DexLab Analytics

One forecast isn't enough for ECL. Illustrative: base case alone = ₹24,000. Weighting base, upside and downside = ₹32,800 (+37%). RBI's ECL framework for bank:dexlabanalytics.com/blog/mac… #IFRS9 #CreditRisk

10 lakh loan. 2% PD. 40% LGD. Expected loss: ₹8,000. That's EL = PD × LGD × EAD. Our walkthrough covers a portfolio example, Basel vs IFRS 9 and a Python snippet: dexlabanalytics.com/blog/exp… #CreditRisk #IFRS9

Expected loss vs stress test: what's the difference? 🤔 Expected loss = what a bank expects to lose Stress test = what it could lose in a severe recession, and whether it survives. dexlabanalytics.com/blog/str…

Interview question: what's the difference between a credit rating and a credit score? "One is for companies, one is for people" is true but weak. A strong answer covers internal rating models, scorecards, master scales and IFRS 9 staging. Breakdown:dexlabanalytics.com/blog/cre…

From 1 April 2027, Indian banks provision on expected credit loss. The engine behind it: Expected Loss = PD × LGD × EAD A ₹8 cr exposure, 2% PD, 40% LGD: ₹6.4 lakh. Downturn (5% PD, 50% LGD): ₹20 lakh. Same loan. More than 3× the loss.dexlabanalytics.com/blog/cre…

After the ECB reviewed banks' internal models (2016–2021), risk-weighted assets for the models it examined rose about 12%, or roughly €275bn. Credit risk models drive real capital decisions. The 5 main types, explained 👇dexlabanalytics.com/blog/wha… #CreditRisk #creditrisktraining

Your credit model just declined a loan. Can it explain why? "The algorithm said no" isn't an answer. SHAP turns ML predictions into reason codes, and RBI's June 2026 draft on model risk is raising the bar. Full guide: dexlabanalytics.com/blog/mac… #CreditRisk #MachineLearning

Two-thirds of counterparty credit risk losses in the 2007–09 crisis came from CVA, not defaults (Basel Committee). The counterparty was still paying. The losses came anyway. Our guide to SA-CCR, CVA and more: dexlabanalytics.com/blog/cou…

RBI requires Indian banks to hold 9% CRAR — a full point above Basel III's 8% global floor. Add the 2.5% buffer, and the real floor is 11.5%. We broke down RBI's entire credit risk framework, including what changed in April 2026: dexlabanalytics.com/blog/how… #CreditRisk #RBI

An AUC above 0.95 should worry you, not impress you — it usually means target leakage. Real credit risk model validation tests 4 things: discrimination (AUC/Gini/KS), calibration, population stability (PSI), and backtesting. Full breakdown: dexlabanalytics.com/blog/cre…

Why do two loans with identical payment histories sometimes land in different IFRS 9 stages? It's not about missed payments. SICR is about whether default risk has risen — a forward-looking call, not a backward-looking one.dexlabanalytics.com/blog/wha… #ECL #creditrisk #SICR

Basel III credit risk rules just got a real deadline in India: April 1, 2027. RBI's 2026 Directions rewrite risk weights across sovereigns, banks, corporates, MSMEs, retail & real estate. Here's what's actually changing: dexlabanalytics.com/blog/bas… #creditrisk #riskmodeling #basel

Diversifying across 1,000 borrowers doesn't make a loan book safe. Neither does staying under every single-name limit. IL&FS proved it in 2018 — concentration hid inside a book that looked diversified on paper. Correlation is the blind spot. Full story:dexlabanalytics.com/blog/cre…

India's bank GNPA ratio: 1.8% (Mar 2026) In 2018, it was 11.46%. One of the fastest bank deleveraging cycles any major economy has seen-per RBI's own data. Breakdown by bank type, and where the next risk is building 👇 dexlabanalytics.com/blog/cre… #CreditRisk #RBI #BankingIndia

When you model Exposure at Default on a revolving facility, are you using the drawn balance — or the modeled exposure? If it's the balance, you're probably underestimating risk. dexlabanalytics.com/blog/wha… #CreditRisk #ExposureAtDefault #RiskManagement

EAD isn't your outstanding balance — it's what you're exposed to if a borrower draws down the rest of their limit right before defaulting. That's what the Credit Conversion Factor models.Full breakdown:dexlabanalytics.com/blog/exp… #creditrisk #creditriskmodeling #expectedcreditloss

Two loans. Same default probability. One has 10% LGD, the other 70%. Same risk? Not even close. PD tells you if it defaults. LGD tells you what it costs. Here's why that distinction runs Basel III and IFRS 9 provisioning 🔗dexlabanalytics.com/blog/why… #CreditRiskModeling #FRM

What does a borrower’s default really cost a lender? Not the entire outstanding amount.LGD measures the portion of exposure that remains lost after recoveries. Explore LGD modeling, recovery rates, and practical estimation methods.dexlabanalytics.com/blog/lgd… #creditriskmodeling

Freedom inspires. Knowledge transforms. India progresses. Happy 80th Independence Day 📷 #IndependenceDay #dexlabanalytics

Did you know that google uses something called "Web Crawlers" also known as "spiderbot" to index data throughout the internet and make web searching faster? #DataScience #MEMES #MachineLearning