IQ OQ PQ is the three-stage test structure most teams use to show that a computerized system was installed correctly, works as specified, and performs in real use. It is borrowed from equipment qualification, and it fits software only if you change what each stage checks. This post explains what each stage proves for a computerized system, how the July 2025 draft of EU Annex 11 raises the bar on testing, and what an executed script has to contain. For the regulation behind the testing, see 21 CFR Part 11 requirements.
One thing to know first. Neither Part 11 nor the 2011 Annex 11 nor the July 2025 draft uses the terms IQ, OQ or PQ. They ask for validation, qualification and test evidence. The three stages are a convention that organizes that evidence, and you are free to name your stages differently if the evidence is there.
IQ OQ PQ for software, not equipment
With a tablet press, installation qualification checks the machine is bolted down and connected. With software, the same logic applies to something less visible: versions, configuration, environments, accounts and the infrastructure under the application.
| Stage | Question for software | Typical evidence |
|---|---|---|
| IQ | Is the right software, in the right configuration, installed on supported infrastructure? | Version and build checks, configuration compared with specification, platform and patch status, environment and access setup |
| OQ | Does each function work as specified, including at limits and when things go wrong? | Executed test scripts traced to requirements, covering normal, boundary and negative cases |
| PQ | Do the system, procedures and trained people deliver the intended result in the real process? | End-to-end runs of actual workflows by actual users, with real or representative data |
IQ: installed and configured as specified
The draft Annex 11 clause 9.3 says that before any test starts, you should verify that the system and its components are correctly installed and configured to specification and, where applicable, properly calibrated. It adds that operating systems and platforms should be on supported versions with relevant security patches deployed.
For software, an IQ usually checks:
- application version and build against the release you approved;
- configuration settings against the configuration specification, not against someone’s memory;
- operating system, database and platform versions and support status;
- environment separation, so you know whether you are testing in validation, test or production;
- initial roles and accounts, set up before any OQ relies on them;
- connected instruments or interfaces, where installed.
The 2011 Annex 11 principle says IT infrastructure should be qualified, not only the application. Qualify the servers, network and platform once, then reference that record from each application IQ rather than repeating it.
OQ: works as specified, including the hard cases
Annex 11 (2011) clause 4.7 asks for evidence of appropriate test methods and scenarios, and says to consider system parameter limits, data limits and error handling. That is the heart of an OQ. A test that only shows the happy path does not meet it.
An OQ script ties to requirements. If a requirement says “the system shall block release of a result that has not been reviewed,” the test tries to release an unreviewed result and records that the system blocks it. If the test passes without the attempt, the requirement has not been tested.
The draft’s nine test-focus areas for OQ
The draft Annex 11 clause 9.6 expects increased focus on key functional requirements, on functions that make work follow GMP, and on functions that protect data integrity. It lists areas that include, without being limited to, the nine below. Here is what each looks like as a test.
| Focus area | Example OQ test |
|---|---|
| Access privileges | Log in as each role and confirm each can do only what the role matrix allows |
| Release of products and results | Attempt release without the required review or signature |
| Calculations | Enter known inputs and compare with independently calculated results |
| Audit trails | Change a value and check the trail shows who, what, old and new value, when and why |
| Error handling | Submit invalid input and a failed transfer; check the system rejects or flags it |
| Alarms and warnings | Trigger an alarm, acknowledge it, and check the alarm log entry |
| Boundary and negative testing | Test just inside, at and just outside each limit, and with disallowed actions |
| Reports and interfaces | Compare report output with source data; send known data across each interface |
| Restore from backup | Restore a backup and verify the data is accessible and complete |
Audit trail testing connects to audit trail review, where the same trail is used in routine operation.
PQ: the last stage of IQ OQ PQ shows real use
PQ answers a different question from OQ. The functions work, but does the whole arrangement hold together when real users follow real procedures? A PQ runs actual workflows end to end: a sample received, tested, reviewed and released; a batch record executed and signed. The people should be trained on the procedures first, and the procedures should be in their approved form.
Run PQ in the production environment or one you can justify as representative. Include the interfaces, the signatures, the printouts and the backup or archive step where they are part of the process. For a replaced manual process, PQ is also where you confirm the principle that the new system must not lower quality or raise overall risk, which both Annex 11 texts state.
Some teams call this stage user acceptance testing. The name matters less than whether the evidence covers real use.
Executed scripts, evidence and traceability in IQ OQ PQ
The draft Annex 11 sets three expectations for the evidence.
- Approved before execution (9.7). Work follows approved plans, protocols and test scripts, and scripts are detailed enough that a different tester could repeat them.
- Executed scripts and screen captures (9.4). The proof that requirements are met should be completed test scripts, plus screen dumps where they add value.
- Traceable (9.5). Test cases link to individual requirements or specifications, for example through a traceability matrix. The draft says test cases that do not trace to requirements do not meet its qualification and validation expectations.
In an IQ OQ PQ package, a usable executed script records the step, the expected result, the actual result, pass or fail, who ran it, when, who reviewed it, and the reference of any deviation raised. Failed steps become recorded deviations, not erased and rerun. Build the links in a requirements traceability matrix so a reviewer can see every requirement has a test and every test has a reason to exist.
Annex 11 (2011) clause 4.7 also asks for documented assessments of the adequacy of any automated testing tools and test environments. If you automate scripts or test in a separate environment, record why the tool or environment can be trusted.
Finishing the qualification
The draft expects qualification and validation to be completed and reported before the system is approved for use. It allows a conditional approval where acceptance criteria are not fully met only with a documented assessment that the gaps will not affect product quality, patient safety or data integrity, stated in the validation report, with close follow-up. Plan for that route, but do not rely on it.
Scaling IQ OQ PQ to risk
Not every system needs three heavy IQ OQ PQ protocols. The validation master plan should say which tier of system gets which depth. For a configured commercial system, you may rely on vendor testing of standard functions if you reviewed it. The draft says the regulated user stays accountable and must judge whether vendor documentation covers the implemented version. What vendors cannot do for you is test your configuration and your use. FDA’s February 2026 Computer Software Assurance guidance, which covers medical device production and quality system software, discusses scripted and unscripted testing by risk. A drug manufacturer may borrow that logic only with a documented justification. See CSV vs CSA.
The draft’s consultation page, which links the draft text, is on the European Commission site. The draft is not adopted and no date is set, so the 2011 Annex 11 remains the text in force.
The 21 CFR Part 11 Toolkit includes an Installation Qualification Protocol and Report, an Operational Qualification Protocol and Report, a Performance Qualification Protocol and Report and a Test Script and Execution Record, drafted to share one traceable structure. It has 105 templates (68 Word and 37 Excel) for $99.
IQ OQ PQ FAQ
Are IQ, OQ and PQ required by Annex 11 or Part 11?
Not by name. Both texts require validation and test evidence. The three stages are the common way to organize it.
Can we combine IQ and OQ in one protocol?
Yes, if the evidence for each question stays clear and approved before execution. Combining the two is easier to justify for a low-risk system. Keep the installation checks before any functional testing, as the draft expects.
Who should execute and approve the scripts?
Trained testers execute them. The plan should say who reviews and approves. Quality assurance commonly approves protocols and reports, with the system owner and process owner involved. Roles are set in the plan, not in the script.
Do SaaS systems need IQ OQ PQ?
The regulated user still needs evidence for the implemented version and configuration. You may use vendor evidence after reviewing it, then test your configuration and workflows.


