It has been tested in user research and shown to work with a representative sample of users, including those with disabilities. Components and patterns which are not proven usable can be published as experimental. But they must be clearly based on relevant user research from other organisations and best practice, and meet the other criteria.