How to talk an interviewer through a troubleshooting problem
When an interviewer asks how you would troubleshoot something, they are marking your order of work, not whether you guess the cause. Clarify the problem, find out who else is affected, gather facts, test the likely and cheap causes first, change one thing at a time, confirm the fix with the user, write it down, and escalate with evidence when it is beyond you.
"Talk me through how you would troubleshoot this" is the question in an IT interview that most rewards a method. The interviewer may have a cause in mind, but finding it is not the point. They want to hear you ask before you act, narrow the problem down, test the cheap and likely explanations first, change one thing at a time, check the fix with the person who reported it, and know when to hand it on. Say each step out loud, including the ones that feel obvious.
What the interviewer is marking
A strong answer sounds like someone who could be left alone with a real user. Three habits come through in it. You find out what is actually wrong before touching anything. You work out how far the problem reaches, because one user and a whole office are different jobs. And you can say why you are doing each check, so that if the interviewer interrupts with "that came back fine", you know what it rules out and where to go next.
That last part matters if the interviewer plays the user or feeds you results as you go. Treat it as a conversation, and ask for the information you would ask a real user for.
A method you can say out loud
Microsoft's guide to troubleshooting DHCP opens with questions rather than fixes. Its checklist asks when the problem started, whether there are error messages, whether it worked before and what changed, whether it is constant or intermittent, and whether it affects all clients or only some. Those questions work for almost any fault.
- Clarify. What exactly happens, what error appears, and what were they trying to do? Get the exact wording of any message.
- Scope. One user, one team, one site or everyone? One device or every device that user has? Since when?
- Gather facts. Look before you change anything: the device's network settings, recent changes, logs, whether a colleague has the same problem.
- Form hypotheses and test them. Start with what is both likely and quick to check. A loose cable or a locked account comes before a server rebuild.
- Change one thing at a time. If you change three settings and it starts working, you do not know which one fixed it, and you may have broken something else.
- Verify. Ask the user to try again, and confirm it works for them, not just for you.
- Document. Write what you found and what you did in the ticket so the next person does not start from zero.
- Escalate with evidence. When the cause is outside what you can see or change, hand it on with what you checked and what you found.
The four scenarios below are the kind interviewers use. The commands are Windows ones; the reasoning carries across to other systems.
One user cannot reach a website that everyone else can
Start by asking what the browser shows. A "can't find this site" message, a security warning and a page that never finishes loading point in different directions. Does any other website work? Does it fail in another browser too?
Then scope it. A colleague in the same office can reach the site, so the internet connection and the website are probably fine. The difference is something about this user or this device.
Now gather facts. Run ipconfig /all to confirm the machine has a sensible address, gateway and DNS servers. Then compare name resolution with a working colleague. Resolve-DnsName followed by the site's name performs a DNS query; if the two machines get different answers, or this one gets none, name resolution is the suspect. Microsoft's ping documentation describes the same test from another angle: if pinging the IP address works but pinging the name does not, you might have a name resolution problem, and the name might be coming from the local Hosts file, DNS or NetBIOS. A stale cached DNS answer can be cleared with ipconfig /flushdns, which flushes the DNS client resolver cache.
If the name resolves correctly, test the connection itself with Test-NetConnection followed by the site's name and -Port 443, which checks whether a TCP connection to the web port can be made. If that succeeds but the browser still fails, look at what sits between the browser and the network on this one machine: a proxy setting, a VPN or a browser extension.
Finally, have the user load the site, then note in the ticket what the cause was. If it was a Hosts file entry or a proxy setting, record how it got there.
The whole office has lost the internet
This is the opposite case. Everyone is affected, so stop looking at individual PCs. Ask what exactly is down: the internet only, or internal systems too? Can people reach the file server or the printer? Do phones on mobile data work? If internal systems are fine, the local network is up and the problem sits at the edge, between the office and the internet.
On one affected PC, ipconfig shows whether it still has a normal address. Microsoft's Wi-Fi troubleshooting guidance notes that an address starting 169.254 means Windows could not get one from the router, which would point at DHCP rather than the internet line. If the address is normal, ping the default gateway. Microsoft's guidance explains the reasoning: if the router replies, your connection to it is fine and the problem may be the modem or the internet provider, so check for an outage from another device.
Next, ping a public IP address and then a public name. If the address answers and the name does not, the line is up and DNS is the problem. tracert shows how far traffic gets before it stops, by sending packets with an increasing time-to-live and listening for the routers that reply. Read its output with care: Microsoft notes that some routers never send those replies and show as a row of asterisks for that hop, so a gap in the middle does not always mean traffic stops there.
Ask what changed: a firewall update, building work, a new device. Then tell people. One short message to the office saying the internet is down, that it is being worked on and when the next update will come means far fewer people ring to ask. A good answer also says you would raise it as a high-priority incident and give the provider your evidence: when it started, that the gateway responds, and where tracert stops.
