Hiring a Salesforce administrator means evaluating how a candidate translates a business process into usable, maintainable configuration. A user who cannot see a record may have a different problem from a user who cannot edit a field, and an automation that works for an administrator may fail for its intended audience. Recruiters should look for candidates who trace those differences before changing access or adding another rule.
This guide to 100+ Salesforce Admin Interview Questions and Answers contains 120 questions across beginner, intermediate and experienced levels. It focuses on declarative configuration, user access, data quality, Flow, reporting, releases and support. Feature behavior, available tools and limits depend on the org’s licenses, enabled features and release; the interview should reflect the environment the administrator will support.
How can recruiters assess Salesforce Admin judgment when a candidate knows Setup navigation but struggles to explain effective access or automation failures?
Give the candidate a specific user, record and failed action, then ask which access layer or automation path they would inspect first. Introduce an import or a second user with different permissions and ask how they would test the proposed fix. Strong answers identify the smallest justified change, representative exceptions and a recovery path rather than assuming administrator success proves the process works for everyone.
Sherlock AI helps hiring teams assess signals of potential AI cheating, deepfakes, impersonation and unauthorized assistance during remote interviews. Combine those signals with configuration walkthroughs, scenario exercises and structured follow-up questions. They provide context for human review rather than conclusive proof of misconduct.
Salesforce Admin Interview Questions
Start with the data model and access layers, then select declarative automation or operational scenarios relevant to the role. Each question has a matching numbered answer below. Ask candidates to explain how they would verify their answer using a representative user and record.
Salesforce Admin Interview Questions for Beginners
Establish the data model, access layers and declarative tools.
What does a Salesforce administrator do?
What is a Salesforce org?
How does metadata differ from business data?
What is a standard object?
What is a custom object?
What is a record?
How does a field's label differ from its API name?
What should guide field-type selection?
How do lookup and master-detail relationships differ?
What is a junction object?
What is a formula field?
What is a roll-up summary field?
What is a validation rule?
How does a page layout differ from field-level security?
What is a Lightning record page?
What is a record type?
How do picklists improve data quality?
What is a dependent picklist?
What is a global value set?
What is a Salesforce user license?
What is a profile?
What is a permission set?
What is a permission set group?
What is a muting permission set?
How does a role differ from a profile?
What are organization-wide defaults?
How does the role hierarchy affect access?
What is a sharing rule?
What is a public group?
What is a queue?
How do object permissions and record access interact?
Why is hiding a tab insufficient to secure an object?
What is a list view?
What is a report type?
How does a report differ from a dashboard?
What is a dynamic dashboard?
What is Flow Builder?
How does a screen flow differ from a record-triggered flow?
What is a sandbox?
Why should an administrator document changes?
Salesforce Admin Interview Questions for Intermediate
Probe configuration choices, troubleshooting and automation behavior.
How would you investigate a user who cannot see a record?
How would you investigate an unexpectedly visible record?
What is the difference between View All Records and ordinary sharing?
How would you troubleshoot a missing field?
How should onboarding access be designed?
How do freezing and deactivating a user differ?
How would you offboard a user safely?
How would you review permission set group access?
When should you use a custom permission?
How should a validation-rule exception be managed?
How would you prevent an invalid date range?
How do required fields at different layers differ?
How would you design a cross-object formula?
When is a formula insufficient for a business requirement?
How would you decide whether to create another record type?
What should be checked before changing a field type?
How would you retire an unused field?
How do before-save and after-save record-triggered flows differ?
How should record-triggered flow entry criteria be designed?
What is the purpose of the prior record value in a flow?
How do you avoid automation recursion?
Why should database operations be avoided inside loops?
What is a flow collection variable?
What is a flow fault path?
How would you test a flow?
What does Flow Trigger Explorer help with?
Why does flow execution context matter?
When should asynchronous flow work be considered?
How would you migrate legacy automation to Flow?
How do approval processes differ from ordinary field updates?
How would you select a data import method?
What is an external ID useful for?
How does upsert differ from insert and update?
How would you prepare a CSV import?
How do matching rules and duplicate rules differ?
How should import failures be handled?
Why is an export not automatically a complete recovery plan?
How would you troubleshoot a report missing records?
How do dashboard running-user settings affect security?
How do row-level and summary formulas differ in reports?
Salesforce Admin Interview Questions for Experienced
Assess governance, releases, incidents and sustainable administration.
How would you design an access model for a new business unit?
How would you handle an urgent request for broad administrator access?
How would you investigate access that changed unexpectedly?
How should integration-user access be governed?
How would you manage sensitive data in a sandbox?
How would you plan a sandbox refresh?
What can change sets move?
How would you prepare a release package?
Why does deployment validation not replace business testing?
How would you plan rollback for declarative changes?
How would you manage dependencies between multiple flows?
How would you investigate a flow that fails only during bulk loads?
How would you investigate duplicate task creation?
How should an administrator respond to a data-quality incident?
How would you plan a large record reassignment?
How would you evaluate ownership skew or data concentration?
How would you troubleshoot row-lock errors?
How would you govern permission set proliferation?
How should a shared administrator team avoid conflicting changes?
How would you prioritize an administration backlog?
How would you turn a vague request into a testable requirement?
How would you measure user adoption?
How would you decide between configuration and custom development?
How would you assess a managed package before installation?
How would you prepare for a Salesforce release?
How would you handle an undocumented legacy customization?
How would you audit dashboard access?
How would you govern metric definitions?
How would you troubleshoot a timezone-related reporting discrepancy?
How would you design a consent-related configuration change?
How should a production data repair be tested?
How would you manage an automation bypass for a migration?
How would you respond when a user reports incorrect results after a release?
How would you assess administrator documentation quality?
How would you manage data retention requirements?
How would you plan training for a new workflow?
How should an administrator handle an AI-generated configuration suggestion?
How would you explain an edition-dependent limitation?
What should an administrator handover contain?
What demonstrates senior Salesforce administrator judgment?
Salesforce Admin Interview Questions and Answers
Use these answers as a practical rubric for administration decisions. Reward candidates who explain effective permissions, test exception paths and connect configuration to a business outcome. Confirm the target org’s supported features and behavior before applying a procedure.
Salesforce Admin Interview Questions and Answers for Beginners
These answers distinguish the platform’s data, access and presentation layers. Look for concrete examples that show why a visible field or assigned role does not by itself establish the permissions required for an action.
1. What does a Salesforce administrator do?
An administrator translates business requirements into maintainable configuration, access controls, data processes and user support. The role includes testing and change governance rather than simply granting permissions or creating fields on request.
If a sales manager asks for another status field, I would first clarify the decision it supports and inspect existing fields, reports and automation. The right solution might be a controlled picklist, a report filter or training rather than a duplicate field. I would test the chosen configuration with affected user personas and document ownership so the org remains understandable after the immediate request is complete.
2. What is a Salesforce org?
An org is a Salesforce environment containing its configuration, users and data. Production and sandbox orgs serve different purposes, and availability of features depends on licenses, editions and enabled settings.
Two orgs can contain objects with the same names but different records, permissions and enabled features. An administrator must verify which org is open before changing configuration and know how a sandbox relates to production. A sandbox login is not evidence that every integration points to a test service, so connected systems and notification destinations also belong in the environment review.
3. How does metadata differ from business data?
Metadata defines structures and behavior such as objects, fields, permissions and automation. Business data consists of records such as accounts and cases; moving configuration does not automatically move those records.
Deploying a custom field definition creates the structure, but it does not automatically populate the new field on existing customer records. A change may therefore need both a metadata deployment and a separately planned data migration. Keep those activities distinct: test the configuration dependencies, then validate record counts and values for the data work rather than assume one successful deployment covers both.
4. What is a standard object?
A standard object is supplied by Salesforce for a platform or product purpose, such as Account or Contact. Prefer understanding its existing relationships and behavior before replacing it with a custom object.
Before creating a custom Customer object, I would assess whether Account and Contact already represent the required organization and person relationships. Standard objects may carry product-specific behavior, reporting and integrations that a replacement would have to reproduce. The choice still depends on the actual business model and licenses; using a standard object for an unrelated purpose solely to avoid a custom object can be equally confusing.
5. What is a custom object?
A custom object models information not adequately represented by the available standard model. Its design should consider ownership, sharing, reporting and lifecycle rather than only the fields needed on one screen.
For equipment inspections, a custom Inspection object might store the inspected asset, date, result and responsible team. I would define whether an inspection can exist independently, who owns it and which users need access before deciding its relationships. Reporting and retention requirements may reveal that several repeated fields on an Asset record should instead be separate related inspection records.
6. What is a record?
A record is an instance of an object with field values and a Salesforce identifier. Its visibility and editability depend on multiple access layers, not just whether a user can open the object's tab.
An Account record for one organization can contain a name, owner and related contacts, while another Account is a different instance of the same object. A user who can read Accounts may still lack access to this particular account, and a user who can see it may lack edit permission on a sensitive field. Diagnose those layers separately instead of solving every access complaint with a broad administrator permission.
