Trust and security

How Signalment keeps study records.

What an auditor, a sponsor’s QA team or your IT reviewer will ask, answered with the product itself where it can be. The screens are from the example study.

Example data
The audit trail of the example study: 2,917 entries with the chain verified, each with its number, time, person, action and the reason given.

Audit history

Every action is a numbered entry: who, when, what and why, each bound to the one before it by its hash. The chain is checked from the first entry to the last.

Example data
The workspace team: each person’s role, the sites their access covers, and whether they see allocations.

Role-based access

Six roles, access limited by site, and blinding by person. Monitors, assessors and viewers, such as a sponsor’s, never see owner identities.

Example data
A physical examination at version 3: the query on the body weight that prompted the correction, closed when the data changed, and the signatures, each bound to the record’s hash.

Versions and signatures

A correction makes a new version with its reason. A signature asks for the password again and binds to the version signed.

Example data
Export and audit for the example study: the study archive, the JSON bundle with audit trail, subjects CSV, data dictionary, SAS program and the offline check.

Exports you can check

One CSV per form, a data dictionary, a SAS program and a signed JSON bundle that anyone can verify offline.

Who does what

The software keeps the records. The study team runs the study.

Signalment is a data system. The conduct of the study, its approvals and any filing stay with the sponsor and the investigators.

What Signalment does

  • Keeps every entry, change and signature on an append-only audit trail
  • Limits access by role, by site and by blinding
  • Versions forms, protocol, consent and every record
  • Produces exports that anyone can verify offline

What your study team does

  • The protocol, ethics review and any regulatory filing
  • The content of consent and the veterinarian–client–patient relationship
  • Adverse-event reporting on the protocol’s timelines
  • Procedures, training and validation for your intended use

Security

Access and accounts.

  • Sign-in by password or passkey; passwords are stored only as salted, peppered PBKDF2 hashes
  • Idle sign-out and password-age policies set per workspace
  • Every signature asks for the password again
  • Six roles, with access limited by site and study, and blinding by person
  • A strict content security policy and no third-party scripts in the app, apart from cookieless Cloudflare Web Analytics
  • Report a security concern to security@signalment.app

Validation and releases

A validation package, and releases under change control.

  • Validation plan, user requirements, risk assessment, test protocol and report, traceability matrix and summary report, with SOPs for access, incidents and training
  • Every release runs the automated test suite and a browser walk-through, then goes to a separate staging system before production
  • A schema change adds columns or tables; a release never rewrites study records
  • While a study is live, its validation rules, hash rules, export formats, audit trail, randomization algorithm and blinding rules change only with the customer’s written agreement
  • Customers with a live study hear by email before a release changes a screen their staff or owners use

Request the validation package

Data handling

Where the data are, and who sees them.

  • Hosted on Cloudflare and stored in United States data centers, encrypted in transit and at rest
  • Owner identities held apart from study data; records and exports use subject codes
  • Sub-processors: Cloudflare (hosting and storage), Resend (email), Stripe (billing), Twilio (text messages, only for a study that offers them and an owner who agreed) and, only when a workspace turns on the assistant, Anthropic
  • You own the study data; exports are open formats you keep
  • The data processing addendum and the privacy notice give the details

Backups

If something goes wrong.

  • The database keeps thirty days of history and can be rewound to a point in time
  • A full backup can be restored into a fresh database; the restore is rehearsed and checked row by row against the backup
  • After a restore, every table’s rows and every workspace’s audit head are compared with the backup
  • A study’s own exports and archive are a copy you hold, checkable without us

Regulations

Built to 21 CFR Part 11 and VICH GL9.

Compliance belongs to a study, not to software alone. These are the parts the software provides.

  • Audit trail with who, what, when and why: 21 CFR 11.10(e)
  • Signatures that re-check the password and bind to what was signed: 11.50, 11.70, 11.200
  • Access by role and by site: 11.10(d)
  • Idle sign-out and password-age policies: 11.300

Ask the questions your QA team will ask.

Book a demo and bring them. We will show the records, the exports and the validation package.