In an IATF 16949 audit, the calibration folder usually comes out clean. The finding comes from the software behind the device. The CMM measurement program has been revised, a new version installed on the test stand’s PC, the SPC tool updated; but there isn’t a single line of record showing the measurement result didn’t change. When the auditor asks, “Which version is running on this station, and when was it last validated?”, quality looks at IT and IT looks at quality. Below we explain the scope, division of responsibility, validation methods and the record to show in the audit, in the order that works in the field.
What does the requirement ask for? The scope of measurement software
Clause 7.1.5.2.1 of IATF 16949 (calibration/verification records) sets two software expectations alongside device calibration: verifying that the software version used for product and process control is as specified, and verifying the production-related software used for product and process control. The clause explicitly states that this also covers software on equipment owned by employees, the customer or a supplier on site. Software running on the machine manufacturer’s own test PC is within your scope too.
The quickest way to determine the scope is the control plan. If there’s software behind every system listed as a measurement or test method in the control plan, that software goes into the validation inventory. A typical list in the field looks like this:
- CMM and optical measurement software: the measurement program (part program), alignment and geometry calculation algorithms, report templates.
- Test stand software: leak tightness, EOL (end-of-line) electrical testing, torque/angle monitoring; any software carrying limit values and pass/fail logic.
- SPC and statistics tools: software that calculates Cp/Cpk, control limits and rule violations, including Excel macros.
- Image processing and error-proofing (poka-yoke) systems: camera-based inspection software and parameter sets.
- Data collection middleware: interfaces that take the measurement result from the device and write it to the MES or SPC database.
The last item is the one most often forgotten. The device measures correctly, the interface misreads the decimal separator: 12,05 mm lands in the database as 1205. Or no unit conversion is done, and inches and millimeters get mixed up. A CSV file moved between two PCs with different regional settings is the classic source of this error. Validation must cover the chain up to the table where the value is finally stored.
Division of responsibility between IT and quality
The most common problem in this work is that nobody owns it. Quality says, “Software is IT’s job”; IT says, “We can’t assess measurement accuracy.” Both are right, and still no record gets created. The model that works: metrological validation sits with quality, version and environment control with IT; the two meet in a single process for the goal of IT compliance in manufacturing.
| Activity | Quality / Metrology | IT |
|---|---|---|
| Software inventory and scope decision | Owner | Support |
| Validation method and acceptance criterion | Owner | Informed |
| Approved version list (baseline) | Approves | Owner |
| Installation, update, backup | Informed | Owner |
| Carrying out the validation test | Owner | Provides the environment |
| Detecting unauthorized changes | Informed | Owner |
IT’s job isn’t to prove the calculation algorithm; it’s to preserve the validated state. Which version is installed on which computer must be known. Unauthorized updates must be blocked. Operating system patches must be tied to the change process, and measurement programs must be backed up. Metrology, for its part, shows that the measurement result is still reliable when the version changes. When the roles are clear, who speaks in the audit is clear too.
Validation methods: reference parts, comparative measurement
Don’t turn measurement software validation into a software testing project. You’re not doing a code review. There’s only one thing you need to show as a user: the software produces correct results the way you use it. For this, the four methods below are used alone or together.
Measuring with a reference (master) part
A reference part whose values were determined with a calibrated system is measured with the new version, and the result is compared with the certificate value. The acceptance criterion is written before the test, taking measurement uncertainty and tolerance into account. On test stands, the same logic works with samples known to be “good” and “bad”: a limit sample, a test part with a deliberate leak. The software must pass the good part and reject the bad part. A test stand that passes the bad sample can’t be used, however new it is.
Comparative measurement (old and new version)
The same set of parts is measured with the old and new versions, and the differences are evaluated statistically. If the difference is within the uncertainty limit, the version change hasn’t affected the measurement system. This requires the old version’s environment to stay up for a while. If no image of the station, or at least a backup of the installation and program files, was taken before the update, this method is lost from the start.
Calculation test with a reference data set
In SPC and statistics software, a data set whose results were calculated independently is entered into the software, and Cpk, mean, standard deviation and control limits are compared. The same data set can be reused with every version change. On the CMM side, ISO 10360-6 defines testing geometric element fitting algorithms with reference data sets; a test report in this scope can be requested from the software supplier.
The link to measurement systems analysis (MSA)
Clause 7.1.5.1.1 requires statistical studies for each type of measurement system in the control plan. If a software change can alter the measurement system’s behavior (if the filtering or alignment algorithm changed), whether to repeat the MSA is part of the validation decision. This decision must be recorded. “Not considered necessary” is a decision too, with its justification.
Version and configuration control
Validation is valid only as long as the software stays in the state in which it was validated. For this, an “approved configuration” record is opened for each measurement station: software name and version, the hash of the installation file, license/option modules, measurement program files and parameter sets, operating system version and connected device drivers. Don’t skip the driver line; updating a measurement card or serial port driver can also affect the result.