7. How does a field's label differ from its API name?
The label is the user-facing name, while the API name identifies the field in formulas, integrations and metadata references. Changing a label is usually less disruptive than changing a referenced API name, but both changes require impact review.
A custom field might display “Renewal Date” while its API name is Renewal_Date__c. A report reader mainly sees the label, but an integration or formula may depend on the API name. Before renaming, inspect dependencies and communicate changes to consumers. Labels should also remain unambiguous: two fields with nearly identical labels can cause bad data entry even when their API names are technically distinct.
8. What should guide field-type selection?
Choose a type that represents the business meaning and supports needed validation, reporting and integration. Storing numbers or dates as free text creates avoidable sorting and data-quality problems.
A renewal date belongs in a Date field if the requirement is a calendar day, while an external account code may need Text to preserve leading zeros. A numeric field is appropriate for quantities that require arithmetic, but not every digit-only identifier is a number. Consider field length, precision, requiredness and downstream mapping together, then test imports and reports with representative values before committing to the type.
9. How do lookup and master-detail relationships differ?
A lookup provides a reference with configurable relationship behavior, while master-detail tightly couples the detail's lifecycle and access to its master. Consider ownership, deletion behavior and roll-up requirements before choosing or converting a relationship.
An inspection that must be independently owned may fit a lookup to Asset, whereas a line item that should depend entirely on its parent may fit master-detail. Master-detail affects access and deletion behavior as well as enabling eligible roll-ups. I would walk through parent deletion, reassignment and reporting scenarios before choosing it; a relationship should not be made tightly dependent merely to obtain one convenient total.
10. What is a junction object?
A junction object represents a many-to-many association, commonly through two master-detail relationships. It can also hold attributes of the association, such as participation status, rather than duplicating them on either parent.
For students attending courses, an Enrollment junction record can link one student and one course while storing attendance status or completion date. Those attributes belong to the association, not globally to the student or course. If using the common two-master-detail pattern, review the resulting access and lifecycle behavior. Also define how duplicate enrollments are prevented if one student-course pair should occur only once.
11. What is a formula field?
A formula field calculates a read-only value from an expression when evaluated. It is not an editable stored snapshot, so time-dependent formulas can change their displayed result without a user editing the record.
A formula showing days until renewal changes as TODAY() changes, even if nobody edits the customer record. That is useful for a current display but unsuitable as a permanent historical statement of what was shown last month. For an audit snapshot, use an intentionally stored value with a defined capture event. Formula references, return type and null treatment should be tested against the relevant record scenarios.
12. What is a roll-up summary field?
A roll-up summary aggregates related detail records using supported operations and criteria in an eligible master-detail relationship. Ordinary lookup relationships do not automatically support the same native roll-up field behavior.
A parent order can summarize eligible detail line amounts using a supported SUM roll-up with criteria that exclude cancelled lines. I would test creating, changing and deleting detail records to confirm the total follows the intended rule. A lookup from line to order does not automatically provide the same native configuration; changing the relationship solely for aggregation can introduce unwanted ownership or deletion consequences.
13. What is a validation rule?
A validation rule prevents a save when its formula evaluates to true under the applicable save process. Write an actionable error message and test valid as well as invalid paths, including imports and automation.
For a hypothetical custom object, a rule can reject a quantity below zero with an error explaining the allowed range. Because the rule evaluates to true when it should block, reversing that logic is a common mistake. Test ordinary edits, bulk imports and automation that changes the field. If a legitimate integration needs an exception, define a narrow controlled bypass rather than deactivate the rule for everyone.
14. How does a page layout differ from field-level security?
A page layout organizes the interface and can mark fields read-only or required in that layout. Field-level security governs field access more broadly, so hiding a field from a layout is not a security control.
Removing Salary from a page layout may tidy the screen, but it does not stop a user with field access from reaching the value through another supported path. Field-level security is the appropriate layer for restricting that field. Conversely, a field hidden by security cannot be made readable merely by adding it to a layout. Verify the effective result using representative user permissions, not only the administrator's page preview.
15. What is a Lightning record page?
A Lightning record page arranges components and experiences for viewing or working with a record. Activation and assignment determine who sees the page, and component visibility does not replace underlying data permissions.
A service record page might show a case timeline, related records and a component that appears only for a relevant status. Page activation can differ by app, profile or record type under the supported assignment options. If one user sees an older page, inspect its assignment before rebuilding the component. Visibility rules improve the experience but do not authorize data the user is otherwise forbidden to access.
16. What is a record type?
A record type supports differences such as business processes, available picklist values and layout assignments for an object. It does not by itself create record-level security or justify duplicating every minor business variation.
Two opportunity processes may need different sales stages and layouts, making record types useful. A minor cosmetic difference may be better handled through page configuration or guidance rather than another record type and its maintenance burden. Record-type availability influences creation and UI behavior, but it should not be treated as a mechanism that hides all records of another type from a user.
17. How do picklists improve data quality?
Picklists guide users toward controlled values and consistent reporting. Use restricted picklists when approved values must also be enforced for API or app writes; unrestricted picklists can accept additional values through those paths. Define ownership and change procedures so categories remain meaningful.
A support-reason picklist can prevent reports from splitting “Billing,” “billing issue” and “Invoice” into accidental categories when one approved category is intended. Decide whether values are restricted, who may add them and how old values will be handled. Integrations and imports must use the agreed values too; a tidy selection menu alone does not establish consistent data across every write path.
18. What is a dependent picklist?
A dependent picklist filters available values based on a controlling field under configured dependencies. Test record types, existing records and integrations as well as the interactive selection experience.
A controlling field for product family can narrow an issue-type picklist to relevant choices, reducing incorrect combinations during entry. I would test a user changing the controller after a dependent value is already stored and review existing records that predate the dependency. Record types and integration paths can add constraints, so the on-screen dropdown is only one part of validating the intended relationship.
19. What is a global value set?
A global value set shares a governed set of picklist values across eligible fields. It improves consistency but also creates shared dependencies, so a value change can affect several objects or processes.
If several objects use the same approved region categories, a global value set can prevent separate lists from drifting apart. The shared definition also means a proposed rename or deactivation needs a dependency review across those objects, reports and integrations. Reuse is helpful when the meaning is genuinely shared; two teams using the same word for different concepts may need distinct controlled sets instead.
20. What is a Salesforce user license?
A user license determines the baseline capabilities available to a user, subject to additional permissions and feature licensing. Assigning a permission set cannot grant a capability that the user's licensing does not support.
When a user cannot access a feature, I would check the assigned license and any required feature or permission-set license before adding more permissions. A license enables a supported capability boundary; the user still needs the relevant permissions and record access. Promising that a permission set can turn any license into a full-featured user ignores the platform's licensing constraints and can lead to an invalid solution.
21. What is a profile?
A profile supplies a user's baseline settings and permissions within the license model. Each user has one profile; use a deliberate baseline and modular permission assignments rather than proliferating nearly identical profiles.
A lean baseline profile can establish required settings, while task-based permission sets add access for activities such as case handling or report creation. This reduces the need for a new profile whenever one user's responsibilities change. I would still inspect profile contributions when debugging excessive access, because adding modular sets does not remove broad permissions that were already granted in the baseline.
22. What is a permission set?
A permission set grants additional supported permissions to assigned users. Permissions are generally additive, so assigning another permission set does not remove access granted elsewhere.
A temporary assignment to a report-creation permission set can grant that capability to an analyst without changing every user on the same profile. Removing the assignment only removes that source of access; another permission set or the profile may still grant it. Document why the permission exists, who owns it and how access should be reviewed when the user's responsibilities change.
23. What is a permission set group?
A permission set group bundles permission sets around a job function or access package. It simplifies assignment, but administrators must still understand the combined effective permissions and applicable licenses.
A Service Agent group might bundle case handling, knowledge access and a required application permission set. That makes onboarding more consistent, but changes to a shared component set can affect several groups. Before adding a permission, inspect which users receive it through every assignment route and test the resulting access. A business-friendly group name is useful documentation, not proof that its contents still match the job.
24. What is a muting permission set?
Muting suppresses selected permissions contributed through its permission set group. It does not revoke the same permission if a profile, separate assignment or another group grants it.
Suppose a reusable permission set includes Delete on Cases, but one job-function group should not receive that operation. Muting can suppress that group's contribution without editing the shared set for everyone. If the user also receives Delete directly from another set, the user still has it. An access investigation must therefore trace all grants; muting is not a universal deny rule that overrides the rest of the org.
25. How does a role differ from a profile?
A role primarily participates in record-access hierarchy, while a profile defines baseline user settings and permissions. A senior business title does not automatically require broad object permissions or a matching role arrangement.
Two users can share the same baseline profile but sit in different roles because they need different record visibility. Conversely, a manager may need access to a team's records without requiring permission to configure the platform. I would design the role hierarchy around who must see whose records, then separately grant supported object and field capabilities. Copying the company organization chart without reviewing data access can overexpose information.
