The Salesforce admin questions that separate doers from certificate holders
There’s a specific question that sorts a room of Salesforce admin candidates faster than anything else on a scorecard: “a user can’t see a record they should be able to see, walk me through how you debug it.” It’s not a trivia question and you can’t memorise the answer. Most salesforce admin interview questions are trying to get at the same underlying thing, which is whether you’ve actually operated an org or whether you’ve passed an exam about one.
Why the certification does not settle it
The Salesforce Certified Administrator credential covers configuration, user management, security, automation and reporting. It’s a reasonable qualification and hiring managers largely treat it as a floor rather than a differentiator, because the exam tests recognition and the job tests judgement.
What that means for you as a candidate: your certification gets you the screen, and everything after it is about decisions you’ve made in a live org with users complaining. Prepare accordingly. If your Salesforce experience is a personal developer org, say so honestly and lean on what you built in it, because being caught inflating org experience ends the process.
Security and sharing, which is where most interviews start
Explain the sharing model from the top down. Organisation-wide defaults set the baseline, role hierarchy opens access upward, sharing rules open it sideways, manual sharing handles exceptions, and teams and territories layer on top for specific objects. Salesforce’s own sharing overview is the reference. Say it in that order, because the order is the debugging method.
Profiles versus permission sets: when do you use each? The answer interviewers want is that profiles should be minimal baselines and permission sets carry the variation. If you tell them you clone a profile per team, expect a follow-up about how that scales past twenty teams, and it doesn’t.
What’s the difference between a role and a profile? A profile is what you can do. A role is what you can see through the hierarchy. People conflate these constantly.
A user can’t see a record. Debug it. Check the object-level permission on the profile or permission set first, then the org-wide default for that object, then the role hierarchy position, then sharing rules, then whether it’s a private object with manual sharing, then field-level security if they can see the record but not the data. Working top down and saying so out loud is the whole point of the question.
What are the risks of View All and Modify All? They bypass sharing entirely. Name that, and name the audit consequence.
Salesforce admin interview questions on the data model
Master-detail versus lookup. Master-detail cascades deletes, inherits sharing, allows roll-up summaries, and the child can’t exist alone. Lookup is loose and optional. The follow-up is usually when you’d convert one to the other and what breaks when you do.
What’s a junction object and when do you need one? Many-to-many, two master-detail relationships. Have a real example ready.
Record types versus page layouts. Record types drive business process and picklist values, page layouts drive presentation, and they’re assigned together through profiles. Candidates who describe record types as “different layouts” get marked down.
How would you prevent duplicate accounts? Duplicate and matching rules first, validation rules where the logic is simple, and a conversation about whether the duplicates come from integrations rather than users, because if they do, the fix isn’t in the UI.
Automation, and the question about Workflow
Which automation tool would you choose, and why? Flow is the default now for record-triggered automation. Apex is for what Flow can’t do: complex bulk logic, callouts with specific handling, anything needing real transaction control. The answer that scores is one that mentions maintainability, because the person interviewing you is probably maintaining somebody else’s automation right now.
Before-save versus after-save flows. Before-save updates fields on the triggering record without a second DML operation, so it’s faster and the right choice for same-record field updates. After-save handles related records, emails and anything needing the record ID.
How do you avoid hitting governor limits in Flow? Keep DML and queries out of loops, bulkify, use collection variables and a single update at the end. If a candidate has never hit a limit, they haven’t run automation at volume.
An automation is misfiring in production. What now? Debug logs, Flow error emails, the fault path, and whether you can deactivate safely without orphaning records mid-process. Bonus points for mentioning that you’d check what changed recently before assuming the automation is at fault.
The scenario questions, which carry the most weight
These salesforce admin interview questions are where senior candidates separate themselves.
Sales leadership wants a new required field on Opportunity. Reps say it slows them down. What do you do? The wrong answer is to build it. The right shape is to ask what decision the field supports, whether the data exists somewhere already, whether required-always or required-at-stage is what’s actually needed, and to make it required at the stage where reps have the information. Requiring data people don’t have yet produces garbage entries, which is worse than no field.
You’ve inherited an org with 60 profiles, 200 custom fields on Account and no documentation. Where do you start? Field usage analysis, login history to find which profiles are live, and a freeze on new customisation until you’ve mapped it. Anyone who says “rebuild it” without asking about the business calendar is answering as an engineer rather than an admin.
How do you handle deployments? Change sets, sandboxes, or a proper source-driven pipeline. Say what you’ve actually used. Claiming CI experience you don’t have is easy to unpick with one follow-up about the deployment that failed.
Reporting, which gets underestimated
Reporting questions come late in the loop and candidates treat them as a formality. They aren’t. Whoever hires you will be asked for numbers within a fortnight.
What’s a joined report and when would you use one? Multiple report blocks with different report types sharing a common field. Useful for comparing, say, opportunities against cases for the same accounts. Know its limits too, because joined reports can’t be used as the source for some dashboard component types.
Explain a roll-up summary versus a formula field. Roll-ups aggregate child records up a master-detail relationship and are stored. Formula fields calculate on the fly from the record itself. The follow-up is what you do when you need a roll-up across a lookup, and the honest answers are a declarative rollup tool, Flow, or Apex.
A dashboard shows different numbers to different users. Why? The running user setting. Dashboards run as a specified user or as the viewer, and this is one of the most common support tickets an admin actually receives.
How to prepare, and the part people skip
Rebuild the sharing model in a developer org from scratch and break it deliberately, then debug your own breakage. That exercise teaches the debugging order better than reading it does, and the debugging order is what’s being tested.
Then practise saying the answers out loud, under interruption. The scenario questions in particular go wrong not because candidates lack the knowledge but because they answer at length and never reach the point. An interviewer who cuts in at forty seconds is testing whether you can be redirected, and rehearsing alone trains exactly the opposite habit.
Craqly’s mock interview mode runs that on your desktop. It asks out of order, interrupts where you’re running long, and gives you a transcript so you can see which answers wandered. The free Starter plan is 20 credits a month, one credit being a minute of live session, resetting monthly, with paid plans from $19 a month billed yearly, checked on 6 September 2026. Our notes on answering questions you don’t know are worth reading before a scenario round, since admitting a gap well is a scored behaviour here rather than a penalty, and the STAR structure covers the inherited-org story you’ll be asked for.
If you prepare one thing, make it the record-visibility debug walkthrough, spoken aloud, in order, in under ninety seconds.
