CSV vs CSA
Distinguish computer system validation from computer software assurance without treating them as competing compliance programs.
Watch video and read transcript →Short Videos
Watch a short explanation of a validation or assurance topic. Captions and written transcripts accompany each recording.
20 short videos
Does Every GxP System Need Validation?
Decide validation scope from the actual regulated use and failure consequence rather than from the product label.
Watch video and read transcript →200 Tests Passed. Would You Approve?
Evaluate a validation result from risk coverage, deviations, limitations, and residual risk rather than a test count.
Watch video and read transcript →Scripted vs Unscripted Testing
Choose scripted, unscripted, exploratory, and automated testing according to the risk and behavior being evaluated.
Watch video and read transcript →Can Supplier Testing Be Used?
Use supplier testing without assuming it covers the customer configuration, process, data, interfaces, and intended use.
Watch video and read transcript →Does Part 11 Apply?
Assess 21 CFR Part 11 from the regulated record, electronic reliance, and signature use rather than a generic software checklist.
Watch video and read transcript →Does This Spreadsheet Need Validation?
Assess spreadsheet validation from intended use, calculation and data risk, change exposure, and existing controls.
Watch video and read transcript →Does Every SaaS Release Need Regression Testing?
Choose no impact, targeted confirmation, or broader regression from actual release exposure and remaining uncertainty.
Watch video and read transcript →Is Human Review Enough for AI?
Determine whether human review can reliably detect and correct important AI errors before regulated impact.
Watch video and read transcript →Is 95% AI Accuracy Acceptable?
Judge AI performance from critical error, subgroup behavior, uncertainty, human recovery, and intended use rather than one aggregate score.
Watch video and read transcript →What Changes Require Regression Testing?
Set regression scope from affected GxP functions, dependencies, failure scenarios, controls, and evidence rather than the administrative change label.
Watch video and read transcript →Vendor Validated Does Not Mean Customer Validated
Use supplier assurance without assuming it covers the customer’s intended use, configuration, interfaces, data, roles, and regulated process.
Watch video and read transcript →Backup Passed. Can You Restore the Record?
Distinguish a completed backup job from demonstrated restoration, retrieval, interpretation, and reconciliation of required regulated records.
Watch video and read transcript →Is a Successful Data Migration Just a Record Count?
Verify values, metadata, signatures, attachments, relationships, access, and retrievability instead of relying only on matching totals.
Watch video and read transcript →Does a Low-Code App Need Validation?
Assess the regulated use and locally configured functions rather than assuming the platform label determines validation need.
Watch video and read transcript →When Can Automated Tests Be Used as Validation Evidence?
Establish that automated tests are suitable, controlled, traceable, and capable of detecting the failure relevant to the validation decision.
Watch video and read transcript →User Still Employed. Is the Access Appropriate?
Review authorization against current responsibility, least privilege, segregation of duties, privileged access, and approved exceptions.
Watch video and read transcript →What Should a Periodic Review Actually Review?
Use current operating evidence to explain whether the approved intended use, controls, and validation conclusion remain supported.
Watch video and read transcript →Can You Retire the System After Exporting PDFs?
Determine whether exported records preserve required content, metadata, signatures, audit history, relationships, retention, access, and interpretation.
Watch video and read transcript →AI Model Updated. Do You Need Revalidation?
Assess whether a model or service update changes approved use, failure modes, acceptance criteria, human controls, or previously demonstrated performance.
Watch video and read transcript →For a sequence of lessons, see the Foundation videos or AI Assurance Academy.