JPMorgan Chase Data Analyst Interview Questions
A data analyst at a large bank for people and businesses works with card spending, account activity, branch and app usage, and fraud signals across a very wide customer base. Prepare to describe queries in words, to choose measures for things like fraud alerts and card applications, to question a sudden change in a chart, and to show care with personal data. What to show here: careful numbers about customers and respect for their data. Each question below has what the interviewer is checking, an answer frame and a sample spoken answer to adapt.
How JPMorgan Chase hires, in its own words, and its questions for every job →
1. Tell me about a time your analysis showed that a product or campaign people were proud of was not working. How did you tell them?
- Describe the product and what people believed.
- Explain what the data showed and how you checked it.
- Describe how you presented it.
- Share what the team decided.
“In a previous role, the team had launched a reminder email and believed it was bringing customers back. When I compared customers who received it with a similar group who did not, the return rate was almost the same. Before saying anything, I checked my groups were fairly matched and asked a colleague to review the query. Then I showed the result to the owner privately first, with the method on one page and two ideas to try next. She was disappointed but glad to know. We tested a different message, and your result.”
Weak answers soften the result until it says nothing, or announce it in a large meeting with no warning to the owner.
2. Describe a time someone asked you for customer-level data, with names or account details, when a summary would have answered their question. What did you do?
- Describe the request and who made it.
- Explain what they really needed.
- Show what you gave instead and why.
- Mention the rule or process you followed.
“A colleague in marketing asked me for a full list of customers with their contact details and spending, to see which age groups used a feature most. I asked what decision he was making. He only needed counts by age band. I explained that I could not send personal records by email, and that he did not need them. I built a summary table with groups large enough that no single person could be identified, and shared that. He got his answer the same day. I also pointed him to the access request process in case he needed more later.”
Weak answers send the full extract to be helpful, or refuse without offering anything useful.
3. You have a table of card transactions with the card, the time and the city. How would you find cards used in two different cities within one hour, which may point to fraud?
- Order each card's transactions by time.
- Bring the previous transaction beside each row with a window function.
- Keep rows where the city differs and the time gap is under an hour.
- Exclude online purchases and explain remaining false alarms.
“I would use a window function partitioned by card and ordered by time, with LAG to bring the previous transaction's city and time onto each row. Then I filter for rows where the city is different from the previous one and the gap between the two times is less than one hour. Before that, I would keep only payments made with the card present, because an online purchase has no real city. I would also expect false alarms from neighbouring cities, so I would add distance between the cities if we have it, and treat the result as a list to review, not proof.”
Weak answers join the table to itself with no time limit, or forget that online payments are not tied to a place.
4. From a table of account transactions, how would you list customers who were active every month last year but have had no transactions in the last three months?
- Define active, for example one customer-started transaction in the month.
- Count active months per customer in the earlier period.
- Find customers with no qualifying rows in the recent period.
- Combine the two and exclude closed accounts.
“First I would agree what active means. I would count only transactions the customer started, not interest or fees posted by the bank. Then I group by customer and count distinct months with activity in the earlier year, keeping those where the count equals twelve. For the recent three months, I need customers with no rows, so I use NOT EXISTS against the same table with the recent date range. Joining the two gives the list. I would remove accounts that were closed, since those customers did not go quiet, they left, and that is a different question.”
Weak answers count bank-posted entries as activity, or use an inner join that cannot find customers with no rows.
5. The bank sends a text message asking the customer to confirm a suspicious card payment. Which measures would tell you whether these alerts are working well?
- State the goal: stop fraud without blocking honest payments.
- Measure how many alerts were real fraud and how much fraud had no alert.
- Measure reply rate, reply time and cards blocked in error.
- Watch the customer cost: complaints and reduced card use.
“There are two ways to fail: missing fraud, and troubling honest customers. So I would measure the share of alerts that the customer confirmed as fraud, and the share of later fraud claims that never had an alert. I would track how many customers reply and how fast, because an alert nobody answers protects nothing. For the honest side, I would count payments declined that the customer later confirmed as their own, and whether those customers use the card less afterwards. I would split all of this by payment type, since online and in-person patterns differ.”
Weak answers count alerts sent or fraud blocked alone, and ignore the honest customers who were interrupted.
6. You test a shorter application form for a credit card and more people complete it. Why is that not enough to call the test a success, and what else would you look at?
- Separate completing the form from being approved and from using the card.
- Name the later outcome: repayment behaviour.
- Check whether the new form changed who applies.
- Propose early signals and a longer follow-up.
“More completed forms only means the form is easier. What matters is what happens next. I would compare approval rates between the two groups, because if the extra applicants are mostly declined, we have wasted their time. Then I would look at whether approved customers activate and use the card. The real outcome is repayment, which takes months to see, so I would keep the two groups tagged and follow them. In the meantime I would compare the mix of applicants, such as income band and existing relationship with the bank, to see whether the shorter form drew a different kind of applicant.”
Weak answers declare a winner on form completion alone, or never mention that repayment is seen much later.
7. Your dashboard shows card spending in one region fell sharply overnight. A manager is about to raise it with senior leaders. What do you check before anyone acts on it?
- Check that the data is complete: load times, row counts, late files.
- Split the drop by shop type, card type and channel.
- Look for outside causes such as a holiday or an outage.
- Tell the manager what is known and what is not.
“I would ask the manager for a short pause and check the data first. Did all files for that region arrive, and do the row counts look like a normal day? A late file is the most common cause of a sudden fall. If the data is complete, I split the drop by shop type, card type and whether the payment was online or in person. A fall in one payment network points to an outage. A fall everywhere points to a holiday or bad weather. Then I tell the manager plainly what I have confirmed, what is still open, and when I will know more.”
Weak answers explain the drop with a story before checking whether the data had fully loaded.
8. What interests you about analysing data on how people and businesses use their accounts and cards at JPMorgan Chase?
- Say what questions about banking behaviour interest you.
- Connect them to accounts, cards and payments.
- Show awareness of privacy and fairness.
- Link to your skills and what you want to learn.
“Money data is a very honest record of how people live: when they are paid, what they spend on, when things get tight. At JPMorgan Chase that record covers a huge range of people and small businesses, so an analyst can ask questions that matter, like which customers are about to struggle with a payment and how the bank could help earlier. I also know this data is personal, and I want to work somewhere that takes that seriously. My strength is careful SQL and clear writing, and I want to learn more about fraud patterns and lending.”
Weak answers talk about big data in general and show no thought for the people behind the rows.
No figures on this page: pay depends on the level, the city and the offer in front of you. When the offer comes, the free salary negotiation script helps with that part.
The questions above are the prep. On the real call, ClapAssist helps with the ones you did not practice.
ClapAssist is an AI interview assistant for Mac and Windows. It listens to the interview on your computer and shows you what to say, in short lines you can read while you talk. Your live interview audio and screen are never stored. Your resume and notes are saved to your account so the app fills them in on any computer. It is hidden from screen share on every plan, including Free; only you can see it.