In pharma, SPC has a regulatory name: continued process verification. The FDA's 2011 process validation guidance splits the lifecycle into three stages — process design, process qualification, and continued process verification — and Stage 3 is the standing obligation to keep demonstrating that the process remains in a state of control during commercial manufacture. EU GMP Annex 15 and the ongoing process verification expectations in Annex 1 and ICH Q10 say the same thing in European.
Nothing in those documents says "control chart". Everything in them describes one. A trended parameter, statistically evaluated, with predetermined criteria for when a trend needs investigating, and evidence that somebody investigated it.
A batch attribute, charted across a validation baseline
Batch
Limits frozen on the 18 batches of the validation baseline. From batch 19 — a new excipient lot — the mean steps up by about 2 mg and stays there: batch 24 crosses the upper control limit, and the nine-in-a-row shift rule completes at batch 27.
Every batch is comfortably inside the ±5% specification, the baseline Cpk is 3.6, and every one of these batches would pass release testing. Continued process verification asks a different question — is this still the process that was validated — and from batch 19 it is not.
Illustrative data — generated to behave like this process, not taken from anyone's plant. The limits, the signals and the capability indices above are computed at page load by the same engine that draws a customer's chart.
Every batch on this chart releases. The specification is 250 mg ±5%, the process sits comfortably inside it, and the certificate of analysis for every one of these batches is clean. That is exactly why the chart matters: Stage 3 is not asking whether the batches conform. It is asking whether the process is still the process you validated.
From batch 19 it is not. A new excipient lot has moved the mean by about 2 mg and it has stayed there. On a chart with frozen limits that is a run of points on one side of the centre line — an unambiguous, documented, dated signal you can attach to a deviation and a change control. On a spreadsheet whose limits recompute with every added batch, the mean quietly moves and the chart never says a word.
The other thing worth staring at: the tolerance lines are drawn here because this is an individuals chart, where the points are the measured units. Never draw them on a chart of subgroup averages — averages vary less than the parts they came from by a factor of √n, so a chart that looks comfortably inside tolerance can be produced by a process making out-of-tolerance units. Our chart generator will show you that on your own numbers.
What gets trended
- Solid dose — tablet weight, hardness, thickness, friability, disintegration, content uniformity, granulation moisture (LOD).
- Sterile fill — fill volume or weight, container closure integrity, environmental monitoring counts, particulate counts.
- API and bulk — yield, assay, impurity profile, residual solvents, reaction time and temperature.
- Utilities — WFI conductivity and TOC, compressed-air particle counts, cleanroom differential pressure.
- Cleaning validation — swab and rinse residues, batch after batch.
Two of those lists behave completely differently, and it matters. Batch attributes give you one point per batch, which is an individuals chart (I-MR) whether you like it or not — a subgroup of one, because two tablets from the same batch tell you about within-batch variation, not batch-to-batch. In contrast, EM counts and particulates are counts of rare events, which need a c-chart or, better, a rate chart, and are famously not normal.
Why the usual tooling hurts here
Excel. Almost every CPV programme in a small or mid-size site starts as an Excel workbook maintained by one person in QA. It fails in three specific, documented ways: limits that move with the data, no audit trail on the calculation, and an annual product review that takes three weeks of copy-paste. And an unvalidated spreadsheet performing a GMP calculation is itself an inspection finding.
Full CPV suites. Genuinely good, genuinely expensive, and typically a 12-month implementation with a validation package the vendor writes and you pay for. Correct for a large site with dozens of products. Heavy for one line, one product, and a Stage 3 protocol that is due.
Statistical packages. A statistician's tool, used quarterly, by the one person who knows how to drive it. Nothing is watching between the quarters.
Where this platform fits
- Baseline windows are explicit and frozen. You choose the validation batches, the limits are computed from them, and they are stored with the window, the author and the timestamp. Re-baselining after an approved change is a recorded event with its own history — which is precisely the story a Stage 3 protocol needs to tell.
- Every limit change, tolerance edit and deletion is logged and can be exported. When an inspector asks why batch 41 was judged acceptable, the answer is a record, not a reconstruction.
- Rules with severities. Which rules are active on which parameter is a named, reusable rule set — the "predetermined criteria" the guidance keeps asking for, written down instead of implied.
- Signals have owners. Acknowledge, assign a cause, close. The log is the investigation trail.
- Read-only views for people who should not be able to change anything — production, engineering, a QP who wants the picture rather than the workbook.
What we do not do
Worth being precise about in a GMP context:
- Validation. No SaaS product arrives validated — validation is something you perform, for your intended use, in your environment. What we can supply is the input to it: an IQ/OQ template pack and the calculation-engine test results mapped to the published worked examples they reproduce. The qualification effort is still yours.
- 21 CFR Part 11 signatures. Actions are attributed and timestamped and the change log is exportable, but there is no signature meaning or manifestation and no closed-system controls package. Where signed records are required, this works alongside your eQMS rather than in place of it.
- p, np and u charts. Proportion with a varying denominator is not supported yet, and several EM and rejects applications want exactly that.
- Non-normal capability, transformations, CUSUM and EWMA. Content uniformity and impurity data often ask for these.
- LIMS, MES and batch record integration. Data arrives as CSV or over the API and stays a chart.
In short: a good tool for seeing whether your commercial process is still behaving, alongside the records system you already have.
Where to start
Take one product and one parameter with at least 20 batches of history, export it, and upload the CSV. Pick the validation batches as your baseline window and watch what the last two years did afterwards. It takes about ten minutes, and it is the cheapest possible answer to "is our process still the one we validated?"
The deeper background on what Stage 3 asks for is on the continued process verification page.