The user requirement specification is the document everyone wishes were already written and no one wants to write. In a process-analytics project it sits between the process owner and every downstream party - the vendor preparing a quote, the engineer drafting the P&ID change, the chemometrician planning the calibration set, the validation lead writing the IQ/OQ/PQ. When the URS is concrete and bounded, those parties converge. When it is vague, they diverge in expensive directions, and the differences only surface during factory acceptance testing or, worse, on the production line.

This guide describes the structure of a URS for an inline or atline process analyzer. It assumes a regulated context (pharma, biopharma, or a specialty chemical line operating under customer-imposed GMP equivalents) and references GAMP 5 (2nd edition, 2022), ICH Q14, and the FDA PAT framework as the underlying regulatory expectations. The same skeleton works for unregulated chemicals work; you simply prune the compliance sections.

Why the URS deserves the effort

A URS that is two pages long is not a URS. A URS that is forty pages long is also not a URS - it is a copy of someone else’s template with the wrong process pasted in. The useful range is ten to twenty pages of dense, specific content, written by the process owner with input from quality, automation, and analytics. The cost of preparing it is one to three weeks of effort; the cost of not preparing it is most commonly a Stage 2 rebuild, which routinely runs ten times that.

The URS is also where the feasibility study results land. If the project skipped feasibility, this is where the gaps in evidence become visible. A URS that requires a sub-1% RMSEP with no prior demonstration that the measurement is even feasible is a URS that has skipped a step.

The required sections

A defensible PAT URS contains the following sections. Each is a contract clause; each must be testable.

1. Process and product description

Two to three pages. What process, what unit operation, what product or product family, what production rate, what existing instrumentation, what existing automation platform. Include a process flow diagram or, at minimum, a sketch of the relevant unit operation with the proposed measurement point marked.

The reason this section exists is that vendors and contract chemometricians do not understand your process. They understand spectrometers and models. The process description is what allows them to size, scope, and price the work correctly. A URS without it produces quotes that are wrong by a factor of two in either direction.

2. Measurement requirements

This is the core technical section. For each measurand:

  • the analyte and units;
  • the operating range and the alert/action limits;
  • the required accuracy and precision, in the same units as the reference method;
  • the required cycle time, specified concretely (sub-second, 1 minute, 5 minutes - never real-time);
  • the reference method and its precision;
  • the criticality classification (critical quality attribute, critical process parameter, monitoring-only).

The accuracy and precision targets must be defensible. Targets pulled from a vendor brochure are not defensible. Targets derived from process control needs - the multivariate statistical-process-control limits, the release specification tolerance, the regulatory specification - are defensible and survive contract negotiation.

3. Probe, sampling, and process interface

Where the probe sits, how it is held, how it is cleaned, how it is removed for maintenance. Wetted-materials list. Operating temperature and pressure envelope. Cleaning-in-place and steam-in-place compatibility. Hazardous-area classification (ATEX zone, NEC class/division). Vibration, ambient temperature at the analyzer cabinet, ingress protection rating.

A surprising fraction of PAT failures trace to this section, not to the spectrometer. A probe specified without a CIP-survival clause is a probe that will be replaced after the first hot caustic cycle.

4. Data, control system, and IT integration

The OT/IT integration is where regulated PAT projects most often slip schedule. Specify:

  • the control system (DCS, PLC, MES) and the integration protocol (OPC UA, Modbus TCP, Ethernet/IP, 4-20 mA + HART for legacy lines);
  • the data tags exchanged, in both directions;
  • the historian, its retention policy, and the raw-spectra archival approach;
  • the time-synchronisation source;
  • the user authentication model and the audit-trail requirements (21 CFR Part 11 or EU Annex 11 as applicable);
  • the network segmentation (the analyzer cabinet typically sits in a Level 2 zone with a documented data diode or firewall to Level 3).

5. Chemometric model lifecycle

The model is part of the validated system, not an afterthought. The URS specifies:

  • the analytical target profile per ICH Q14;
  • the calibration strategy (number of samples, variability sources covered, reference-method assumptions) - see the PCA vs PLS decision guide for context on technique selection;
  • the validation acceptance criteria, referencing ASTM E1655 and USP <1039>;
  • the change-control trigger thresholds (when does a calibration update require revalidation?);
  • the ongoing performance verification schedule;
  • the data and model retention period.

For a deeper treatment of the validation file an inspector wants to see, the article on validating a chemometric model for GMP covers the acceptance-criteria checklist.

6. Quality, regulatory, and compliance

A short section enumerating the regulatory framework, the GMP classification of the equipment, the GAMP 5 software category (typically Category 4 or 5 for PAT systems), and the validation deliverables expected (IQ, OQ, PQ, plus the model qualification documents). A risk-assessment summary per ICH Q9(R1) covers the criticality decisions made in section 2.

7. Maintenance, training, and support

Who maintains the analyzer, with what training, at what frequency. Spare-parts stocking strategy. Vendor support response-time targets. The end-of-life and obsolescence policy - particularly relevant for spectrometers, where firmware and reference-database support windows are short relative to the equipment’s mechanical life.

8. Acceptance criteria

Every requirement in sections 2 through 7 is referenced here in a traceability matrix. Each row: requirement ID, test method, acceptance criterion, responsible party, FAT/SAT/PQ stage. This is the section that converts the URS into a defensible procurement and validation document. Without it the URS is a wish list.

Common failure modes

Three patterns recur:

The vendor-template URS. A URS lifted from a vendor’s sales engineer reflects the vendor’s product capabilities, not the process needs. Symptom: the URS is suspiciously well-aligned with one vendor’s spec sheet. Remedy: rewrite the measurement requirements section from process-control first principles before any vendor is engaged.

The aspirational URS. Accuracy targets that exceed the reference-method precision, cycle times that exceed the spectroscopy physics, environmental tolerances that no commercial analyzer meets. Symptom: the URS reads like a marketing wish list. Remedy: anchor every numeric target to either a feasibility-study result or a published specification with a citation.

The incomplete URS. Sections 1 and 2 are well-written; section 4 (IT integration) is a single sentence saying to be defined. Symptom: the procurement closes on time but the integration scope is renegotiated three times. Remedy: refuse to issue the URS until all eight sections are complete, even at the cost of a delayed kickoff.

Sign-off and change control

The URS is signed by the process owner, the quality lead, the automation lead, and the analytics lead. After sign-off it enters change control: revisions go through a formal review with the same signatories. Loose URS edits - a paragraph rewritten in a hallway conversation - produce the divergence that the URS was meant to prevent.

A well-written URS is short, dense, and frequently amended in its first two months of life. The amendments are signs that the document is being read and used. A URS that goes through procurement and FAT untouched is more likely to have been ignored than to have been perfect.