Morgan Stanley Data Analyst Interview Questions
A data analyst at a global investment bank and wealth management firm works with trade, transaction and client account data where every figure may be checked by a reviewer or a regulator. Prepare to show clean SQL, careful reconciliation between systems, respect for confidential client data, and the ability to explain a number to a banker or financial advisor in plain words. What to show here: accuracy, audit trails and care with client data. Each question below has what the interviewer is checking, an answer frame and a sample spoken answer to adapt.
How Morgan Stanley hires, in its own words, and its questions for every job →
1. Tell me about a time the same figure came out different in two systems or two reports. How did you work out which one was right?
- Name the two sources and the figure that did not match.
- Explain how you narrowed the gap: same date, same filters, same definitions.
- Say what the root cause was and who owned the fix.
- Show how you recorded it so the next person can check it.
“In my last project, a monthly balance in our reporting database did not match the figure in the finance team's spreadsheet. I did not assume either one was right. First I checked that both used the same cutoff date and the same list of accounts. The gap shrank but did not close. Then I compared the two sources account by account and found a handful of closed accounts still counted in one of them. I told the data owner, they fixed the filter, and I wrote a short note explaining the rule. After that, your result.”
Weak answers pick the number they like better, or blame the other team without showing how the gap was traced.
2. Describe a time a report had a hard cutoff, like a month-end close or a morning deadline, and something broke late. What did you do?
- Set the deadline and why it could not move.
- Describe what broke and when you noticed.
- Explain who you told and what trade-off you proposed.
- Share what you changed so it did not happen again.
“We had a report that had to reach managers before the start of the working day. Late the night before, a source file arrived with missing rows. I did not want to send numbers I could not trust, so I messaged my lead right away and suggested two options: send the report on time with the affected section clearly marked as pending, or send it a little late in full. We chose the first. I filled the gap once the file was reloaded. Afterwards I added a row-count check that alerts us before the report runs.”
Weak answers hide the problem until the deadline passes, or quietly send numbers they had not checked.
3. You have a trades table and an end-of-day positions table for client accounts. How would you write SQL to find accounts where today's trades do not explain the change in position?
- Sum today's trade quantity per account and security.
- Compute the change between yesterday's and today's position per account and security.
- Join the two with a full outer join so missing rows still show.
- Keep rows where the values differ or one side is missing, and explain how you would handle cancelled or late trades.
“I would build two small result sets. The first groups today's trades by account and security and sums the signed quantity, with buys positive and sells negative. The second takes today's position minus yesterday's position for the same account and security. Then I full outer join them on account and security, and use coalesce to treat a missing side as zero. I keep the rows where the two numbers are not equal. I would also ask whether cancelled or amended trades sit in the same table, because they are a common reason for false breaks.”
Weak answers use an inner join and silently drop accounts that only exist on one side.
4. A daily series of trading volume for one desk shows a huge spike on a single day. How do you decide whether it is a real market event or a data error?
- Look at the raw rows behind the spike, not just the total.
- Check for duplicates, wrong units or a late file loaded twice.
- Compare with related series and known market news for that day.
- Decide, document the decision, and flag it in the report.
“First I would open the raw records for that day and look for duplicate trade IDs, or the same file loaded twice. I would also check units, because a quantity entered in the wrong unit can look like a spike. If the rows look clean, I would compare other desks and the wider market that day, since a real event usually shows up in more than one place. If it is real, I keep it and add a note. If it is an error, I raise it with the data owner rather than just deleting the point myself.”
Weak answers delete the outlier straight away or report it without checking where it came from.
5. How would you build an audit trail into a recurring report so a reviewer can trace any figure back to its source data?
- Keep the queries or code in version control.
- Record which source tables and extract dates each run used.
- Save a snapshot or run log with input counts and totals.
- Write down business rules and changes so a reviewer can follow them.
“I would make every run repeatable. The SQL or Python lives in version control, so anyone can see exactly what logic produced a figure and when it changed. Each run writes a small log: which tables it read, the as-of date, row counts and control totals. I would keep the output from each run instead of overwriting it. Finally, I would keep a plain document of business rules, like how cancelled trades are treated. If a reviewer asks about a number months later, I can rerun the same logic on the same inputs and show the same answer.”
Weak answers rely on manual spreadsheet steps that nobody else could repeat.
6. When would you choose Excel, SQL or Python for an analysis on client account data, and how does the data being confidential change that choice?
- Match the tool to the size of the data and whether the work repeats.
- Explain where each tool is strongest.
- Explain how confidentiality limits copying data to files or personal machines.
- Give a short example of your choice.
“For a quick one-off look at a small, already approved extract, Excel is fine. When the data is large or lives in a database, I use SQL so the work stays close to the source. For repeated work, joins across many sources, or statistics, I use Python so it can be scheduled and reviewed. With client data I also ask where the data will end up. I avoid exporting full client files to spreadsheets that get emailed around, I only pull the columns I need, and I mask names or account numbers when the analysis does not need them.”
Weak answers name a favourite tool for everything and never mention access rules or data masking.
7. What draws you to data work inside a firm that combines investment banking, trading and wealth management, and why Morgan Stanley?
- Say what interests you about financial data specifically.
- Connect it to the mix of businesses at the firm.
- Show you respect the regulation and accuracy that come with it.
- Link it to skills you already have and want to grow.
“I like work where a number has real weight. At a firm that runs investment banking, trading and wealth management together, the same data can support a trader, an advisor and a risk reviewer, and each needs it to be right. That appeals to me more than analysis where nobody checks the result. I also know a regulated bank expects careful controls and clear records, and I enjoy that discipline. I have been building my SQL and reconciliation skills, and I want to use them where accuracy matters every day.”
Weak answers talk only about the brand name or pay and show no idea what the firm actually does.
8. A financial advisor calls you shortly before a client meeting, asking why a client's portfolio value in your report differs from the client's statement. What do you do?
- Confirm the account and the two values using approved channels.
- Check the usual causes: valuation date, pricing time, pending trades, cash.
- Explain the likely reason in plain words and say what is still unconfirmed.
- Follow up in writing once confirmed.
“First I would confirm which account and which two figures we are comparing, using our internal systems rather than asking for details by personal message. Then I would check the common causes: my report may use a different valuation date or price time than the statement, or include a trade that has not settled yet. I would tell the advisor in simple words, for example, the statement shows last month's close while this report shows yesterday's prices. If I am not sure yet, I say so clearly, and I follow up by email once I have confirmed it.”
Weak answers use jargon the advisor cannot repeat to a client, or guess an explanation without checking.
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.