QC Consistency — How to Use

• Builds a reference library from qualified historical batches you already trust.
• Compares each incoming sample with that material-specific reference library.
• Shows a graded result together with spectral and numerical evidence.
• Highlights important spectral differences and monitoring regions.
• Keeps a dated QC history for traceability and later review.

• Incoming raw-material screening.
• Batch-to-batch consistency checks.
• Supplier or process-change investigation.
• Stability or retained-sample comparison.
• Routine checks where the question is: does this spectrum look like my qualified material?

Create a separate library for each material or grade that you want to monitor. The reference set should represent normal, qualified variation rather than a mixture of unrelated materials.

• Material name — Use a clear name such as 'Aspirin API' or 'Polyethylene Grade A'.
• Reference batches — Upload historical batches that have already passed your normal qualification process.
• Context — Keep composition and batch notes where they help explain expected variation.

Good practice: Add representative qualified batches over time, but do not add a new batch simply because it failed the current comparison.

The default mode changes with reference-library size:

• Similarity: 1–19 references — Direct spectral similarity and peak-difference evidence.
• PCA: 20–49 references — Adds multivariate modeling of the variation in the qualified reference set.
• SIMCA-style class model: 50+ references — Adds class-model diagnostics for larger qualified reference sets.

Note: These are the default thresholds. A deployment can configure different thresholds, and the workbench shows the active mode.

Do not rely on the headline label alone. Review the evidence that produced it:

• Result label — The wording is mode-dependent; similarity mode can report Highly Consistent, Moderately Consistent, or Inconsistent.
• Spectral overlay — Shows the incoming spectrum against the qualified reference pattern.
• Difference peaks — Shows where the incoming sample differs most from the reference.
• Similarity or multivariate diagnostics — The exact numerical evidence depends on whether Similarity, PCA, or the class model is active.

Strongly consistent result — The spectrum falls within the configured reference criteria. Apply your normal QC and SOP review before disposition.

Borderline or moderate result — Review the overlay, difference peaks, acquisition conditions, and batch context before deciding what to do next.

Inconsistent result — The spectrum falls outside the configured reference criteria. This tells you that the sample differs; it does not by itself identify the cause.