# Salesforce Health Check: Security Review and Org Audit Checklist

A Salesforce health check should combine the native security assessment with an operational review of access, data, automation and integrations. Use the built-in score to inspect security settings against a named baseline. Then test the customer workflows the business depends on, with evidence, repair owners and acceptance conditions that explain what still needs attention.

Canonical URL: https://www.martechdevs.com/blog/salesforce-health-check

By [Jafrin Akter Alo](https://www.martechdevs.com/authors/jafrin-akter-alo)

Data Engineer and Co-founder at [Martechdevs](https://www.martechdevs.com) and [Growth and Analytics Partners](https://www.growthandanalytics.com)

[LinkedIn](https://linkedin.com/in/jafrinalo)

Last updated: 2026-11-15

Next step: [CRM workflow and reporting repair](https://www.martechdevs.com/crm-lifecycle). Connect verified findings to owned changes and repeatable acceptance checks.

## Review Scope

Keep the native security result and the broader Salesforce audit visible as separate outputs.

| Review area | Main evidence | Useful outcome | Limitation |
| --- | --- | --- | --- |
| Native Health Check | Selected baseline and setting differences | Prioritized security configuration review | A score does not verify customer workflows |
| User access | Permission sources and role tests | Approved reading and editing scope | An administrator's view is insufficient |
| Data quality | Defined record population and identity rules | Reliable ownership, matching and required values | A cleanup does not repair the producer |
| Automation | Flow version, inputs and resulting actions | Correct completion and error handling | A successful run may have the wrong business result |
| Integrations and reports | Matched records, source values and totals | Explained downstream decisions | Current connectivity does not prove complete coverage |

For an established sales org, begin with the workflows that matter now: enquiry assignment, lead conversion, opportunity progression and customer handoff. A renewal operation may need contract dates, account ownership and billing status instead. Use those workflows to select records and users for the review.

The [Salesforce implementation checklist](https://www.martechdevs.com/blog/salesforce-implementation-checklist) covers an initial release. An ongoing health review should examine drift: old permissions, changed field ownership, automation with no current owner and new business processes that no longer fit the original configuration.

When the findings span several systems, [CRM and lifecycle services](https://www.martechdevs.com/crm-lifecycle) can connect the security, data and workflow repairs into one operating plan. Each change still needs a concrete acceptance test and a responsible owner.

## Native Security Health Check

Salesforce Health Check compares security settings with the Salesforce Baseline Standard or a selected custom baseline. Its score uses a proprietary formula; it is not a simple percentage of settings that pass. Salesforce supports up to five custom baselines, and new signals can affect a score without a corresponding business-process change. [Security Health Check](https://help.salesforce.com/s/articleView?id=sf.security_health_check.htm&language=en_US&type=5).

Record the org, selected baseline, review date, score and individual setting differences. If the next review uses a different baseline, preserve that distinction. Two scores from different baselines do not establish a like-for-like improvement.

Treat the score as a route into the settings. A high result does not establish that the right user can close a deal, that an integration cannot overwrite a contract date or that a report includes every intended record. Those outcomes need their own checks.

Salesforce lists Health Check for Professional, Enterprise, Performance, Unlimited and Developer editions. Viewing it requires View Health Check, View Security Center or Manage Security Center; importing a custom baseline requires Manage Health Check or one of the listed Security Center permissions. [Availability and permissions](https://help.salesforce.com/s/articleView?id=sf.security_health_check.htm&language=en_US&type=5).

### Security Review Steps

1. Open Setup and search for Health Check.
2. Confirm the baseline selected for this review.
3. Capture the score and the settings that differ from the baseline.
4. Assign each material difference to a security or platform owner.
5. Inspect the users, sessions and integrations affected by a proposed change.
6. Rehearse the change in a suitable test environment before its production release.
7. Recheck the setting and the affected workflows after release.

The native page provides Edit and Fix Risks routes for remediation. Salesforce's training guidance recommends considering changes individually because a broad settings change can affect integrations or user access. Custom baselines should be aligned with the organization's security requirements. [Health Check remediation guidance](https://trailhead.salesforce.com/content/learn/modules/security_basics/security_basics_healthcheck).