On Windows-based stations, collecting installed software versions and hashes of critical files with PowerShell reduces comparison with the approved list to a few minutes’ work:
# Installed software and versions (64-bit and 32-bit registry keys)
$paths = 'HKLM:\Software\Microsoft\Windows\CurrentVersion\Uninstall\*',
'HKLM:\Software\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\*'
Get-ItemProperty $paths |
Where-Object DisplayName |
Select-Object DisplayName, DisplayVersion |
Sort-Object DisplayName |
Export-Csv "C:\Validation\$env:COMPUTERNAME-software.csv" -NoTypeInformation -Encoding UTF8
# SHA-256 hashes of measurement program and parameter files
Get-ChildItem 'D:\MeasurementPrograms' -Recurse -File |
Get-FileHash -Algorithm SHA256 |
Select-Object Path, Hash |
Export-Csv "C:\Validation\$env:COMPUTERNAME-hash.csv" -NoTypeInformation
If you take this output periodically (monthly or every calibration cycle) and compare it with the approved list, unauthorized changes become visible on their own. When a difference shows up, the first question isn’t technical but a quality question: when did the change happen, and which parts have been measured on this station since then? The answer feeds into the nonconformity assessment.
The measurement programs themselves (CMM part programs, test recipes) must also be put under version control. Editing programs on the shop floor without a revision number, the person who changed it, the reason for the change and an approval record is a gap you may come across in audit preparation. An operator changes the approach direction of a point, saves it and tells nobody. The change management in production systems guide, which covers all program and recipe changes in production, explains this flow in detail.
Revalidation triggers
Validation isn’t done once and shelved. If which event requires revalidation is written down in advance, the decision doesn’t change from person to person. The table below is an example decision framework; adapt the tiers to your own risk assessment.
| Trigger | Validation required |
|---|---|
| Major version upgrade | Reference part + comparative measurement; MSA decision |
| Minor version / bug-fix package | Reference part measurement; review of release notes |
| Measurement program or limit change | Reference part for the relevant characteristics; approval |
| Operating system upgrade | Function test + reference part |
| Security patch | Function test; record of the result |
| Hardware change (PC, card, sensor) | Function test + reference part |
| Data interface change | End-to-end data comparison |
| Detection of a hash/version deviation | Nonconformity assessment + full validation |
Engineering changes are triggers too: 7.1.5.2.1 requires revisions following engineering changes that affect measurement systems to be recorded. When a new characteristic is added or a tolerance changes, the measurement program is updated. That update requires its own validation. Don’t underestimate the “security patch” row in the table either: some patches change .NET or driver components, and the measurement software may behave differently the next time it starts.
Sample validation plan and record
The validation record is the first document the auditor asks for. Keep it short, but traceable. The fields below are the skeleton of a sample record prepared for test stand software:
- Identification: station/device no., the related characteristic in the control plan and the operation number.
- Software: name, previous and new version, installation file hash, measurement program revision.
- Trigger: change record number and reason.
- Method: IDs of the reference parts/samples used and their calibration certificate numbers, the comparison method.
- Acceptance criterion: a numerical or pass/fail criterion written before the test.
- Results: raw data attachments, comparison table, deviations.
- Decision and approval: the release decision, signatures of the metrology and process owners, date.
- IT closure: the date the approved configuration list was updated and the new hash values.
The annual validation plan keeps the inventory and periodic checks together: station, software, approved version, last validation date, next periodic check and owner. Linking the plan to the calibration calendar makes the job easier. When the device goes for calibration, the software version and hash check are done on the same work order. A separate calendar means a separate thing to forget.
Weak points commonly seen in audits
- There’s a validation record, but the acceptance criterion was written after the test, or never written at all.
- There’s an approved version list, but it has never been compared with the actual version on the shop floor.
- The machine manufacturer’s test PC was left out of scope as “their responsibility”.
- The formulas in the Excel file used for SPC aren’t locked, and the file has no version history.
- Updates made over a remote maintenance connection weren’t reflected in the change record.
- There’s no backup of the measurement programs; if a station fails, nobody knows which revision to restore.
The remote maintenance item needs particular attention. The machine manufacturer connects at night and updates the software, and the morning shift carries on measuring without knowing about the difference; the approved configuration has been broken without anyone knowing. Opening remote access only on request and with approval, and recording the session, keeps this risk under control. For technical methods, see the machine manufacturer remote access guide.
Measurement software validation is part of the evidence IT prepares for the IATF audit; the overall evidence set is in the article IT evidence in the IATF 16949 audit. Showing that measurement results are stored unaltered in the MES and SPC database is a separate task, explained in the article MES and CMMS record integrity. If you’d like to build the inventory and responsibility model together, take a look at our IATF 16949 IT requirements and audit preparation and change, configuration and patch management services. For the current text of the standard and its interpretations (SIs and FAQs), go by the IATF Global Oversight website.
Frequently asked questions
Must measurement software validation be repeated from scratch with every update?
No. A tiered approach based on the trigger is enough. A major version upgrade calls for a reference part and comparative measurement; for a security patch, a function test and a record are enough in most cases. What’s required is that these tiers are written down in advance and every decision is recorded.
Who is responsible for validating test software that belongs to the machine manufacturer?
IATF 16949 7.1.5.2.1 states that verification also covers software on supplier-owned or customer-owned equipment on site. You can do the test together with the manufacturer; responsibility for the record and the control of the approved version lies with the organization doing the production.
Do the Excel files used for SPC fall within the validation scope too?
They do if they’re used for product or process control decisions. Formulas are tested with a data set whose results are known, cells are locked, the file is put under version control, and where the approved version lives is clear.
Is the software supplier’s validation certificate enough?
Not on its own. The supplier’s report shows the general accuracy of the algorithm. You need to do your own reference part or comparison test for your measurement program, your parameters and your data transfer.