26. What are organization-wide defaults?
Organization-wide defaults establish baseline record access for users who do not otherwise receive access through ownership or other mechanisms. Choose a justified baseline and then grant necessary access through the appropriate sharing design.
If representatives should initially see only their own opportunities, a restrictive baseline can be opened for managers and collaborating teams through the appropriate access mechanisms. Starting with a broadly public baseline and adding a sharing rule cannot take access away. Review the object's supported settings and other broad permissions before treating the default alone as the full effective visibility model.
27. How does the role hierarchy affect access?
The hierarchy can extend access upward according to the object's sharing behavior and settings. It should reflect data-access requirements, and administrators must check object-specific exceptions rather than assuming every object behaves identically.
For a standard sales use case, a manager above a representative can receive access through the hierarchy according to the object's sharing behavior. That does not automatically grant missing object or field permissions. Custom-object hierarchy settings and other object-specific rules need review. I would test the intended subordinate and manager personas rather than assume a role label such as “Director” produces the correct access by itself.
28. What is a sharing rule?
A sharing rule grants additional record access to a defined group based on ownership or criteria. It opens access beyond the baseline; it does not make an already broader baseline more restrictive.
A criteria-based rule might share eligible regional accounts with a specialist group that needs to collaborate, while an owner-based rule can share records owned by a defined population. Both grant additional access. If a record remains visible after removing one rule, inspect ownership, hierarchy, other rules, teams and broad permissions before concluding that the change failed; another valid access route may still apply.
29. What is a public group?
A public group collects users and supported groups or role-related memberships for use in sharing and other features. Review nested membership and business ownership because a membership change can affect several access paths.
A public group named Regional Support can be used in a sharing rule, so adding a member may immediately affect access to records covered by that rule. Nested groups and role-related membership make the effective population less obvious than a short list of names. Assign an owner to review membership and dependencies, especially when employees transfer roles or a group is reused for a new purpose.
30. What is a queue?
A queue holds records for supported objects so members can work or take ownership through an agreed process. Queue membership, assignment and access behavior need testing; creating a queue alone does not guarantee balanced workload distribution.
A queue can hold incoming cases until eligible team members take or are assigned work through the configured process. I would define which records enter it, who monitors aging and how ownership transfers. A queue with no active monitoring can become a backlog rather than a routing solution. Automated balancing or omnichannel behavior requires the relevant additional configuration and supported features; it is not implied by creating the queue.
31. How do object permissions and record access interact?
Object permissions determine whether a user can perform an operation on that object, while record access determines which records are available for it. Field permissions and other controls still apply, so one layer cannot be used to infer the whole access result.
A shared Case can be visible to a user who has Read but not Edit on Cases, so sharing alone does not make it editable. The user also needs access to the particular fields involved, and additional controls may constrain the operation. For troubleshooting, test one operation on one record and trace object, record and field layers instead of granting Modify All to bypass an unexplained failure.
Access layer | Question it answers | Example limitation |
|---|---|---|
Object permission | May this user perform this operation on the object? | Read does not grant Edit. |
Record access | Which records can the user access through applicable sharing? | Object Read alone does not expose every private record. |
Field access | Which fields may the user read or edit? | A layout cannot override denied field access. |
32. Why is hiding a tab insufficient to secure an object?
Tab visibility controls navigation, not the complete authorization model. A user with underlying access may reach records through search, related records, reports or APIs.
If the Account tab is hidden but the user retains the necessary permissions, a related-record link or report may still expose accounts. That means the change affects navigation rather than revoking the underlying capability. To remove access, identify and adjust the actual supported permission and sharing grants, then test representative entry points. A clean navigation bar is useful usability work, but it is not a security boundary.
33. What is a list view?
A list view presents records matching filters within the viewer's access context. Sharing a list view makes the view available; it does not independently grant access to the underlying records.
A team can share a “My overdue cases” view definition while each viewer receives only records their access permits and the filters select. If a user sees fewer rows, investigate ownership filters and record access before changing the view's sharing. “Share the view” answers who can use the saved configuration; it does not answer who can read every case that another viewer sees.
34. What is a report type?
A report type defines the objects, relationships and fields available to a report. It influences which records can appear, so a missing record may reflect the report type rather than only a filter problem.
An Accounts with Contacts report type can omit an account with no contact depending on the relationship definition. Relaxing a date filter will not recover a record excluded at that level. I would confirm the intended population first, then inspect the report type's object relationships, available fields and filters. A report's row count can reflect child-record multiplicity as well as missing parents.
35. How does a report differ from a dashboard?
A report queries and presents data using a report type, filters and formatting. A dashboard visualizes source reports under its configured viewing context, which must be reviewed for appropriate data exposure.
A report can list and group open cases, while a dashboard can visualize that report as workload by team. For a dashboard using a fixed running user, the viewer may see data according to that user's access, so folder sharing and running context both require review. A visually correct chart is not enough; check whether its audience should receive the underlying aggregation and whether the refresh timestamp is suitable.
36. What is a dynamic dashboard?
A dynamic dashboard uses the viewing user's context rather than a single fixed running user for its data visibility. Availability and limits vary, so confirm the org's supported configuration before promising one for every audience.
One dynamic dashboard can show each salesperson their permitted pipeline without maintaining a separate dashboard per person. It still depends on appropriate source reports, permissions and supported configuration. Test at least two representative viewers with different access, and review any option allowing users to choose another viewing context. Avoid promising a fixed number of dynamic dashboards without checking the org's edition and applicable limits.
37. What is Flow Builder?
Flow Builder is Salesforce's declarative tool for designing supported automated and interactive processes. Flows require careful entry criteria, execution-context review, error handling and tests just as coded automation does.
A flow can guide a user through case triage or update related records after a defined event. Before building it, identify the trigger, expected record changes and failure path, then choose the appropriate flow type. Declarative configuration can still create recursion, unexpected access or bulk failures. Test the actual entry conditions and record populations rather than judge the automation only by a successful single-record debug run.
38. How does a screen flow differ from a record-triggered flow?
A screen flow guides a user through an interactive sequence, while a record-triggered flow runs in response to a supported record event. Their invocation and execution contexts differ, so copying logic between them requires review.
A screen flow can ask a user for a service choice and validate the entered details before continuing. A record-triggered flow can react when a case is created or updated without presenting screens. The latter should use precise entry criteria so unrelated edits do not repeat work. When sharing logic, review execution context, transaction behavior and available inputs instead of assuming identical behavior because both are drawn in Flow Builder.
39. What is a sandbox?
A sandbox is a separate environment used for development, testing or training based on a production org. Sandbox types differ in copied data and capacity; protect sensitive data and connected services even outside production.
A sandbox lets the team test configuration and representative data scenarios before production deployment, but its type determines what is copied and how much data it supports. A refreshed environment can also have different IDs or connection settings. Check email delivery and integrations so a test does not notify real customers or update a live external system. Passing a test in an undersized or differently configured sandbox has limits that should be recorded.
40. Why should an administrator document changes?
Documentation records the requirement, design, dependencies, test evidence and support ownership. It lets another administrator explain or reverse a change without guessing from a field name or an old chat message.
For a new validation rule, document the business requirement, formula, error message, intended exceptions and tested save paths. Include dependent flows, imports or integrations and identify the owner who can resolve a later failure. A rollback note should distinguish disabling the rule from repairing records already affected by related automation. Useful documentation explains the reason and operational consequences, not just which Setup page was clicked.
Salesforce Admin Interview Questions and Answers for Intermediate
These scenarios examine configuration choices and troubleshooting. Strong candidates trace the user, data and automation context, explain the smallest useful test and consider bulk or exception behavior.
41. How would you investigate a user who cannot see a record?
I would confirm the exact user, record and business reason access is expected, then trace the access layers rather than immediately broadening permissions. Check the license and object permission first, followed by ownership, sharing, hierarchy and any applicable restrictions.
For example, a user may have permission to read Cases but no sharing path to one particular case. Another user may have record access but lack permission to a field, which is a different problem. Compare the effective access of representative users and inspect the specific record.
I would test the smallest authorized correction and verify that unrelated records remain protected. Copying an administrator's profile or granting View All Records may hide the cause while exposing much more data than the job requires.
42. How would you investigate an unexpectedly visible record?
I would enumerate every path granting access to the record: ownership, hierarchy, groups, teams, sharing rules, manual or managed sharing and broad object or org permissions. Salesforce access is often additive, so removing one grant does not necessarily remove the effective access.
A user can remain able to read a record after a sharing rule changes because a permission set grants View All Records for that object. A group membership or team assignment can also supply a separate path that is easy to overlook.
Compare the observed access with the approved requirement before changing configuration. Record the actual grant path and test the correction with both intended and unintended users. A more restrictive page layout is not a fix for an underlying record-access issue.
43. What is the difference between View All Records and ordinary sharing?
View All Records is an object-level permission that can bypass ordinary record sharing for that object. It is much broader than adding a user to one record's team or a narrowly scoped sharing rule, and it is different from org-wide View All Data.
For example, granting it for Cases to solve one missing-case complaint can expose cases across the object. It does not simply mean the user can see every field or edit every record; field permissions and edit-related permissions are separate considerations.
