ClinOpsWord audits Word protocols and CSRs for numeric discrepancies, Word/XML corruption and ICH E6 structural gaps — audited entirely in your browser, so nothing ever leaves your machine.
Zero document bytes transmitted — verifiable in your browser's Network tab
Deterministic engine: identical input always yields identical findings
Ineligible documents are rejected, not scored — no invented findings
Don't take our word for it
This is the production audit engine, running in this page. Drop a .docx and it is unzipped and parsed in your tab. If you would rather not use a live document yet, run the bundled sample protocol — its findings are computed from the sample text by the same engine, not hardcoded.
Open DevTools → Network, then run the analysis. No request carries your document.
In-browser sandbox
0 bytes uploadedDrop a protocol or CSR (.docx) here — it stays on this machine.
What this prevents
A discrepancy that reaches a regulator returns as a request for information: a response cycle, a document re-issue, a re-review. The three failure modes below are the ones a machine can catch reliably and a tired reviewer cannot.
Data Integrity
A subject count in the narrative that does not match the disposition table, an alpha level that drifts from the SAP, an adverse-event total that no longer reconciles after a data cut. Each one is a plausible information request, and each request costs review cycles.
Word / XML
Orphaned numbering references that black out bullet numbers, a table of contents frozen at an earlier revision, non-standard fonts inherited from a partner's template. These survive PDF rendering and reach the submission packet.
Scale
Cross-checking dozens of numeric fields across several hundred pages by hand, under a lock deadline, across timezones. Reviewers are good at judgement and bad at exhaustive arithmetic — the failure mode is systematic, not careless.
Illustrative, not a case study: ClinOpsWord is new and we publish no customer outcomes or savings figures we cannot evidence.
Product walkthrough
The workspace shows every processing stage as it completes, including a live counter of network requests made during analysis. The steps below are the real flow — you can walk it yourself in the sandbox above.
Optionally add a trial database export (ADSL/ADAE CSV) for cross-referencing. Files are read from disk, not uploaded.
The tab unzips the .docx, parses WordprocessingML, extracts every numeric token, and runs the enabled checks. You can watch each stage complete.
Each finding carries the verbatim excerpt, its section location, a risk tier, and the regulatory reference it maps to. Drill down to the exact block.
Apply in-memory style/XML repairs, then export the audit report as PDF, CSV or JSON for your submission packet.
Three modules
Declared subject counts, alpha levels and safety totals that disagree between narrative sections, or with a trial database export you supply.
Example finding: “Section 4.2 declares N=300; the disposition table totals 298” — with the excerpt and section shown.
Numeric inconsistency is the most common avoidable trigger for a regulatory information request.
If you supply no database export, cross-referencing is reported as skipped — never guessed.
Orphaned numbering IDs, missing or stale tables of contents, and non-standard fonts inside the Word package.
Example finding: “12 orphaned numId references in word/numbering.xml” — counted from the actual file.
Structural defects render inconsistently at the agency's end and undermine document control.
Repairs are applied to an in-memory copy; your original file on disk is untouched until you export.
Missing or unlabelled sections required by ICH E6 (R3) protocol structure, and absent safety-reporting language.
Example finding: “No section addressing adverse event reporting timelines was located.”
Structural gaps are caught by reviewers before science is even discussed.
Presence checks are text-based; they flag what to inspect, they do not certify compliance.
A .docx is a ZIP archive of XML parts. When you choose a file, the browser hands the page its bytes; ClinOpsWord unzips that archive in memory, reads word/document.xml along with the styles and numbering parts, tokenises the narrative, and runs the checks — all inside the page. There is no upload step because there is no server-side analysis component to upload to.
You can confirm this rather than trust it: open your browser's developer tools, switch to the Network tab, and run an analysis. The workspace also shows a live request counter during processing, read from the browser's own performance timeline.
Documents are read from disk into memory and discarded when the tab closes.
Unzip, WordprocessingML parsing and tokenisation all execute in the page.
Runs as a desktop-class PWA window with no IT deployment package.
Reports export locally as PDF, CSV or JSON; we never hold a copy.
Early access — no testimonials yet
ClinOpsWord is new. There are no customer logos, testimonials or user counts on this page because there are none we can evidence — and this audience checks. Here is what we can put behind the product instead.
It comes out of working inside clinical document workflows — the QC pass where a reviewer reconciles narrative numbers against tables and database exports by hand, days before a lock. That task is mechanical, exhausting and exactly the kind a deterministic tool should own.
Because analysis is local, evaluation costs you nothing and exposes nothing. Run your own protocol through the sandbox above and judge the findings against what you already know is in the document.
Every finding states the check that produced it, the verbatim excerpt, the section, a risk tier and the regulatory reference. The Health Index is a weighted pass rate over the checks that actually ran — skipped checks never count as passes.
Plans
$99/month
$499/month
Contact sales
You won't be charged today. Your 7-day free trial starts after your payment method is verified, and you will be charged the selected plan price after the trial ends unless you cancel before it expires.
FAQ
Early access
Founding users shape the check library and the validation pack. Run your own protocol first; if the findings are useful, join the early access cohort and we will prioritise the checks your programme needs.