82%
2016	44.00%	45.17%	10.83%
2017	44.60%	47.33%	8.07%
2018	55.72%	39.70%	4.58%

                Shift Analysis: low_risk grew from 40.01% to 55.72% (+15.71pp) | 
                medium_risk shrank from 46.76% to 39.70% (-7.06pp) | 
                high_risk shrank from 13.23% to 4.58% (-8.65pp)
            

📉 Cumulative Default Curves Analysis

This analysis shows the cumulative probability of default (PD) over time for each vintage.

Key PD Milestones by Vintage

Vintage	PD6	PD12	PD18	PD24	PD30	PD36	Ultimate PD
2007	0.00%	4.38%	7.97%	11.95%	15.94%	17.13%	17.93%
2008	1.92%	5.83%	9.41%	13.12%	14.66%	15.62%	15.81%
2009	2.25%	5.32%	7.40%	9.69%	11.13%	12.07%	12.60%
2010	1.42%	4.23%	6.14%	8.01%	9.86%	11.14%	12.89%
2011	1.66%	4.42%	7.30%	9.61%	11.90%	13.13%	15.18%
2012	1.82%	5.41%	8.38%	11.40%	13.63%	15.07%	16.20%
2013	1.46%	4.72%	7.55%	10.59%	12.66%	14.01%	15.60%
2014	1.56%	5.20%	8.27%	11.41%	14.27%	15.92%	17.57%
2015	1.72%	5.43%	9.61%	12.82%	15.60%	17.21%	18.25%
2016	2.16%	6.68%	10.37%	13.56%	15.53%	16.33%	16.50%
2017	2.15%	6.27%	8.91%	9.95%	9.98%	9.98%	10.11%
2018	1.46%	2.57%	2.59%	2.59%	2.59%	2.59%	2.72%

                Observations: Highest PD12: 2016 vintage (6.68%) | 
                Highest PD24: 2016 vintage (13.56%) | 
                Highest Ultimate PD: 2015 vintage (18.25%) | 
                Peak Hazard Month: Month 12
            

🔬 Survival Analysis

Default Rate by Segment

Segment	Loans	Defaults	Default Rate	Median Months to Default
high_risk	188,630	57,315	30.38%	13.0
low_risk	1,096,132	70,923	6.47%	16.0
medium_risk	973,157	155,345	15.96%	14.0

Default Rate by Vintage

Vintage	Loans	Defaults	Default Rate
2007	251	45	17.93%
2008	1,562	247	15.81%
2009	4,716	594	12.60%
2010	11,536	1,487	12.89%
2011	21,721	3,297	15.18%
2012	53,367	8,644	16.20%
2013	134,814	21,030	15.60%
2014	235,629	41,408	17.57%
2015	421,095	76,851	18.25%
2016	434,407	71,666	16.50%
2017	443,579	44,854	10.11%
2018	495,242	13,460	2.72%

Default Timing Distribution (Hazard Proxy)

Month Bucket	Default Count	% of Defaults
1-3	11,859	4.18%
4-6	28,837	10.17%
7-9	35,058	12.36%
10-12	39,964	14.09%
13-18	59,131	20.85%
19-24	45,914	16.19%
25-36	49,145	17.33%
37+	11,246	3.97%

Peak Hazard Bucket: 13-18 months (59,131 defaults, 20.85%)

Cumulative PD by Segment

Segment	PD6	PD12	PD18	PD24	PD36	Ultimate
low_risk	0.82%	2.41%	3.80%	4.99%	6.32%	6.47%
medium_risk	2.43%	6.72%	10.09%	12.64%	15.30%	15.96%
high_risk	5.52%	13.96%	19.84%	24.11%	28.69%	30.38%

📊 Macro Economic Analysis

Analysis of macroeconomic indicators and their correlation with default rates.

Macro Indicators Summary

Indicator	Mean	Min	Max	Correlation with Default
Unemployment	5.17%	3.70%	10.00%	0.084
Hy Spread	4.83	2.98	19.88	0.110
Yield Curve	1.70	-0.18	3.79	0.138
Consumer Sentiment	91.1	55.3	101.4	-0.054
Real Income	$14,107	$12,053	$15,506	-0.127

                Key Insights:

                • Positive correlations (red) indicate the indicator increases with default rates

                • Negative correlations (green) indicate the indicator decreases with default rates

                • Yield curve shows the strongest positive correlation with defaults

                • Real income shows a protective negative correlation

🔧 Model Training & Drift Detection

                Note: Model training requires significant memory resources. 
                For full model results including drift detection and P&L analysis, 
                please run the interactive Streamlit app: streamlit run driftbreaker.py
            

The model uses a discrete-time survival analysis approach with logistic regression to predict default probabilities and detect drift between training and test periods.

📊 Macro Economic Analysis

Analysis of macroeconomic indicators and their correlation with default rates.

Macro Indicators Summary

Indicator	Mean	Min	Max	Correlation with Default
Unemployment	5.17%	3.70%	10.00%	0.084
Hy Spread	4.83	2.98	19.88	0.110
Yield Curve	1.70	-0.18	3.79	0.138
Consumer Sentiment	91.1	55.3	101.4	-0.054
Real Income	$14,107	$12,053	$15,506	-0.127

                Key Insights:

                • Positive correlations (red) indicate the indicator increases with default rates

                • Negative correlations (green) indicate the indicator decreases with default rates

                • Yield curve shows the strongest positive correlation with defaults

                • Real income shows a protective negative correlation

🔧 Model Training & Drift Detection

                Note: Model training requires significant memory resources. 
                For full model results including drift detection and P&L analysis, 
                please run the interactive Streamlit app: streamlit run driftbreaker.py
            

The model uses a discrete-time survival analysis approach with logistic regression to predict default probabilities and detect drift between training and test periods.