Use the broad grant only for a role whose approved responsibilities require it. Review where it is assigned through profiles, permission sets or groups, and test effective access. Prefer a suitable sharing path when the requirement concerns a subset of records.
44. How would you troubleshoot a missing field?
I would identify the exact page, app, record type and user, then distinguish field security from page presentation. A field can be absent because of field-level permissions, a page layout, Dynamic Forms visibility conditions or the assigned Lightning record page.
If the field appears to an administrator but not a sales user, first check the user's effective field permission. If it appears on one record type but not another, inspect assignment and visibility conditions before changing security.
I would also confirm whether the user is opening the expected component or related record. Hiding a field in one layout does not protect it across APIs or reports, while granting field access does not guarantee every page displays it. Correct the layer responsible and test representative entry points.
45. How should onboarding access be designed?
Map approved job tasks to a license, minimal baseline and permission sets or groups that represent named responsibilities. Onboarding should answer what the employee needs to do, not merely copy whoever happens to sit nearby.
A support agent may need case handling and a specific knowledge capability without sales administration or bulk export privileges. Test a representative workflow using that access package, including actions that should remain denied.
Assign an owner for approval and later review, and document any temporary exception. License and feature entitlements constrain which permissions can be assigned, so verify the actual org configuration. Reusable role-based packages simplify onboarding only when they remain understandable and are reviewed as duties change.
46. How do freezing and deactivating a user differ?
Freezing prevents login while leaving the user active, which is useful when a dependency prevents immediate deactivation. Deactivation changes active-user status and makes the user license available according to the platform's process. Freezing alone does not release the license.
For example, a departing employee referenced in a required configuration may need to be frozen while that dependency is reassigned. That is an interim access-control step, not completion of offboarding.
Review sessions, identity-provider access, integrations and ownership through the organization's approved procedure. Neither action means deleting the user's historical business records. Test the relevant access removal and preserve the audit trail rather than assuming one user-screen change covers every credential or scheduled dependency.
47. How would you offboard a user safely?
I would follow an offboarding checklist covering login removal, identity-provider access, active sessions, integration credentials, scheduled work and ownership. Coordinate with the relevant owners so access ends promptly without unexpectedly stopping a business process.
If the user owns a scheduled report or serves as an automation reference, identify a suitable replacement before treating deactivation as finished. Record ownership and sharing may need reassignment under a business rule rather than transferring everything to an administrator.
Preserve the user's history and records. I would not delete customer data because its owner has left. Verify the resulting access and monitor dependent jobs, and document any temporary freeze or unresolved dependency with an owner and completion date.
48. How would you review permission set group access?
Review the group's included permission sets, its muting configuration, assignments and resulting effective permissions. A muted permission is suppressed through that group; it is not a global denial that removes the same permission granted elsewhere.
If a user still has an export capability after it is muted in one group, check the profile, directly assigned sets and other groups. Editing the muting set repeatedly will not fix an independent grant path.
Test the actual persona's required and prohibited tasks, and verify that the group's recalculation or status is appropriate after changes. Keep responsibility packages understandable rather than accumulating overlapping exceptions. A group's name is documentation, not proof of the permissions its current contents deliver.
49. When should you use a custom permission?
A custom permission represents a named capability that formulas, flows or code can check. It separates the business exception from a particular person's username, making assignment and review more explicit.
For example, an approved data-steward capability can allow a narrowly defined correction process to bypass one validation condition through $Permission. Assign that capability through a governed permission set rather than embedding a list of employee names in the formula.
A custom permission does not automatically grant object or field access; the consuming logic gives it meaning. Document where it is checked and who may receive it. Broadly assigning a bypass capability can defeat the control even when the underlying formula is technically correct.
50. How should a validation-rule exception be managed?
Use a narrow exception with an explicit business purpose, owner and assignment policy. A custom permission can make a validation bypass visible and reviewable, but only if it is granted to the identities that genuinely require the exception.
For example, a controlled migration account may need a temporary bypass for one historical-data rule. Disable the exception when the migration is complete and reconcile the loaded records. Turning off the entire rule lets unrelated users and integrations create the same invalid state.
Test normal entry, authorized exception and denied exception cases across relevant save paths. Avoid a blanket administrator exclusion without checking which automation or integrations run as that identity. The exception should preserve the rule's purpose, not become the default route around it.
51. How would you prevent an invalid date range?
First define whether either date may be blank and whether equal dates are valid. For two Date fields where blanks are allowed but an end before the start is invalid, the validation should return true only when both dates exist and the end is earlier.
The following Salesforce validation formula uses two custom Date fields:
Test end before start, equality, end after start and each blank combination. If both dates are mandatory, enforce that requirement separately or include it deliberately. DateTime values introduce time-zone and time-of-day questions, so do not apply a Date-only rule without confirming the actual business interval.
52. How do required fields at different layers differ?
Requiredness at the page-layout layer controls that interface, while schema-level requirements and validation rules can enforce conditions across additional save paths. Choose the layer from the business invariant rather than assuming a red marker on one page prevents all incomplete records.
If every newly created record must have a business identifier, test API creation, imports and automation as well as the standard edit page. A condition such as ‘reason required only when status is Rejected’ usually needs an appropriate conditional rule rather than making the field universally required.
Check existing data and downstream integrations before tightening enforcement. A new requirement can break previously valid updates or loads. Give users a useful error message at the correct field or page location, and verify that the intended exceptions are explicit.
53. How would you design a cross-object formula?
A cross-object formula reads through a supported relationship path to calculate a value from a related record. Confirm the field types, missing-relationship behavior and business meaning before building it, and review the sensitivity of the resulting value.
For example, showing a parent account attribute on a case can simplify the agent's view, but the formula may expose that attribute even when the user cannot directly access the parent record. Source-record sharing alone is therefore not a sufficient exposure review.
Protect the formula field appropriately and test as the intended user. A formula reflects its referenced data rather than storing a historical snapshot, so changes to the parent can change the displayed result. Use a stored, governed value when the requirement is to preserve what was true at a past event.
54. When is a formula insufficient for a business requirement?
A formula calculates a value; it does not by itself create related records, send a notification or preserve a historical event snapshot. Those requirements need appropriate automation or storage in addition to, or instead of, the calculation.
For example, a formula can display the current account tier on an opportunity. If reporting must retain the tier at the moment the opportunity closed, a live formula can later show the wrong historical meaning after the account tier changes.
Define whether the value should be current, recalculated on a specific event or immutable after capture. Choose the simplest supported mechanism that preserves that behavior, with entry criteria and permissions reviewed. Do not add Flow merely because a formula feels limited if the requirement is only a current derived value.
55. How would you decide whether to create another record type?
Create a record type when there is a real business-process or experience distinction that its capabilities support, such as different picklist availability or relevant process configuration. It is not a generic answer to every request for another page layout or reporting category.
If two teams use the same process but want a different presentation, Lightning page configuration or other simpler options may be sufficient. If the underlying lifecycle and allowed values differ materially, a record type may be justified.
Consider maintenance across permissions, defaults, reports, integrations and automation. Test record creation and conversion paths with the intended personas. Record types do not replace record-level security, so do not use them as the sole mechanism for separating confidential populations.
56. What should be checked before changing a field type?
Review both stored values and every dependency before changing a field type. Formulas, flows, reports, integrations, validation and external mappings can rely on the current representation even when the field looks simple on a page.
Converting a text identifier to a number can lose leading zeros or reject alphanumeric values. A date-related conversion can alter meaning if the original process relied on time-of-day. Supported conversions and restrictions depend on the field and platform configuration.
Test representative records in an appropriate environment, preserve recoverable source data and plan a migration path if an in-place conversion is unsuitable. Validate integrations and reports after the change. A successful metadata save is not proof that the business data retained its intended meaning.
57. How would you retire an unused field?
Confirm usage through metadata dependencies, integration owners and stakeholders before retiring a field. No recent edits does not prove it is unused: a field may still drive a report, export, formula or external matching process.
Mark it as deprecated through the approved process, remove it from new entry paths where appropriate and monitor consumers before removal. A staged replacement can provide time to migrate an integration that reads the old API name.
Preserve data and recovery information according to policy, and obtain the required approval for removal. Check supported retention and restoration behavior rather than assuming deletion can always be undone indefinitely. The retirement plan should name the replacement or explain why the information is no longer needed.
58. How do before-save and after-save record-triggered flows differ?
Use a before-save record-triggered flow for supported changes to the triggering record when those changes belong in the original save. After-save logic supports additional work such as creating or updating related records and actions that require the new record's ID.
For example, normalizing a field on the triggering case is a good before-save candidate. Creating a related follow-up task requires an appropriate after-save design. Updating the triggering record again after save adds another update path that may invoke more automation.
After-save does not automatically mean a separate committed asynchronous transaction. Review the actual path and save order. Choose from supported capabilities and test interaction with other automation rather than assuming every action is available in every record-triggered flow mode.
