Guide

How to move from CSV to CSA

Moving from computer system validation (CSV) to computer software assurance (CSA) means changing your procedures, your risk method and your evidence rules so that testing effort follows patient and product risk. You still validate every GxP system for its intended use, and you stop spending the same effort on a report layout as on an electronic signature. In my experience the set-up takes a few months and the habits take a year or more to settle, because they live in SOPs, templates and reviewers' heads.

Last checked: 12 November 2026 · By Sachin Bhandari

I've led paperless validation across 40+ sites, and the lesson from that work applies here too: a new method sticks when the procedures, the tools and the reviewers all change together. Change one and leave the others, and people drift back to what they know. If you want the comparison first, read CSV vs CSA: what changes.

The regulatory footing

FDA's guidance "Computer Software Assurance for Production and Quality System Software" came from CDRH and CBER as a draft in September 2022. It was finalised September 2025, updated February 2026 to align with the QMSR, and it's now titled "Computer Software Assurance for Production and Quality Management System Software". Its stated scope is medical device production and quality management system software under 21 CFR Part 820.

Pharma applies the same principles through ISPE's GAMP 5 Second Edition, published in July 2022. Your obligations under 21 CFR Part 11, 21 CFR 211.68 and EU GMP Annex 11 don't change. The FDA guidance contains no percentage targets for effort reduction, so don't put one in your business case as if it came from FDA. Base your case on your own baseline. My cost of poor quality calculator and time-to-market calculator help you put a first number on it.

Step 1: Assess where you are

Start with an honest picture of current practice. I'd pull 3 recent validation packages, one for a low-risk system, one for a medium-risk system and one for a high-risk system, and lay them side by side.

Look for these signs. Are the 3 packages roughly the same size? Do testers take screenshots of every step? Are supplier test results ignored and then repeated? Does the risk assessment change what gets tested, or does it sit at the front of the binder and change nothing? How long does a typical change to a configured system take from request to release?

Write this up in 2 pages. It becomes your baseline, and you'll need it in step 7.

Step 2: Fix the procedures

This is the step most teams want to skip, and it's the one that decides whether the change survives an audit. Your validation master plan, your CSV SOP, your risk assessment SOP, your supplier assessment SOP and your test execution SOP all need to say what CSA allows.

The SOPs should define how you classify risk, which assurance activities fit each risk level, what counts as acceptable evidence (including system logs and audit trails), how you use supplier evidence, and who approves the risk decisions. If the SOP still requires a scripted test and screenshot for every requirement, nobody can use CSA without breaking procedure.

Keep the SOP changes under your normal change control, with QA approval. An inspector will want to see when your approach changed and why.

Step 3: Classify by risk

FDA's guidance splits software functions into 2 groups: high process risk, where a failure to perform as intended may result in a quality problem that foreseeably compromises safety, and not high process risk. GAMP 5 Second Edition gives you more detail on assessing impact on patient safety, product quality and data integrity.

I'd classify at the function level, inside each system. Inside a LIMS, result calculation, specification checks and result approval are high risk, while the instrument list screen and the report header carry much less.

Make the classification a short, repeatable exercise: intended use, list of functions, risk per function, reason. If 2 trained people classify the same function and land in different places, your criteria need tightening.

Step 4: Choose assurance activities

Once a function has a risk level, the assurance activity follows from your SOP. A typical mapping looks like this:

Intended useList the functionsCould a failureforeseeably compromisesafety?YesNoHigh process riskScripted tests,expected results,negative testsNot high riskUnscripted tests,short record ofwhat was coveredPlatform behaviour alreadytested by an assessedsupplier: their evidenceEvidence from the systeminto one summary report
Figure 1. The CSA decision flow I teach. FDA frames the question as high process risk or not; the supplier route depends on a documented supplier assessment.

Record the result in proportion to the risk. For a lower-risk function, a line saying what was tested, by whom, when, and the outcome is often enough.

Step 5: Pilot on 2 systems

Pick 2 systems for the pilot. I'd choose one configured off-the-shelf system with a reasonably mature supplier, say a SaaS document management system, and one system closer to the product, such as a LIMS module or an MES change. The first shows you the supplier evidence route. The second tests whether your risk method holds where the stakes are higher.

Run both through the new SOPs end to end, and have QA review them with the same rigour they'd apply to any release. Keep a log of every place the SOP was unclear or the template didn't fit. Expect to revise the SOPs once after the pilot, and plan for it.

Step 6: Train the people who do the work

Train in 3 groups with different needs. Validation engineers and testers need practice in risk classification and in running and recording unscripted tests. QA reviewers need to know what "enough evidence" looks like now, because a QA reviewer trained on CSV will reject a lean package by reflex. System owners and IT need to know their part in intended use, supplier management and change impact.

Use your own pilot packages as the training material. People learn faster from their own systems than from a generic slide, which is how I run my in-house CSA training.

Step 7: Measure

Go back to the baseline from step 1 and measure the same things: package size by risk level, time from change request to release, number of test cases on high-risk functions versus low-risk ones, and review cycles per package. Add one quality measure, such as defects found after go-live or deviations linked to computer systems.

You want to see effort shift towards the high-risk functions, and total time on low-risk changes come down. If high-risk test effort fell too, look again at your classification.

