Preparing calibration records for an ISO/IEC 17025 audit

What the assessor will trace

Assessors rarely read records front to back. They sample: a handful of certificates, each followed back to its evidence. For every one they expect to find, without anyone having to explain:

1. The readings as first recorded, before and after any adjustment.
2. The reference standard used, and proof it was in calibration on that day.
3. How that standard is traceable, through its own certificate.
4. The procedure followed, and the conditions at the time.
5. How the uncertainty on the certificate was arrived at.
6. Who did the work, who checked it and that they were authorised to.

A gap at any step is a finding, even if the calibration itself was done well. The rest of this guide takes the steps in the order the standard sets them out. Clause numbers are from ISO/IEC 17025:2017 and the summaries are ours, so read the clause itself before changing a procedure.

Technical records: clause 7.5

The technical record for a calibration has to hold enough to repeat it under conditions as close as possible to the original. In practice that means:

• Original observations, recorded at the time they are made and identifiable to the job.
• The date, and the identity of the person responsible for each activity and for checking the results.
• The equipment used and anything else that affects the result or its uncertainty.

The second part of the clause is about changes. An amendment has to be traceable to the earlier version or the original observation. Both the original and the amended data are kept, with the date of the change, what was changed and who changed it.

Where labs get caught: readings jotted on scrap and copied out neatly later, sheets with no technician name, and spreadsheets where a corrected cell leaves no trace of what it said before.

Equipment records: clause 6.4

For each reference standard and any other equipment that influences results, keep a record with:

• Its identity: type, manufacturer, serial number or other unique identification.
• Calibration dates and results, any adjustments, the acceptance criteria and the next due date or interval.
• Its location, and evidence it was verified as fit for use before going into service.
• Maintenance carried out, and any damage, malfunction, modification or repair.

Where labs get caught: a standard used a few days past its due date. One overdue standard puts every calibration made with it in question, so the assessor will ask what you did about those certificates.

Traceability: clause 6.5

Each result has to link to the SI through an unbroken chain of calibrations, each contributing to the uncertainty. The record that proves it is the certificate for your reference standard, from a competent laboratory, on file and matching the serial number you used. Your own certificate then names that standard.

Where labs get caught: a calibration record that says which type of standard was used but not which one, or a standard whose certificate cannot be found.

Certificates: clause 7.8

The certificate is the part of your records the customer sees, and the part assessors check first. Common gaps are a missing date of issue, uncertainty missing or in the wrong unit, a pass or fail with no decision rule, and a calibration due date the customer never asked for.

The free certificate checklist goes through clause 7.8 item by item, and the uncertainty budget calculator produces a table you can keep as the evidence behind the uncertainty you report.

Keeping and protecting records: clauses 8.4 and 7.11

Records have to stay legible and be identified, stored, protected, backed up, retrievable and kept for a retention period you have defined. Where they are held on a computer, the system has to be protected from unauthorised access and from tampering or loss, and changes have to be recorded.

Where labs get caught: one field notebook that is the only copy, a database on a single office computer with no backup, and a shared login that makes it impossible to say who entered a result.

A plan for the weeks before the audit

1. Six weeks out. List every reference standard with its due date and locate its current certificate. Send anything due before the audit for calibration now.
2. Four weeks out. Pick ten issued certificates at random, across technicians and instrument types. Trace each one back through the six steps at the top of this page and write down what you could not find.
3. Three weeks out. Fix the cause of each gap in the template or procedure, not only in the ten records. Raise the gaps you cannot fix as nonconformities yourself. A lab that found its own problems is in a better position than one that had them found.
4. Two weeks out. Check that each technician's authorisation covers the work they sign for, and that it is on file.
5. One week out. Practise retrieval. Ask someone to request a record from last year by instrument serial number and time how long it takes.

What software changes

None of this requires software. What a system built for calibration does is remove the ways records go missing: a required field cannot be left empty, a reading is stamped with who entered it and when, a correction keeps the original, an overdue standard cannot be selected, and a record from two years ago is a search away. In Probatus the audit package gathers the calibration records, reference standards and instrument register for a date range into one PDF.

For what to look for in a system, see calibration software for small labs.