Learn / Basics

Credit risk analyst interview questions with answers, including SQL

By the Bucketwise team · 11 min read · Updated 13 November 2026

These are the questions worth being ready for if you are interviewing for a credit risk analyst role at a bank or an NBFC. The answers are short, in the way you would say them aloud. Where a topic needs more, the guide for it is named.

What basic credit risk questions are asked?

Question	A short answer
What is DPD and how is it calculated?	Days past due: the number of days since the oldest unpaid instalment fell due, with the due date counted as day 1. It runs from the oldest unpaid due date, and a later missed instalment does not start a second clock. The allocation rule and grace days change the result, so state them.
When does a loan become an NPA?	A term loan becomes an NPA when principal or interest is overdue for more than 90 days, which means 91 days or more. Cash credit and overdraft accounts use the out-of-order test.
What are the asset categories after a loan becomes an NPA?	Sub-standard, for an NPA of up to 12 months in the bank and middle-layer NBFC rules, then doubtful, then loss. Provisioning rises at each step.
What are SMA-0, SMA-1 and SMA-2?	Special mention accounts that are overdue but not yet NPAs: SMA-0 up to 30 days, SMA-1 more than 30 and up to 60, SMA-2 more than 60 and up to 90, for non-revolving loans.

Full guides: DPD, NPA classification, SMA categories.

What portfolio and regulation questions are asked?

Question	A short answer
Gross NPA, net NPA and provision coverage ratio: what is the difference?	Gross NPA ratio is NPAs over advances. Net NPA ratio takes provisions off NPAs and advances. PCR is provisions over gross NPAs. A worked example is 40 gross NPAs on 1,000 of advances with 25 of provisions: 4.0%, 1.5% and 62.5%.
What is a roll rate?	The share of loans in one bucket this month that are in another bucket next month. The flow from 1-30 to 31-60 is a roll rate, and the share that returns to Current is the cure rate.
Why use vintage analysis when you already track the portfolio NPA?	A portfolio ratio mixes loans of all ages, so a worse new cohort can hide behind older loans that are performing. A vintage follows each disbursal month as it ages, and compares cohorts at the same age.
What is collection efficiency?	Amount collected over amount due. Say whether the due amount is only the month's instalments or includes overdue brought forward, and say how payments are allocated.

Full guides: gross NPA vs net NPA, provision coverage ratio, roll rates, vintage analysis, collection efficiency.

What model questions are asked?

Question	A short answer
Why do scorecards use logistic regression?	It is explainable, it gives a calibrated probability of default, its coefficients turn into points that every team can read, and it is stable when the portfolio mix shifts.
What are weight of evidence and information value?	WoE for a bin is the natural log of the share of goods divided by the share of bads in the bin. IV adds up, across bins, the gap between those shares times the WoE. It ranks how useful a variable is.
What do Gini, KS and PSI tell you?	Gini and KS measure how well the score ranks goods and bads. PSI measures whether the population scored now looks like the one the model was built on.
Gini fell from the development value. What do you check?	First that the cohort is mature and the bad definition has not changed. Then PSI and CSI to see whether the population or a variable moved. Then calibration by band. Only then decide to keep, recalibrate or rebuild.
What is expected credit loss?	PD times LGD times EAD, measured over 12 months in Stage 1 and over the lifetime in Stages 2 and 3.

Full guides: why scorecards use logistic regression, validating a scorecard, expected credit loss.

What SQL questions are asked?

In an SQL round, ask about the table and the definition before you type. Say what you assume, then write.

Task: show loans and principal by bucket at a month-end.

SELECT bucket,
       COUNT(*)            AS loans,
       SUM(pos_amt)        AS pos
FROM   loan_month
WHERE  month_end = '2026-09-30'
GROUP  BY bucket
ORDER  BY MIN(dpd);

Task: find the loans that moved from 1-30 in August to 31-60 in September.

SELECT a.loan_id
FROM   loan_month a
JOIN   loan_month b
  ON   b.loan_id = a.loan_id
 AND   a.month_end = '2026-08-31'
 AND   b.month_end = '2026-09-30'
WHERE  a.bucket = '1-30'
  AND  b.bucket = '31-60';

Task: find duplicate rows in a snapshot table.

SELECT loan_id, month_end, COUNT(*) AS rows_found
FROM   loan_month
GROUP  BY loan_id, month_end
HAVING COUNT(*) > 1;

Question: your gross NPA ratio doubled overnight. What do you check first? Duplicate or missing rows in the snapshot, a changed bucket boundary (more than 90 versus 90 and above), a change in the as-of date, and a change in what the principal column holds. Run the duplicate query first.

All the queries here were run on sample data. The SQL guide has ten more.

How do you answer well in the interview?

Give the definition first, then the number.
Say what you assume before you write a query.
When you do not know an RBI rate, say that you would take it from the current direction.
Offer a check for every result: a row count, a total tied to finance, ten loans traced by hand.

Try it on your own tables

The free tools give you the prompt, the checks and how to present the result. You enter table names only.

See the free tools

Frequently asked questions

What are common credit risk analyst interview questions?

Definitions such as DPD, NPA and SMA, portfolio measures such as roll rates and vintages, model measures such as Gini and PSI, ECL, and SQL tasks on bucket distributions and roll rates.

Do credit risk interviews include SQL?

Many do for analyst roles. Expect to write a query on a loan table and to explain your assumptions and checks. Practise on a table you understand.

How should I prepare for a credit risk interview?

Know the definitions precisely, work through each ratio on a small example, run the common queries on a sample table, and read the current RBI text for classification and provisioning.

How do I answer a question about a regulatory rate?

State the rule you know, say where you would confirm it, and do not quote a rate from memory if you are not sure.

Keep reading

Data guide