The audit gap that lives between calibrations.
Most QA programs schedule instrument calibration on a calendar — quarterly, semi-annual, or annual. Between those events, instruments drift. A pH probe slowly polarizes. An HPLC detector ages. A checkweigher belt stretches. By the time the next calibration rolls around, weeks of production data have been generated by instruments that no longer meet their as-qualified specification.
For auditors under BRC Global Standard for Food Safety Issue 9 clause 4.10, SQF Edition 9 module 11, and FSSC 22000 additional requirement 2.5.3, that gap is the question. Not “is the instrument calibrated today?” but “how do you prove it stayed in spec between calibrations?” If your evidence is a printed certificate dated three months ago, the answer is unsatisfying.
The instruments that drift without warning.
Three categories drive most of the audit findings our customers see:
- Release-testing instruments. HPLC columns lose efficiency over hundreds of injections. NIR calibrations drift with lamp age and ambient humidity. Conductivity probes polarize. pH electrodes fatigue. Drift here means a released lot may not have been in spec at the moment of release.
- In-process controls. Weight-checkers drift with belt tension and product build-up. Metal detectors lose sensitivity as the aperture fouls. X-ray inspection drifts with product density. These instruments are the line of defense against a recall — they cannot silently degrade.
- Environmental dataloggers. Temperature and humidity dataloggers in cold storage and clean rooms drift too. A 1 °C drift over six months can invalidate an entire stability study.
Between-shift checks: why the human walk-through is not enough.
The standard answer is “the operator runs a between-shift check.” In practice that means the QC technician pulls a standard, runs it on the instrument, and initials a log. It happens once per shift, on the technician's schedule, on a representative sample. If the drift is monotonic — slow creep toward the edge of the spec — the check passes for weeks before it suddenly fails.
Continuous monitoring changes the cadence. Every sample, every reading, every batch record adds a data point to a trend. The drift curve becomes visible before the spec limit is crossed. The deviation opens in the system, not in the technician's notebook.
Mapping drift events to the standards auditors actually read.
A drift event by itself is noise. What auditors care about is the chain: the deviation was opened, the recalibration action was dispatched, the post-event verification ran, and the lot release decision was signed. That chain maps cleanly onto existing regulatory language:
- BRC Issue 9 clause 4.10 — calibration of measuring equipment. Evidence requirement: continuous or risk-based intervals, not calendar-only. See the dedicated BRCGS compliance page →
- SQF Edition 9 module 11 — calibration and conformity. Evidence requirement: trend records and timely corrective action.
- FSSC 22000 AR 2.5.3 — measurement and monitoring. Evidence requirement: documented calibration, traceability, and records of action taken when results indicate loss of control.
- GAMP 5 — pharma CSV / computer system validation. Evidence requirement: validated systems, audit trails, and electronic-record integrity (21 CFR Part 11 in the US).
When each drift event lands as a signed packet — deviation, recalibration action, post-event verification, lot release decision — the auditor reads it as one continuous record, not five separate screenshots from a CMMS.
Where to start.
For most lines, the highest-leverage starting point is the release-testing instruments — the ones that decide whether a lot ships. Add continuous watch on the pH probe on the bioreactor, the HPLC on the QC bench, the checkweigher at the end of the packaging line. Those three cover most of the audit exposure. From there, in-process controls and environmental dataloggers fall in naturally.
We're onboarding the first pharma and food-manufacturing customers now. The manufacturing vertical is in early access; if your team is feeling the between-calibration gap, the page below is the easiest way to start a conversation.