Do not change a baseline merely to improve the displayed result. If the organization accepts a deviation, record the requirement, affected setting, approver and review date. That makes the exception explainable without pretending it has been remediated.

## Access and Identity Review

Review human users and application identities separately. A sales representative, a support operator and an integration account need different operations, even when they touch the same customer records.

For a selected user, Salesforce's User Access Summary shows assigned permissions and includes Access Granted By to identify the profile, permission sets or permission set groups providing a permission. It does not list every possible access feature, and the field tab excludes fields that do not support configurable field-level security. [User Access Summary](https://help.salesforce.com/s/articleView?id=sf.users_access_summary.htm&language=en_US&type=5).

Use the summary to trace grants, then test the actual operation. The [Salesforce field permissions guide](https://www.martechdevs.com/blog/salesforce-field-permissions) explains how field, record, page and execution-context checks fit together.

| Identity | Expected operation | Negative case | Evidence |
| --- | --- | --- | --- |
| Sales representative | Edit their allowed deal fields | Restricted field update | User, record and field result |
| Sales manager | Review the approved team's pipeline | Unrelated restricted data | Report and record access |
| Integration account | Update owned properties | Unauthorized property write | Request and destination result |
| Administrator | Perform approved configuration work | Unexplained standing privilege | Grant source and responsibility |

For each integration identity, name its business owner, application, authentication route and replacement process. A credential with no human owner is an operating gap even if its current requests succeed.

Before removing access, inspect dependent jobs and scheduled work. Test the planned replacement identity on the approved scope and confirm that error alerts reach an active owner. The acceptance condition is continued authorized operation with the intended restriction in place.

## Configuration History

Salesforce Setup Audit Trail records configuration changes. Its page displays the 20 most recent changes, while Download provides the available setup history for the previous 180 days. Older setup entity records are deleted after that period. Viewing the history requires View Setup and Configuration. [Setup Audit Trail](https://help.salesforce.com/s/articleView?id=sf.admin_monitorsetup.htm&language=en_US&type=5).

Export the relevant history before the review window expires, under the organization's storage and access policy. Connect suspicious or unexplained changes to a change request, owner and intended result. The absence of a recent entry cannot prove an older configuration was never changed.

Separate setup evidence from business-record evidence. A changed permission set and an edited opportunity amount require different histories. Keep the evidence source in the finding so a reviewer understands what it can establish.

For a recent incident, construct a narrow timeline: configuration change, first affected record, reported symptom and repair. Include the actual org and time zone. Similar names across production and test environments can otherwise make an unrelated event appear relevant.

Avoid turning the configuration inventory into a deletion list. An apparently unused field or automation may support an infrequent renewal, an external API consumer or a quarterly report. Confirm its dependencies and responsible team before proposing retirement.

## Data Quality and Duplicate Controls

Choose records that participate in current work. For open opportunities, review owner, stage, close date, amount and required account association. For enquiries, review stable identity, routing inputs and the field that distinguishes a new prospect from an existing customer.

Use record IDs to preserve the population while calculating findings. If multiple issues affect the same record, report both the issue counts and the unique affected count. Do not add defect percentages that overlap.

Salesforce duplicate rules also need a user-specific check. A rule can behave unexpectedly when the updating user cannot access a field used by its matching rule. On edits, the rule runs when the edited fields participate in the associated matching rule. Test the actual record path and identity, not only an administrator's manual creation. [Duplicate rule considerations](https://help.salesforce.com/s/articleView?id=duplicate_rules_overview.htm&language=en_US&type=5).

Select several deliberate cases: an exact known duplicate, a similar but different company, a record with an empty matching field and an update to an existing record. Agree the expected result before changing a rule. Stronger blocking is not automatically better if it stops a legitimate enquiry from reaching an owner.