59. How should record-triggered flow entry criteria be designed?
Entry criteria should describe the event and state transition that justify the work, not merely a broad condition that remains true forever. For updates, decide whether the flow should run whenever criteria are met or only when the record changes into the qualifying state.
If a follow-up task should be created when a case first becomes Escalated, an unrelated later description edit should not create another identical task. Compare old and new state or use the supported transition setting that matches the requirement.
Test creation, the intended transition, repeated saves and transitions away and back. Re-entry may be legitimate, so do not suppress it accidentally. Keep criteria understandable and include a duplicate-prevention rule where the side effect requires stronger protection than the trigger condition alone.
60. What is the purpose of the prior record value in a flow?
Prior record values let supported record-triggered logic distinguish a change from a value already present. Comparing $Record with $Record__Prior can prevent actions on unrelated edits when the business event is a transition.
For example, a notification for moving a deal to a particular stage should compare the previous stage with the current one. Checking only the current stage can notify again when an agent updates a phone number or note.
Handle creation and missing prior values according to the flow's event and available variables. Decide whether returning to the stage should generate another action. Prior-value checks explain the transition, but they do not replace idempotency when an external or related-record effect could be replayed through another path.
61. How do you avoid automation recursion?
Design automation so repeated evaluation converges instead of reapplying the same change. Narrow entry criteria, compare prior state where useful and avoid updating a field to the value it already has without a reason.
A case flow that updates an account can trigger another automation that updates the case, forming a loop across objects. Inspect the whole chain, including Apex and legacy automation, rather than only the flow visible in Flow Builder.
Use a small transaction trace or controlled test to identify the repeated transition and its owner. A broad bypass flag may stop the symptom while suppressing legitimate work. Clarify which automation owns each field and side effect, then verify repeated saves and bulk updates do not recreate work.
62. Why should database operations be avoided inside loops?
Queries and record updates inside a loop can consume shared transaction limits and add avoidable work. Retrieve the needed records in a suitable operation, apply per-record assignments in memory and perform supported collection updates outside the loop.
For example, if a flow needs to set a field on many related contacts, build a record collection containing the intended changes rather than issuing one Update Records element for each contact. The transaction may also contain other automation using the same limits.
Bulk design is not simply moving a box outside the loop: preserve the correct record IDs and values, and avoid retrieving an unnecessarily large population. Test a representative bulk invocation as well as a single record. Limits and supported features should be checked for the current org and release.
63. What is a flow collection variable?
A collection variable holds multiple values or records so a flow can process them together. A record collection is useful for preparing several changes and then passing them to a supported bulk element.
When looping through records, distinguish the current-item variable from the collection that will be updated. Changing a field on the current item does not mean the database has been changed. Follow the supported assignment and collection pattern so the intended modified records reach the final persistence step.
Keep the collection's purpose clear: records to inspect, records to create and records to update have different requirements. Ensure update records have the correct IDs, avoid unintended duplicates and verify the no-record case. A collection that contains records is not evidence that the flow saved them successfully.
64. What is a flow fault path?
A fault path handles failures from supported Flow elements and can route to a useful user message, diagnostic record or recovery process. It should distinguish a failed required action from a completed business operation rather than quietly continuing as though everything succeeded.
If creating a follow-up record fails, a screen message saying ‘Request completed’ is misleading even if the fault connector prevents a generic error page. Explain the recoverable next step to the user and retain a safe correlation reference for the support team.
Avoid exposing raw technical details or sensitive field values. Review transaction behavior: logging in a transaction that later rolls back may not leave durable evidence. Test the fault path itself, including whether the selected user and execution context can perform its logging or notification actions.
65. How would you test a flow?
Test the flow's entry and non-entry cases, boundary values, permissions, bulk behavior and failure paths in an appropriate environment. A successful debug of one administrator-owned record is only a narrow check.
For a stage-change flow, include creation directly in the target stage, transition into it, an unrelated edit while already there and a transition back into it. Confirm related-record counts so duplicate work is visible, and test a representative bulk update with other automation active.
Use realistic user permissions and the actual invocation context. Verify outcomes rather than only that an interview completed. Preserve test records and expected results in a repeatable plan, and check rollback or asynchronous behavior where the work crosses transaction boundaries.
66. What does Flow Trigger Explorer help with?
Flow Trigger Explorer helps inspect record-triggered flows for an object and event and manage supported execution ordering. It provides a useful view of overlapping flow responsibilities and potential sequencing assumptions.
If two after-save flows both update related work, reviewing them together can reveal duplicate ownership or a dependency that should be made explicit. The display is a starting point for understanding that flow set, not a complete transaction trace.
It does not show or reorder every automation mechanism in Salesforce's save sequence. Validation, Apex, approvals and other automation can still affect the outcome. Use the documented order of execution and controlled tests when diagnosing a broader interaction, and avoid relying on an order setting to repair circular dependencies.
67. Why does flow execution context matter?
Execution context affects how object permissions, field permissions and sharing are applied for the flow type and invocation. Determine the actual documented context instead of assuming all flows run exactly as the current user or all run with unrestricted access.
A screen flow launched in one way and record-triggered automation may have different access behavior. A design that works for an administrator can therefore fail for a service user—or expose data unexpectedly if it runs with broader access than assumed.
Test representative users, record ownership and denied cases. Review every input and output crossing a broader context, especially record IDs supplied by a user. Context is part of the security design, not merely a troubleshooting setting to broaden until the flow works.
68. When should asynchronous flow work be considered?
Consider an appropriate asynchronous path when work does not need to complete in the originating transaction and the feature supports the requirement. This can separate suitable post-commit work, but it changes the user's completion guarantee and failure handling.
For example, a noncritical downstream notification may be deferred while the record save completes. The user should not be told the external action has finished when it has only been scheduled. Track failure and retries through a defined owner or monitoring route.
Review availability, limits and supported actions for the actual flow type. Design idempotency when retries or repeated qualifying events can duplicate a related effect. Asynchronous execution is not a way to ignore transaction limits, guarantee immediate delivery or make a required validation happen after the decision.
69. How would you migrate legacy automation to Flow?
Inventory the legacy automation's conditions, actions, timing and dependencies before translating it to Flow. Define equivalent observable outcomes, including exceptions and failure behavior, rather than mapping each visible box mechanically.
An old rule may interact with another process that updates the same field. Activating the replacement while leaving both responsible for the action can create duplicate tasks or notifications. Plan a controlled handover and verify which version owns each behavior.
Use supported migration tools where applicable, then test the generated or rebuilt flow in context. Include bulk operations, existing records and scheduled behavior. Preserve a recovery plan and evidence of parity; successful conversion or activation does not establish that the whole business process behaves identically.
70. How do approval processes differ from ordinary field updates?
An approval process coordinates a governed decision: submission, eligible reviewers, pending state and defined approval or rejection outcomes. Field updates can support those transitions, but changing a status field alone does not provide the same review controls or history.
For a discount exception, the process should establish who can submit, who may approve, what happens if it is rejected and how the requester is informed. Consider record locking or edit behavior and delegation according to the chosen approval feature.
Do not let an unrelated flow set Approved merely because a condition looks satisfied if a human decision is required. Verify current feature capabilities and org configuration, including the applicable approval technology. Test alternate, missing and unavailable approver scenarios rather than only a straightforward approval.
71. How would you select a data import method?
Select an import method from the object, operation, volume, relationships and recovery requirements, then verify current tool support and permissions. The Data Import Wizard, Data Loader and integration APIs do not have identical object coverage or capabilities.
A small supported-object import may fit the wizard, while a controlled upsert with detailed success and error files may call for another tool. Consider which validation, duplicate rules and automation will run, because an import is not merely a passive file copy.
Start with a representative sample, map keys deliberately and preserve reconciliation information. Check current limits for the org and selected tool rather than quoting a memorized number as universal. The best method is the one whose behavior and failure recovery match the task.
72. What is an external ID useful for?
An external ID stores a governed identifier from another system and can support matching for upserts and relationships. It helps avoid relying on names or other mutable descriptive fields when synchronizing records.
For example, an ERP customer code can identify the corresponding account across repeated loads. Decide its format, source ownership and whether uniqueness must also be enforced. Marking a field External ID does not automatically make every configuration a unique business-key constraint.
Profile duplicates and missing values before using the field for matching. Changing external IDs can break relationships or create new records on the next load, so treat changes as an integration contract issue. Preserve the source-to-Salesforce ID mapping needed for reconciliation and recovery.
73. How does upsert differ from insert and update?
Insert creates records, update changes identified existing records, and upsert uses a supported key to choose between those outcomes. Upsert is useful for synchronization, but its correctness depends on an unambiguous matching key and deliberate field mapping.
If a source customer ID matches one account, the load can update it; an unseen ID can create a new account. Duplicate matching values or inconsistent formatting can instead produce errors or unintended new records. A changed name is not a reliable substitute for stable identity.
Test new, existing, duplicate and missing-key cases. Replaying a file safely also depends on what its updates do and whether related automation creates repeated effects. Upsert does not by itself prove that the whole integration is idempotent.
