Testing against real cases

A test checks that the code does what it was told to do. An evaluation checks that the product delivers on what we promise the client — that's a different question, and the second is never inferred from the first.

Before going into production, a project is measured against a 66-point reference framework across seven dimensions: legal and GDPR, search visibility, security, accessibility, observability, proven quality, and cost. A blocking criterion is never waived, however minor.

We measure what most don't, and we don't grant ourselves exceptions.