Who owns what

RoleWhat they own in the move to CSA
Head of QA or Quality ITSponsorship, the decision to change the approach, approval of the revised SOPs
Validation leadThe risk method, the SOP and template rewrite, the pilot, and the training plan
QA reviewersAgreeing evidence standards, reviewing pilot packages, approving risk decisions
System ownersIntended use statements and confirming function-level risk for their systems
IT and supplier managementSupplier assessments, access to supplier evidence, change notifications from SaaS vendors
Testers and validation engineersRunning scripted and unscripted testing and recording results to the new standard

Write this down in your validation master plan. In an inspection, unclear ownership of risk decisions is one of the first things an inspector will pick up.

A 90-day plan

Days 1 to 30. Complete the baseline assessment of 3 packages. Agree the sponsor and the working team. Draft the revised validation master plan and SOPs, and the risk classification criteria. Start supplier assessments for the 2 pilot systems.

Days 31 to 60. Take the SOPs through change control and QA approval. Train the pilot team. Run the risk classification for both pilot systems and agree it with QA. Execute the pilot testing.

Days 61 to 90. Review the pilot packages and log every problem. Revise the SOPs and templates once. Train QA reviewers and the wider validation team. Measure the pilot against the baseline, and write a short report for the sponsor with the plan for rolling out to the next group of systems.

1Assess where you are2Fix the procedures3Classify by risk4Choose assurance activities5Pilot on 2 systems6Train the team7Measure the changeDays1-30Days31-60Days61-90
Figure 2. The 7 steps against the 90 days. Step 2 starts in the first month and finishes when the SOPs clear change control in the second.

90 days gets you a working method and 2 real packages. Rolling it across a portfolio takes longer, because legacy systems come into scope as they hit periodic review or change.

What an inspector will ask

Across 300+ inspections, zero critical findings, I've found inspectors ask questions that test whether your approach holds together. Prepare for these:

  1. Show me the procedure that allows this level of testing, and when it was approved.
  2. How did you decide this function was lower risk? Who made that call, and what was the reason?
  3. What is the intended use of this system in your process?
  4. You relied on the supplier's testing here. Show me your assessment of that supplier.
  5. This test was unscripted. How do I know what was covered and what the result was?
  6. Show me the audit trail for this record, and how you know the audit trail itself is reliable.
  7. How do your Part 11 and Annex 11 controls work on this system: access, e-signatures, audit trail review?
  8. What happens when the supplier pushes a release? How do you assess the change?
  9. How do your QA reviewers know what good evidence looks like under this approach?

If you can answer each of these from a document, you're ready. If an answer lives only in someone's head, write it down.

What the regulators in Europe are doing

EU GMP Annex 11 (2011) has asked for risk-based validation and supplier assessment for years, so CSA thinking fits it well. A revised draft Annex 11, and a new draft Annex 22 on artificial intelligence, went to consultation from 7 July to 7 October 2025. On 12 November 2026 neither has been finalised, so keep your procedures referencing the 2011 Annex 11 and watch for the final texts. I keep a dated record on my Annex 22 status tracker, and if AI systems are next on your list, my book on validating AI in GxP carries the same risk logic into models and data.

How I run this with clients

I use a method called DRIVE for this move, and I run it alongside a client's own validation and QA team so the procedures and the habits change together. My CSV to CSA consulting page explains how an engagement is set up. For the underlying discipline, see my guide to computer system validation in pharma.

Frequently asked questions

How long does a CSV to CSA transition take?

Setting up the method and piloting it on 2 systems takes about 90 days if the sponsor is committed and QA is in the room from the start. Rolling it across a full portfolio takes longer, because you'd normally bring legacy systems in at their next change or periodic review.

Do I need to revalidate existing systems under CSA?

In my experience, no. Systems validated under your old procedure stay validated, and I apply the new approach when they next change or come up for periodic review. That's my practice; no regulator text says it in these words, so write your own position into the validation master plan.

Where do I start, procedures or tools?

Procedures. A new test tool or template won't help while the SOP still demands the old evidence. Fix the SOP, then the templates, then think about tooling.

Can we adopt CSA if our QA team is cautious?

Yes, and a cautious QA team helps. Bring QA into the risk criteria and the evidence standard from day 1. When QA has agreed the rules, they'll defend them in an inspection.

Does CSA work for SaaS and cloud systems?

It fits them well, because the supplier does most of the platform testing. The condition is a proper supplier assessment, covering their quality system, release testing, change notification and data integrity controls.

What should we measure to show the transition worked?

Package size by risk level, time from change request to release, and the share of test effort on high-risk functions. Add one quality measure, such as post-go-live defects, so you can show quality held while effort moved.

Train your team on your own SOPs

If you'd like your team trained on CSA using your own SOPs and one of your own systems, the in-house programme is built for that. If you want to talk through where to start, book a 30-minute call.

Sources

Last checked: 12 November 2026. I checked the sources listed below on that date.

  1. FDA, Computer Software Assurance for Production and Quality Management System Software (guidance page, February 2026)
  2. FDA, CSA guidance text (PDF, February 2026)
  3. ISPE, GAMP 5 Second Edition
  4. European Commission, consultation on Chapter 4, Annex 11 and new Annex 22 (7 July to 7 October 2025)
  5. European Commission, draft Annex 22 consultation document
  6. European Commission, EudraLex Volume 4

Related guides and pages