Organised by problem class, not by logo. We don't belong to one industry — we belong to a class of business problems where work repeats, decisions are rule-shaped, and information moves between systems that don't talk.
And rather than list them flat, here they are plotted against the evidence ladder — including the band that is still empty.
The systems work. What is missing is a signed pre-deployment record to measure them against — and once a process has changed, its baseline is gone for good. That is not a gap in the engineering; it is a gap in the sequencing, and it is exactly why the Baseline Week now opens every engagement.
Marichi is the exception worth reading: its engine was validated statement by statement against independent commercial reference software, with two values left flagged rather than fitted. That is a correctness figure, not a client outcome, and the page says so.
The consistency is itself an argument. It says these are engagements run through a method, not projects that happened to go well and got written up afterwards.
Escalation rate, correction rate, incorrect-action rate. Nobody else prints this.