74. How would you prepare a CSV import?
Prepare the CSV against the target schema and operation: normalize dates and other formats, map fields, validate required values and resolve relationship keys. Preserve original source identifiers so each submitted row can be matched to its result.
A column containing numeric-looking customer codes should retain meaningful leading zeros, and a relationship field should use the supported matching representation rather than a display name that may be duplicated. Check blank-field behavior for the selected tool so an update does not erase values unintentionally.
Run a small representative sample that includes ordinary and exception cases. Reconcile successes, errors and target records before the full load. Keep the approved input and results with appropriate access controls, and do not treat an upload completion message as proof that every row was accepted.
75. How do matching rules and duplicate rules differ?
Matching rules define how Salesforce identifies possible duplicates. Duplicate rules define what to do with those matches under configured conditions, such as alerting or blocking a save. The detection criteria and enforcement behavior are separate decisions.
For example, a rule can compare contact details using supported exact or fuzzy matching, while the duplicate rule determines the user or import experience when a candidate match is found. A potential match is not automatically proof that two people are the same entity.
Test false positives, missed duplicates and different entry channels with representative data. Review whether the chosen operation and user context apply the expected rules. Avoid broad bypasses merely to make an import pass; resolve the matching policy and exceptions with the data owner.
76. How should import failures be handled?
Separate successful rows from failed rows and classify each error before retrying. A failed batch can contain substantial committed success, so resubmitting the whole file as inserts can create duplicates.
A validation failure may need corrected source data, a missing relationship may need an earlier parent load, and a permission failure may indicate an inappropriate execution account. Each requires a different fix rather than simply increasing the retry count.
Retain source keys and returned identifiers, correct only the intended records and use the appropriate matching method for replay. Reconcile total submitted, successful, failed and retried records. Also check downstream automation effects, because correcting a row and saving it again can trigger additional actions.
77. Why is an export not automatically a complete recovery plan?
An export is one recovery input, not proof that the organization can restore service. Recovery may also require metadata, relationships, attachments, integration configuration, ownership mappings and a tested order for restoring dependent objects.
A CSV of child records is difficult to reload correctly if parent identifiers cannot be mapped, and an export of data does not recreate the automation that makes the process work. Define the recovery point and time objectives before choosing the backup approach.
Test restoration in an appropriate environment, including permissions and reconciliation. Protect backup data and document who can perform recovery. A file that exists and opens successfully is not equivalent to a complete, current and usable recovery procedure.
78. How would you troubleshoot a report missing records?
Trace one known missing record through access, report type and filters. Confirm the running user, relationship requirements, date field, scope and record-state assumptions before broadening permissions or rewriting the entire report.
An account may be omitted because the report type requires a related contact, or because the date filter uses Created Date when the business expects Close Date. Those are report-definition issues, not evidence that the user needs View All Records.
Compare the record's actual values with each condition and test a minimal report if needed. Check report-folder access separately from access to the underlying data. Document the intended population so a fix does not quietly include unrelated records while restoring the missing example.
79. How do dashboard running-user settings affect security?
A dashboard's running context determines which data it presents, while folder sharing controls who can access the dashboard. A fixed running user can expose aggregate or detailed information based on that user's access, which may exceed a viewer's own record access.
Selecting an administrator as the running user to make every component populate can therefore reveal more than intended. A dynamic dashboard uses the viewer's access, subject to available features and configuration, but should still be tested with representative personas.
Review the audience, component details, drill-down behavior and refresh context together. Folder access is not a complete security review. Verify edition, permission and feature availability in the actual org rather than assuming every dashboard can use the same dynamic configuration.
80. How do row-level and summary formulas differ in reports?
A row-level formula calculates for individual report rows, while a summary formula operates on aggregates or grouping levels supported by the report. Choose the calculation level from the metric's meaning, not simply from which formula editor is easiest to find.
For a margin percentage, averaging each row's percentage can differ from total profit divided by total revenue. A summary formula can express the latter, while a row-level formula answers a per-record question. Both can be mathematically valid but describe different business measures.
Test zero denominators, missing values and grouping changes on a small known dataset. Confirm the report format's supported formula capabilities and limits. Label the metric clearly so users do not interpret an unweighted average as an organization-wide weighted result.
Salesforce Admin Interview Questions and Answers for Experienced
These answers address controlled change, incidents and sustainable administration. Look for evidence-based decisions, clear ownership and recovery planning that accounts for data changes as well as metadata.
81. How would you design an access model for a new business unit?
Start with personas, data sensitivity, ownership and required collaboration, then map them to licensing, permissions and sharing. Validate both allowed and forbidden actions with representative users before rollout.
For example, sales representatives may edit their own opportunities, regional managers may need regional visibility, and finance may need selected commercial fields without sales-administration powers. Separate object and field permissions from record access; a sharing rule cannot compensate for missing object access or hide records already exposed by a broader grant. Include transferred employees and cross-region collaboration in the test matrix. A model that works only for the initial organizational chart will become difficult to maintain after the first restructure.
82. How would you handle an urgent request for broad administrator access?
Clarify the blocked task and provide the narrowest approved capability or supervised operational path that meets it. Document any justified temporary elevation and its removal; urgency does not establish a need for unrestricted access.
If the request is to correct a report filter before a meeting, identify whether folder management or report-editing permission solves the problem. Giving System Administrator access would also expose unrelated settings and data. For genuine emergency elevation, use the organization’s approved process with a named owner, bounded duration and verification of removal. Capture the original failure so the team can fix the underlying access design; otherwise the temporary exception can silently become the permanent operating model.
83. How would you investigate access that changed unexpectedly?
Review user assignments, group membership, sharing configuration and relevant administrative history around the change. Preserve evidence and identify the actual grant or restriction before making broad compensating changes.
Suppose a representative can suddenly edit another team’s opportunities. Compare an affected record and user with an unaffected pair, then trace the access path rather than changing organization-wide defaults immediately. A new group membership, permission assignment or ownership relationship may explain the difference. Check administrative history against the incident time, while recognizing that one log does not capture every possible cause. Removing one grant is insufficient if another assignment still provides equivalent access.
84. How should integration-user access be governed?
Use a dedicated identity and supported authentication design with only the required object, field and operation access. Assign ownership, monitor usage and manage credential rotation without sharing a human administrator's login.
An order-sync integration might require read access to selected account fields and update access to a shipment-status field, but no ability to administer users. Document that contract and test attempted writes outside it. Verify the available integration-user license and authentication capabilities in the actual org rather than assuming every dedicated identity has the same entitlements. Include failed authentication and expired credentials in operational monitoring so a security improvement does not become an unnoticed outage.
85. How would you manage sensitive data in a sandbox?
Limit who can access it, use masking or synthetic fixtures where appropriate and review outbound integrations and messaging. A nonproduction environment can still expose real information or contact real recipients if copied settings remain active.
A copied customer email address remains real even when the record is in a sandbox. Before testers run a flow that sends notifications, confirm the destination and messaging controls, and replace sensitive values where policy requires it. Masking one object is insufficient if related notes, files or exports retain the same information. Prefer synthetic cases for routine training and reserve realistic restricted datasets for justified tests. Verify the post-refresh state instead of assuming previous sandbox protections survived.
86. How would you plan a sandbox refresh?
Identify work that could be overwritten, required data and configuration, external endpoints and post-refresh tasks. Communicate timing and preserve needed artifacts before the refresh rather than treating it as a harmless synchronization button.
A developer sandbox may contain an unreleased flow version that is absent from production, so refreshing first could erase the only copy of that work. Inventory and preserve pending changes before proceeding, and agree which environment becomes the source for ongoing development. After refresh, check access, test identities, scheduled work and integration endpoints before reopening it to users. Treat completion of the platform refresh and readiness for safe testing as separate milestones with different evidence.
87. What can change sets move?
Change sets move supported metadata between appropriately connected related orgs. They do not move ordinary business records, and dependencies or unsupported components may require another documented deployment approach.
For a new approval-related field and flow, the package may need the field definition, referenced resources and relevant supported access metadata. It will not bring the customer records used during testing. Check component support and dependencies in the source and target editions, then document any required data preparation separately. A change set is also not a general transfer mechanism between arbitrary unrelated orgs: the deployment connection and org relationship must support the intended route.
88. How would you prepare a release package?
Identify the exact components, dependencies, access assignments and manual steps, then validate them in the intended target path. Include test evidence, deployment ownership and a recovery plan appropriate to the changes.
A release that adds a required field may need configuration, access and a plan for existing records before the new rule becomes active. Put those steps in execution order with a responsible person and an observable completion check. Distinguish deployment from activation and user assignment where the component requires it. Rehearse the package against a representative target state; a successful development-org test does not prove that production dependencies or existing data will accept the same sequence.
89. Why does deployment validation not replace business testing?
Technical validation can confirm supported deployment requirements without proving that users can complete their actual work correctly. Test business outcomes, permissions and data behavior with realistic scenarios.
