Skip to content

Trust

Trust

An institution cannot use a number it cannot interrogate. These pages document how Phoeniks produces an estimate, which sources feed it, how it is validated and how the platform is secured. Nothing here requires a leap of faith in a vendor claim.

What this document commits to

Documents
Four: methodology, sources, validation, security
Scope
Every published figure
Review basis
Reproducibility ledger
Audience
Credit, actuarial and regulatory reviewers

Key terms on this page

01

Four documents, one chain#

MethodologyHow is an estimate produced, and from which assumptions?
SourcesWhere does every input come from, and under which licence?
ValidationHow do we know the model is any good, and where does it fail?
SecurityHow is the platform and its client data protected?
02

How to read these documents#

  • Start with methodology: it sets the vocabulary — model ensemble, evidence level, confidence band.
  • Move to sources to check the provenance and licence of every input before relying on a figure.
  • Read validation for measured performance and, more importantly, the regimes where the model fails.
  • Finish with security if your review runs through a risk committee or an information-security team.
03

What the evidence chain does not claim#

None of these documents promises universal accuracy. A Phoeniks estimate carries its evidence level — measured, modelled, inferred or untested — precisely so the reader knows what has been verified and what has not. Disagreement between models is published rather than averaged away.

Related pages

See also

Request access

Institutional access. No self-serve trial.

Request access