The [CRM property ownership guide](https://www.martechdevs.com/blog/crm-property-ownership) helps distinguish a bad stored value from a recurring overwrite. If a billing integration owns a renewal date, repairing the Salesforce field alone can be temporary. The audit should identify the producer and the policy needed to prevent recurrence.

Use the [CRM data hygiene audit skill](https://www.martechdevs.com/skills/crm-data-hygiene-audit) for the record-level review, then carry its findings into the broader org register. Keep merges, bulk updates and source-system repairs as separately scoped actions with a before-state and acceptance checks.

## Automation and Integration Review

For each business-critical automation, record the active version, entry condition, input fields, execution context, expected updates and failure owner. Include repeat behavior: a second event or retry should have a defined result.

Salesforce's Process Automation Settings controls flow error email recipients. The configured destination can be the last person who modified the flow or the Apex exception recipients. These emails can contain input data, so the chosen recipients need both operational responsibility and appropriate access. [Process Automation Settings](https://help.salesforce.com/s/articleView?id=platform.automation_about_settings.htm&language=en_US&type=5).

Inspect whether an active person receives the errors. Then select a failed case and trace its version, inputs and resulting records. A visible error can occur after some actions have already taken effect; establish the existing state before replaying work.

For a connector, compare the expected source population with destination records using stable identifiers. Review successes, rejected records, deliberate exclusions and records still waiting. Authentication success answers only whether the connection can operate; it does not establish that the correct records or values were processed.

A hypothetical nightly job expects 500 customer records. The destination contains 470 accepted updates, 18 rejected records and 12 deliberately excluded records. Those categories reconcile to 500, but the 18 rejections still require repair. If only 480 records are accounted for, the remaining 20 are an evidence gap until their status is established.

After repairing a mapping or permission, retry a small affected set and inspect the destination values. Include a second execution to check the intended duplicate behavior. A repeat run should not create another customer merely because the source job was restarted.

For any bulk repair, define the exact IDs, fields and before-state. Use the existing approved data-change process rather than treating the audit as permission to modify everything it discovers.

## Reporting and Operating Coverage

Select the report used for a real business decision. Agree its population, date field, currency and exclusions, then reconcile the underlying IDs to the total. Compare the intended reader's view with the approved result. Keep subscriptions and exports in the audience review because they distribute data beyond the current page.

The [dashboard templates](https://www.martechdevs.com/tools/dashboard-templates) can help organize the review output into outcomes, exceptions and owners. An audit dashboard should distinguish completed checks, open findings and unreviewed areas; those are different measures.

Consider this hypothetical audit plan:

| Measure | Calculation | Result | Meaning |
| --- | --- | ---: | --- |
| Planned checks | Defined review inventory | 120 | Entire agreed scope |
| Completed checks | Evidence collected and assessed | 100 | Reviewed scope |
| Review coverage | 100 / 120 | 83.3% | Work completed, rounded to one decimal |
| Confirmed findings | Completed checks with a defect | 14 | Repair queue at first review |
| Verified repairs | 9 / 14 | 64.3% | Finding closure, rounded to one decimal |
| Remaining findings | 14 - 9 | 5 | Open repair work |

The twenty unreviewed checks remain unknown. The five open findings may also differ in severity, so the closure percentage cannot establish that the org is safe or ready for a release. These are operating measures for the hypothetical review, not Salesforce's native Health Check formula.

Report the highest-impact unresolved issue alongside the counts. A restrictive setting may require planned user communication, while a failed routing flow may need immediate containment. Assign a deadline based on the effect and dependencies rather than a cosmetic score target.

