A unit's credibility depends on its formula being public and recomputable. This is the starting proposal; the principle is non-negotiable: nothing hidden.
The RosettaQ of a (system, problem class) pair is its measured performance, at equal budget, on a public scale where 100 RQ = the best classical solver on that class.
RQ(system, class) = 100 × (solve_quality / best_classical_quality) same instance · same budget · fixed seed · frozen versions
Because it folds the buyer's three questions into one number: does it work? (is it past 100?), by how much? (the distance to 100), and when? (at what size the curve crosses 100). And because, selling no hardware, we hold the neutrality no vendor can. The live calculator is the public proof that there is no black box.
To be tightened rigorously: (1) how a system’s RQ aggregates across instances (mean / median / worst case), (2) whether price enters the unit or sits as a separate axis, (3) how a vendor metric converts to an estimated RQ without over-promising.