A laptop will not join the Wi-Fi after a password change
The first question is whose password changed. If the office changed the Wi-Fi key, this laptop has the old one saved. If the user changed their own account password and the Wi-Fi signs in with their account, the laptop may be offering the old credentials.
Then the scope: is it only this laptop, or the user's phone as well, and can other people join the same network?
If the Wi-Fi key changed, Microsoft's steps are to open Manage known networks, select the network, choose Forget, then connect again and enter the new password. Microsoft also suggests trying the same network on a different device; if that works, the problem is likely the laptop.
If the user's own password changed, this is where to show you know about knock-on effects. A phone or laptop that keeps trying an old password can lock the account. On the domain controller, event 4740 records each lockout, and its Caller Computer Name field names the machine the failed attempts came from. On the laptop, Microsoft points to the WLAN-AutoConfig/Operational log in Event Viewer for wireless connection errors. Where the network checks sign-ins through a Network Policy Server, event 6273 on that server records a rejected attempt and its reason code; Microsoft lists reason code 16 for a wrong password and advises checking whether the account is locked. Forgetting the network and connecting again with the new password is the first thing to try on the laptop. The user's other devices need the new password too, or the lockouts will continue.
If the laptop is offering the right password and the server still rejects it, pass it to the network team with the time of the attempt and the reason code.
A shared drive is slow
Ask what is slow. Opening one document, browsing folders, or copying files? Big files or thousands of small ones? Since when?
Then the scope: one user or everyone, one share or all of them, in the office or only over VPN?
Microsoft's guide to slow SMB transfers is a good answer in itself. Copying large numbers of small files is slow by nature with File Explorer and other single-threaded tools, because each file has to be created before its data moves, and throughput can fall below 1 MB per second. robocopy with /MT copies several files at once, which helps with exactly that job. Antivirus scanning adds a little delay to every file, which adds up when there are thousands. On the server side, the storage can be the bottleneck: Microsoft gives roughly 110 MB per second of sustained storage throughput for each 1 Gbps of network. Packet loss on the network makes TCP slow itself down. SMB signing and encryption also cost performance, how much depending on the hardware.
So the scoping answers steer you. One user copying thousands of small files is probably expected behaviour, and the answer is a better tool. Everyone slow on every share since Tuesday is a server or network problem with a start time, and the next question is what changed on Tuesday.
When it goes to the server team, give numbers, not adjectives: which share, which operation, how long it took, from where, and since when.
Answers that lose marks
- Jumping to a fix ("I'd reinstall the driver") before asking anything.
- Changing several things at once and not being able to say which one worked.
- Rebooting shared equipment such as a router or server in working hours without saying you would check the impact or ask first.
- Calling it fixed without checking with the user.
- Never escalating, or escalating with nothing but "it doesn't work".
Practising before the interview
Say your method out loud on real problems. KBY's Daily Triage exercises are short simulated incidents that ask you to reason from symptoms to a cause, and they are good practice at narrating your checks. KBY runbooks show the same structure on real faults: the runbook on Kerberos sign-in failures caused by clock skew goes from diagnosis through containment and fix to verification. The Enterprise Networking Fundamentals lessons cover the network side, and the port and protocol lookup is a quick reference for which service uses which port.
Best of all is a fault you have caused and fixed yourself. A home lab lets you break DNS, DHCP and account lockouts on purpose, so that when an interviewer describes one, you are describing something you have seen. For the shorter technical questions that come in the same interviews, see IT support interview questions.
Common questions
- What if I don't know what is causing the problem in the scenario?
- That is normal and often intended. Keep narrating: say what you would check next, what each result would tell you, and at what point you would escalate and to whom. A clear method with no final answer scores better than a confident wrong guess.
- Should I name specific commands in my answer?
- Yes, where you know them, with a few words on what each one tells you, such as ipconfig to see the address and gateway, or ping to test whether the gateway answers. Naming a command without knowing what its result means does more harm than good.
Sources
Every figure and rule in this guide comes from these pages, read on 17 November 2026. Pay rates, exams and schemes change: check the source before you rely on one.
- Guidance for troubleshooting DHCP (Microsoft Learn)
- ipconfig (Microsoft Learn)
- Resolve-DnsName (Microsoft Learn)
- ping (Microsoft Learn)
- Test-NetConnection (Microsoft Learn)
- tracert (Microsoft Learn)
- Fix Wi-Fi connection issues in Windows (Microsoft Support)
- 4740(S): A user account was locked out (Microsoft Learn)
- 802.1X authentication issues troubleshooting (Microsoft Learn)
- Guidance for troubleshooting Network Policy Server (Microsoft Learn)
- Slow SMB files transfer speed (Microsoft Learn)
Spotted something wrong or out of date? Tell us and we will correct it.