A deployment can succeed while a sales user cannot save an opportunity because a newly required field is hidden from that persona. Test the actual user journey with representative access, including existing records, missing inputs and an expected rejection. Also confirm that the change creates the correct downstream task or notification exactly when intended. Technical deployment acceptance and business acceptance answer different questions, and neither should be recorded as evidence for the other.
90. How would you plan rollback for declarative changes?
Record the prior configuration and identify which changes are reversible, which require forward fixes and whether data has already changed. Reverting metadata does not automatically undo emails, external actions or record updates already performed.
Suppose a flow sends an incorrect customer notification and changes a status field. Restoring the previous flow version can stop new incorrect actions, but it cannot unsend messages or decide which status values should be repaired. Separate containment, metadata recovery and data reconciliation in the runbook. A successfully deployed change set has no automatic undo button; restoring configuration requires a planned subsequent change. Preserve evidence of which records were affected before making that correction.
91. How would you manage dependencies between multiple flows?
Document shared fields and transitions, define ownership and inspect ordering and transaction effects. Reduce hidden coupling where possible and test the combined behavior rather than approving each flow in isolation.
Imagine one flow updates a qualification flag and another creates a task when that flag changes. Editing the first flow’s timing or entry condition can alter the second flow’s behavior without touching its canvas. Inspect the relevant record-triggered flows and supported ordering tools, then test the entire transition and repeated updates. Avoid relying on an accidental order. When several automations own the same field, agree a clear responsibility or consolidate the decision logic where that improves maintainability.
92. How would you investigate a flow that fails only during bulk loads?
Examine transaction limits, queries or updates in loops, shared automation and data-dependent branches. Reproduce with a controlled representative batch rather than assuming the import tool is defective.
A flow may work for one record but fail when an import activates many interviews in the same transaction. Salesforce can group operations across interviews reaching the same element, but that does not make a Create Records element inside a loop free of repeated work. Look for the failing element and the shared transaction’s other automation before changing batch size. A smaller batch can contain an incident, while the durable fix addresses the repeated operations or branching that consumed the limit.
93. How would you investigate duplicate task creation?
Trace entry conditions, repeated updates, retries and overlapping automation that can create the task. Add a justified idempotency or state-transition design rather than deleting duplicates without correcting the cause.
If a task should be created when a case first enters an escalation state, distinguish that transition from any edit while the case remains escalated. Check whether two active automations independently implement the same requirement or whether a retry repeats a completed action. A marker or existing-record check must match the business event and account for concurrent execution; a simple lookup is not a universal uniqueness guarantee. Reconcile existing duplicates only after preventing the creation path from repeating.
94. How should an administrator respond to a data-quality incident?
Contain the incorrect input or automation, preserve evidence and quantify affected records before repair. Validate a targeted correction and reconcile results; a mass update without a known selection can amplify the incident.
If an import put the wrong region on accounts, stop the import or faulty mapping and retain the input file and result identifiers under the appropriate access controls. Determine whether subsequent flows changed owners or sent communications, since the incident may extend beyond the original field. Repair a verified sample first and compare expected versus actual results. A filter such as “modified today” is rarely precise enough when legitimate users were editing records during the same period.
95. How would you plan a large record reassignment?
Assess sharing recalculation, ownership dependencies, automation and affected users, then stage and monitor the supported operation. Communicate access changes and validate representative records after completion.
Moving thousands of accounts from a departing employee can affect record visibility, dependent processes and related ownership behavior. Establish the expected outcome for each record type and user group before choosing the supported transfer method. Avoid overlapping large ownership changes with other work touching the same records. Validate both that the new owner can work and that previous users retain only the access the business intends. Record totals alone cannot prove that the reassignment preserved the access model.
96. How would you evaluate ownership skew or data concentration?
Identify unusually large concentrations of records or relationships and assess their effect on sharing, locks and operations. Work with architecture guidance and measured symptoms rather than assuming one universal record threshold applies to every org.
If many child records point to one parent, otherwise unrelated updates may contend on that shared relationship. If one owner holds a disproportionate share of records, sharing-related operations can also become expensive. Start from lock failures or slow operations and inspect the actual distribution rather than redesigning solely because a count looks large. A redistribution has its own business and sharing consequences, so involve the appropriate architect and validate a representative approach before moving production data.
97. How would you troubleshoot row-lock errors?
Look for concurrent updates to shared parents, ownership changes and automation that touches the same records. Reduce conflicting work or use an appropriate processing strategy, and retry only with a clear understanding of successful versus failed changes.
For example, two parallel import batches may update different contacts associated with the same account while automation updates that account. The apparent separation by contact does not remove contention on the parent. Review failure identifiers and overlapping jobs, then group or sequence conflicting work using the supported import strategy. Retry only failed work after confirming success results; replaying the complete input can create duplicate side effects. Serial processing may trade throughput for predictability and should be evaluated against the completion window.
98. How would you govern permission set proliferation?
Define reusable business capabilities, naming, owners and review criteria, then consolidate only after dependency and assignment analysis. Fewer sets are not automatically better if consolidation creates broad access bundles.
A permission set named after one employee makes its intended business purpose difficult to assess when that employee changes roles. Reframe reusable sets around finite tasks and use groups for appropriate personas, while retaining legitimate exceptions with owners. Before retiring a set, identify every assignment and compare effective access before and after. Combining unrelated capabilities merely to reduce the count can make least privilege harder, because future users receive permissions they never requested.
99. How should a shared administrator team avoid conflicting changes?
Use a common intake, ownership and release process with tracked configuration changes and a designated source of truth. Communicate work in progress and validate dependencies instead of relying on memory of who last edited Setup.
If two administrators edit the same flow for different requests, the last saved version may omit the other change even though both tickets look complete. Assign component ownership for the release and capture each intended change in the shared review artifact. Compare against the agreed source before deployment and make emergency production edits visible to the team. A process is effective only if the next sandbox or deployment includes those emergency changes instead of accidentally replacing them.
100. How would you prioritize an administration backlog?
Compare business impact, risk, affected users, dependencies and maintenance cost with stakeholders. Separate defects from enhancements and record tradeoffs so the loudest request does not automatically determine the next release.
A small access defect preventing the service team from closing cases may outrank a visually impressive dashboard redesign. Conversely, a quick workaround that grants broad data access may carry a risk larger than its apparent effort. Explain the consequence of deferring each item and identify dependencies that unlock several requests. Reserve capacity for incidents and platform maintenance, then revisit priority when evidence changes rather than allowing an old backlog ranking to become unquestionable.
101. How would you turn a vague request into a testable requirement?
Ask for the actor, triggering event, desired outcome, exceptions and examples of success and failure. Write acceptance criteria before choosing an object, field or flow.
For “notify managers about stalled deals,” define which opportunity stages qualify, how long without activity counts as stalled, who the manager is and whether repeat reminders are allowed. Ask what happens when ownership changes or the opportunity closes before the scheduled action. Those examples expose the decision rules before a flow is built. Agree the observable outcome and negative cases with the requester so acceptance does not depend on whether the automation merely ran without an error.
Case | Expected business outcome |
|---|---|
Open deal crosses the agreed inactivity threshold | Notify the designated manager once for the defined event. |
Deal closes before the scheduled check | Do not send an obsolete stalled-deal reminder. |
Owner changes before the check | Apply the agreed manager-resolution rule. |
102. How would you measure user adoption?
Combine meaningful usage patterns with task completion, data quality and user feedback. Login counts alone do not show that users can complete the intended process or trust its data.
For a new case workflow, compare eligible cases with cases completed through the intended process, then inspect missing required data and common abandonment points. A rise in logins could reflect confusion or repeated failed attempts. Interview a small representative group and observe where they leave Salesforce to finish the task. Separate lack of training from a design defect: neither a dashboard of usage nor a reminder campaign will repair a workflow that cannot handle a common exception.
103. How would you decide between configuration and custom development?
Evaluate capability fit, maintainability, complexity, limits and available support skills. Prefer a design the team can test and operate, and involve developers when declarative workarounds become more fragile than a supported coded solution.
A flow that clearly expresses a modest business process may be easier for the admin team to support than custom code. But a growing network of interdependent flows with complex transaction behavior can become harder to reason about than a developer-supported design. Estimate testing, monitoring and future change costs along with build effort. The administrator’s responsibility is to articulate the requirement and operational constraints, then collaborate on the design rather than treating “no code” as an acceptance criterion.
104. How would you assess a managed package before installation?
Review its purpose, permissions, data access, vendor support, dependencies and lifecycle in an appropriate test environment. Plan removal and data ownership implications rather than assuming marketplace availability eliminates operational risk.
A package may need broad permissions or introduce automation on an object the org already customizes. Test it against existing flows, access personas and a representative data volume before proposing installation. Confirm who owns data created by the package, what happens at license expiry and what an uninstall would remove or prevent. Marketplace review is useful context but does not establish compatibility with this org’s processes. Record vendor support expectations and the administrator’s own support responsibilities.
105. How would you prepare for a Salesforce release?
Review applicable release changes, identify automatically enabled behavior and test critical processes in a suitable preview environment when available. Plan communications and targeted regression checks around the org's actual features.
