An instrument is in place, and it no longer does what was expected of it. The audit says where that comes from, and what the recovery starts with.
It works on the installation as it stands, with your data and your teams. It ends in a written verdict and a recovery plan, ordered by what unlocks the most, the soonest.
Four lines of examination
The hardware
Condition of the probe and its window, stability of the source, temperature and dust conditions, and what the manufacturer has itself documented as a limit.
The model
Domain of validity against the material actually received, the calibration set and how representative it is, residuals, out-of-domain samples, the date of the last update. Keeping a model alive.
The installation
What the sensor sees, and whether what it sees represents the batch. The question arises at the measurement point, and it is handled with the theory of sampling.
The organisation
Who reads the value, who is allowed to act on it, what happens while the measurement is unavailable, and what the dossier promises. After the installation.
Three situations, three different outcomes
The hardware is sound, and it is everything else that is missing. An owner for the model, an update, an alarm instruction. This is the most frequent case, and the quickest to put right.
The question was poorly framed at the start. A quantity with no usable signature, a cadence that does not match the process, a measurement point chosen for convenience. Moving the point or changing the quantity followed often reopens the case.
The answer points elsewhere. Another measurement physics answers better, or the question can be settled without an instrument. What bounds an in-line measurement.
What we ask for on your side
Access to the installation, preferably during a production run — an instrument is judged running, not at a standstill.
What the measurement has produced: raw spectra or signals if you have kept them, predicted values, and the matching laboratory results over the same period. The original calibration set and its documentation, if they exist.
A conversation with the people who use it, or who used to — line operation, quality, maintenance. Operators often know before the instrument does, and their reading is part of the diagnosis.
The history of changes: formulation, equipment, material supplier, software version. A measurement that has stopped working almost always has a before and an after.
None of this is blocking: the audit works with what exists, and states what would be missing to go further.
What calls for trials rather than an audit
The audit works on what is already there. When the conclusion is that another quantity, another measurement point or another physics would answer better, the demonstration calls for trials on real material: that is a proof of concept, and it is decided afterwards.
The boundary is worth knowing before starting: the audit says what is blocking, and what recovery would cost. Demonstrating that a new solution works is the role of the trials, and it is what makes their criteria negotiable at the outset.
You have an instrument that has fallen out of use. Tell us what it was meant to measure.
Forty-five minutes are enough to place where the blockage comes from, and to say whether an audit is warranted or the answer is already known.