## Repair and Regression Flow

![A named org and baseline lead to settings and workflow evidence. Gaps repeat collection; repairs repeat test-environment security and workflow acceptance. Controlled release returns to production evidence collection.](https://www.martechdevs.com/assets/blog/salesforce-health-check.svg)

Diagram connections:

- Org and review baseline to Settings and workflow evidence. 
- Settings and workflow evidence to Evidence completeness. 
- Evidence completeness to Coverage gap and owner: No. 
- Coverage gap and owner to Settings and workflow evidence. 
- Evidence completeness to Finding and repair scope: Yes. 
- Finding and repair scope to Test environment change. 
- Test environment change to Security and workflow acceptance. 
- Security and workflow acceptance to Configuration repair: No. 
- Configuration repair to Test environment change. 
- Security and workflow acceptance to Controlled release and recheck: Yes. 
- Controlled release and recheck to Settings and workflow evidence

The test branch repeats security and workflow acceptance after a repair. The release branch returns to evidence collection for the production recheck. Record the actual outcome rather than closing a finding when the change is merely scheduled.

## Salesforce Health Check Checklist

- [ ] The org, review period and customer workflows are named.
- [ ] The native security score has its selected baseline and setting evidence.
- [ ] Proposed security changes include integration and user impact checks.
- [ ] Human and application identities have current owners and approved operations.
- [ ] Permission sources and ordinary-user outcomes have been checked.
- [ ] Relevant setup history is preserved within the available window.
- [ ] Data findings use a defined population and stable record identifiers.
- [ ] Duplicate controls include the actual user and entry route.
- [ ] Automation failures reach an active, appropriate recipient.
- [ ] Integration categories reconcile to the expected source population.
- [ ] Important reports agree with records, filters and intended audience.
- [ ] Repairs have before-state evidence, acceptance criteria and a follow-up owner.

## Frequently asked questions

### Does a perfect Salesforce Health Check score establish compliance?

No. The score describes security configuration against the selected baseline. A compliance assessment also depends on the applicable requirements, operating processes, evidence and responsibilities. Use the setting results as one input to that assessment. Keep workflow tests, access reviews and accepted exceptions visible rather than treating the score as a certificate for the entire organization.

### Why might the native Health Check score change without a recent admin edit?

The selected baseline or available security signals can change the assessment. Salesforce notes that new signals can affect the current score. Compare the baseline and individual settings with the previous review before assigning a cause. Preserve the earlier evidence so the team can distinguish a configuration change from a changed assessment basis. [Health Check guidance](https://help.salesforce.com/s/articleView?id=sf.security_health_check.htm&language=en_US&type=5).

### Should every audit finding be fixed before the next Salesforce release?

Use the release's affected workflows and risk to decide. A finding that exposes data or breaks an essential customer action can block the relevant release, while an unrelated cleanup may enter a scheduled backlog. Record the decision, owner and acceptance condition. Avoid closing a finding merely because the team has agreed to work on it later.

### Can a sandbox result substitute for a production health recheck?

A sandbox result supports the planned change, but production can have different data, users, credentials and integrations. After a controlled release, verify the affected settings and approved workflow outcomes in production. Keep the scope small and purposeful. The recheck should confirm that the released configuration and its real dependencies match the acceptance conditions established during rehearsal.

### What should a monthly Salesforce audit summary contain?

Include the reviewed scope, native baseline result, material findings, verified repairs and unresolved evidence gaps. Name the owners and due dates for the remaining work. Add a short explanation of the highest-impact open issue and any accepted exceptions. Keep detailed records and configuration evidence linked separately so the summary remains readable while each conclusion can be checked.

## Sources

1. [Salesforce Security Health Check](https://help.salesforce.com/s/articleView?id=sf.security_health_check.htm&language=en_US&type=5)

2. [Salesforce Health Check remediation guidance](https://trailhead.salesforce.com/content/learn/modules/security_basics/security_basics_healthcheck)

3. [Salesforce User Access Summary](https://help.salesforce.com/s/articleView?id=sf.users_access_summary.htm&language=en_US&type=5)

4. [Salesforce Setup Audit Trail](https://help.salesforce.com/s/articleView?id=sf.admin_monitorsetup.htm&language=en_US&type=5)

5. [Salesforce duplicate rule considerations](https://help.salesforce.com/s/articleView?id=duplicate_rules_overview.htm&language=en_US&type=5)

6. [Salesforce Process Automation Settings](https://help.salesforce.com/s/articleView?id=platform.automation_about_settings.htm&language=en_US&type=5)