Build a release impact list from features the org actually uses, then map each relevant change to a critical business journey and an owner. An automatic behavior change deserves different planning from an optional feature that remains disabled. Test in a preview environment when available and appropriate to its instance timing, using representative user permissions and integrations. Communicate observed effects and required action, rather than forwarding every release note and expecting business users to identify the risks.
106. How would you handle an undocumented legacy customization?
Trace its dependencies and observed behavior, interview process owners and capture representative tests before changing it. An unfamiliar component should not be removed merely because its purpose is unclear.
An old formula may feed a report, an integration mapping and a flow decision even if nobody edits it directly. Trace references and compare representative records before deciding it is unused. If ownership is unclear, locate the business process that consumes the result and record the uncertainty. A staged deprecation with monitoring can be safer than immediate deletion when supported by the component’s behavior. Preserve the previous definition and a recovery path before removing a dependency.
107. How would you audit dashboard access?
Review folder sharing, running-user or dynamic context, source reports and the sensitivity of displayed metrics. Test as representative viewers, because access to a dashboard and access to its source report are not the same question.
A dashboard using a privileged running user may expose aggregate information beyond the viewer’s normal record visibility. Conversely, a viewer opening the source report may see a different result because the execution context changes. Check the intended audience, folder access, running context and refresh timestamp together. A dynamic dashboard can align data with the viewer’s access where the edition and configuration support it, but it is not a substitute for reviewing what the audience should learn from the metrics.
108. How would you govern metric definitions?
Assign an owner and document the population, time basis, filters and calculation for each important metric. Reconcile examples across reports so similarly named dashboards do not silently measure different things.
Two “pipeline” reports can disagree because one excludes closed opportunities and uses current ownership while another uses a historical snapshot and a different currency basis. Define those choices explicitly and give the metric a business owner who can resolve disputes. Use a handful of traceable example records to show why each is included or excluded. Version a material definition change and communicate it; an unexplained improvement in the chart may represent a filter change rather than better performance.
109. How would you troubleshoot a timezone-related reporting discrepancy?
Identify the field type, storage and display behavior, user timezone and report filter boundaries. Reproduce with records near the boundary rather than adjusting dates until totals happen to match.
An event near midnight can fall on different local dates for users in different timezones, while a Date field represents a date without the same time-of-day conversion. Compare the stored field type, each user’s timezone and the report’s date-filter behavior using a boundary example. Include daylight-saving transitions when relevant. Do not “fix” the source timestamp by adding hours simply to match one dashboard, because that can corrupt the underlying event for other users and integrations.
110. How would you design a consent-related configuration change?
Work from approved policy and applicable expert guidance, then model scope, status, evidence and downstream enforcement. A checkbox alone does not establish that every communication or integration honors the intended permission.
For example, distinguish consent for a particular communication purpose from a generic contact-level flag, and retain the evidence and effective status required by approved policy. Identify every sending system that consumes the decision and the delay before a change propagates. Test withdrawal and conflicting source updates as carefully as initial consent. The administrator implements and verifies the approved rule; legal applicability and retention periods require the responsible specialists, not an inference from a Salesforce field label.
111. How should a production data repair be tested?
Use a narrowly defined record set, representative before-and-after examples and a reversible plan where possible. Confirm downstream automation and reconciliation before applying the approved repair at scale.
If a repair restores an incorrect picklist value, capture the exact record identifiers and prior values so reviewers can evaluate the proposed change. Test records with different owners and states, because validation rules or flows may affect only some of them. Confirm the repair does not send unintended messages or overwrite legitimate edits made after the incident. Reconcile input, successful updates and failures, then review a sample of downstream results before declaring the incident resolved.
112. How would you manage an automation bypass for a migration?
Define the narrow migration need, approved users or context, duration and audit evidence, then test normal behavior still applies elsewhere. Remove the exception after use and reconcile any safeguards the bypass intentionally skipped.
A bypass intended for a controlled backfill should not disable validation for every user or become a permanent integration entitlement. Where the org’s design supports it, use an explicitly checked, narrowly assigned capability and document which rules it bypasses. Verify that unprivileged users still encounter the normal safeguards and that bypassed records are reconciled afterward. Include removal in the migration completion criteria; a successful load alone does not prove the org is back in its normal control state.
113. How would you respond when a user reports incorrect results after a release?
Capture the user context, exact record or report and expected outcome, then compare with the changed components and prior behavior. Contain confirmed impact while preserving evidence and avoid speculative changes to unrelated configuration.
If one user’s opportunity total changed after a release, reproduce the exact report, filters, access context and data-refresh state before assuming the formula is wrong. Compare another affected and unaffected user to distinguish a permissions change from a calculation change. Freeze or revert the relevant behavior only when the impact and recovery path are understood. Keep the reporter informed about known scope and remaining uncertainty instead of promising a fix based on a screenshot alone.
114. How would you assess administrator documentation quality?
Check whether another qualified person can understand the requirement, reproduce the test and support or recover the feature. Useful documentation explains why the configuration exists and its dependencies, not only screenshots of where to click.
Ask another administrator to explain why a flow has a particular exception and how to disable or recover it safely. If the documentation shows only canvas screenshots, it may omit the business rule, activation state and related automation needed to answer. Good documentation includes representative acceptance cases, dependencies, responsible owners and known limitations. Keep sensitive credentials out of the runbook and point to the approved access process; operational usefulness does not require copying secrets into a shared document.
115. How would you manage data retention requirements?
Translate approved retention rules into classified data scope, ownership and a tested operational process, with appropriate legal or privacy input. Consider copies, exports and integrations; deleting records from one object may not satisfy the whole requirement.
A retention rule for closed service cases may also affect attachments, downstream exports and reporting snapshots. Determine which copies are in scope, which owner executes each action and whether a legal hold changes the permitted operation. Test selection and recovery assumptions on a safe sample before scheduling deletion. Verify the actual product capabilities and policy requirements; an administrator should not equate removing a visible record with certified erasure across every backup and connected system.
116. How would you plan training for a new workflow?
Teach the user's task, decision points and recovery from common mistakes using realistic safe examples. Collect feedback and verify task completion rather than assuming an announcement email produces adoption.
For a new lead-routing process, teach how a user recognizes an assigned lead, what information must be completed and where to send a misrouted example. Include a realistic exception, such as an existing customer submitted as a new lead, rather than demonstrating only a perfect record. Have users complete a safe task and observe errors before rollout. Feed recurring confusion back into configuration decisions; training should not permanently compensate for labels or required steps that are intrinsically misleading.
117. How should an administrator handle an AI-generated configuration suggestion?
Treat it as a proposal that needs documentation checks, dependency analysis and testing in a suitable environment. Verify access and failure behavior, and do not paste credentials or sensitive org data into an unapproved tool.
An AI suggestion may reference a setting that belongs to another edition, invent a permission or omit how a flow behaves on repeated updates. Check the named feature in current official documentation and translate the suggestion into explicit expected outcomes before testing. Inspect negative cases such as unauthorized access or partial failure, not just the generated happy path. Preserve human ownership of the change request and release decision so a plausible explanation never becomes the only evidence that the configuration is safe.
118. How would you explain an edition-dependent limitation?
Identify the actual org's licenses and enabled capabilities, verify current documentation and offer supported alternatives with tradeoffs. Avoid presenting a remembered limit as a universal platform fact.
If a stakeholder asks for dynamic dashboards, verify whether the org’s edition and allocated capabilities support the requested use and volume before designing the solution. Explain what the limitation prevents and whether an existing supported reporting approach meets the real requirement. Avoid promising that buying a license alone resolves every dependency; permissions and configuration can also matter. Record the checked source and org facts so the decision can be reassessed when licensing or the product changes.
119. What should an administrator handover contain?
Include access ownership, architecture and automation maps, release procedures, known issues, backups and operational contacts. Provide recoverable artifacts and tested runbooks so support does not depend on one person's account or memory.
The incoming administrator should be able to identify who owns an integration, locate the approved credential-management process and respond when its authentication fails. Include pending releases, active exceptions and recurring operational work alongside the architecture map. Verify access using the successor’s authorized identity rather than handing over a shared password. Rehearse one recovery or support scenario before sign-off: the ability to find a document is weaker evidence than the ability to use it to restore the intended process.
120. What demonstrates senior Salesforce administrator judgment?
A senior administrator connects business outcomes to secure, maintainable configuration and reliable operations. They diagnose effective access, control change scope, test adverse cases and explain when a requirement needs architectural or specialist input.
Consider an urgent request to make all opportunities visible because a manager cannot see one deal. A senior administrator traces the missing access, clarifies the manager’s legitimate scope and chooses a targeted correction with negative tests. They also identify whether the incident reveals an ownership or role-maintenance problem likely to recur. The judgment is in balancing service restoration, data exposure and future supportability, and in explaining what evidence would make them change the proposed solution.
Salesforce Admin Documentation and Further Reading
These official Salesforce references support the administration concepts discussed here. Confirm feature behavior, licensing and limits in the target org before implementation.
Similar Posts
