Skip to content

Available now

COA Validator

Checks a supplier certificate of analysis for completeness, specification alignment, analytical results, methods, sample quantities, dates, signatures and internal inconsistencies, then produces a traceable report for expert review.

7Report sectionsA fixed structure: identity, outcome, decision summary, comparison against official requirements, supplementary evaluation, open findings, scores and sources.
9Fields on every findingRequirement, evidence, outcome, reason, severity, score impact, reference, required action, applicability. A finding missing any of them is not emitted.
25Sources in the controlled registerEach carries its version and DOI, and the version used is recorded on the report.

What it checks

Grouped the way the work is actually done: first whether the document says what it must, then whether what it says holds up.

Completeness

Is the document whole?

Identity of the material and the batch, the issuing laboratory, the dates, the signatures and the authorisations. A certificate that cannot name its own issuer does not pass this quietly.

Specification

Are the limits the right limits?

The limit printed on the certificate is compared with the limit that applies to that material. A printed limit looser than the official one is a finding in its own right, separate from whether the result passed.

Results

Do the results meet the limits?

Each reported result is evaluated against the limit that governs it, with the comparison shown rather than summarised.

Method

Was the right method used?

The analytical method named for each test, and whether it is the one the requirement expects.

Quantities and dates

Do the numbers hang together?

Sample quantities, manufacturing and retest or expiry dates, and whether they are internally consistent.

Consistency

Does the document contradict itself?

Values that appear twice and disagree, totals that do not add up, a conclusion that the rows beneath it do not support.

Documents it accepts

Supported today

  • Certificate of analysis for an active pharmaceutical ingredient
  • Certificate of analysis for an excipient or raw material
  • Certificate of analysis for a finished product batch

PDF or a scan. Stamped and photographed certificates are common and are handled; a page that genuinely cannot be read produces a hard stop that says so, never a report of empty fields.

Not this product

Certificates of suitability, Drug Master Files, GMP records and medical-device technical documentation are separate problems with separate requirements. They are on the platform roadmap and are not part of the COA Validator.

See the platform direction

How a certificate moves through it

The two halves are separate on purpose: the part that reads the document never decides whether it is good, and the part that decides never re-reads the document.

Stage 1

Read

Values are extracted and each one is checked against the page it claims to come from, the cell it claims to sit in and a confidence floor. A value that fails those checks is not passed on.

Stage 2

Judge

Only verified values are evaluated, against requirements drawn from the controlled source register.

Stage 3

Publish

A report that cannot recount its own headline numbers from the rows beneath them is not published at all.

Walk through it step by step

What comes out

One fixed structure, in HTML and PDF, so a reviewer knows where to look without learning a new document each time.

SectionWhat it holds
Report identityWhich document, which batch, which run, and the software and source versions behind it.
Final outcomeOne of four: accepted, conditional, non-compliant, or insufficient data.
Client decision summaryWhat a decision-maker needs, in plain language.
Comparison against requirementsEach test, its printed limit, the official limit, the result, and the outcome.
Supplementary evaluationDocument-level checks that are not a single test row.
Open findingsEverything unresolved, each with the action that closes it.
Scores and sourcesThe scoring, and every source cited with its version.

Outcomes are a closed set

The distinctions carry weight. A result meeting a limit is not the same as a printed limit being at least as strict as the official one. Evidence being absent is not the same as evidence being present and bad. Collapsing either pair gives exactly the wrong answer.

Ambiguity is an outcome too

Where a damaged scan supports two readings that disagree on the verdict, the report names both and refuses a decisive answer rather than picking one silently.

Where a person signs

An automated analysis is a starting point. A qualified reviewer can examine the certificate and the findings, confirm or correct them with cited sources, and issue an Expert-Reviewed Analysis. Corrections are recorded rather than overwritten, so the reasoning behind a change stays visible.

Limitations, stated plainly

A compliance tool that hides its edges is worse than one that names them.

  • It analyses the document in front of it. It cannot tell you whether the document is genuine, or whether the laboratory that signed it did the work.
  • It does not test material. Nothing here replaces a laboratory result.
  • It works from a controlled source register. A requirement not in that register cannot drive a finding, and the report says when it could not establish one.
  • Automated analysis may contain errors and may not detect every problem. Qualified review is required before a decision that matters.
  • It does not certify compliance and does not grant regulatory approval.

Questions

Does PharmaTrust certify that a batch is compliant?

No. It produces an analysis for a qualified person to act on. It does not certify compliance, does not act as a pharmacopoeia and does not grant regulatory approval. That boundary is enforced in the product, not just stated in marketing.

What happens when the certificate does not contain enough information?

The report says so. An absent value produces NOT_VERIFIED with the evidence that is missing named, and a document that cannot be read at all produces a hard stop rather than a page of empty fields. It never fills a gap with an assumption.

Can I see why a particular finding was reached?

Yes. Every finding names the requirement it rests on, the evidence observed on the certificate, the reasoning, the source and the version of that source, and the action that would close it.

Will two runs of the same certificate agree?

Judging is deterministic, so re-judging a stored reading gives the same result. Each run also records the engine, template, reader model, source versions and rule-registry fingerprint, so an older report resolves to exactly the software and sources that produced it.

Which pharmacopoeias and guidance does it work against?

The controlled source register holds the requirements the engine is allowed to cite, each with its version. A requirement that is not in the register cannot drive a finding.

Validate a certificate now

Five automated analyses every month, free. No card required.