Rate one GenAI system across the six faces of the Value Cube and the five cross-cutting dimensions. You get the average, the guardrail caps that stop a strong face hiding a broken one, the maturity band, and the face to fix first.
The Cube Score is a structured diagnostic, not a scientific grade, and never a stand-in for validated ROI. It travels with an evidence-quality label for exactly that reason.
Score each face 0–5 on the ladder below the calculator. Be honest rather than generous. The shape matters more than the number.
Read across the faces as balancing checks. Quality & Trust is a guardrail: below 2 it caps the overall score.
Two questions that can cap the score, and the honest label for what this assessment rests on.
The share link carries your scores in the address itself, so a colleague opens the page exactly as you scored it. Nothing you enter leaves your browser: the scoring runs in the page itself, and there is no server and no account. Your scores are never sent anywhere or stored. The site counts anonymous, cookieless page visits so I know which tools get used — see privacy.
Cube Score = the average of the six face scores, interpreted with the five cross-cutting dimension checks and the guardrail caps. No weightings, no composite index, no false precision. A simple average a team can compute in a workshop, made honest by the caps rather than by mathematical elaboration.
If governance scores below 2, the overall score cannot exceed 3. If the Quality & Trust dimension scores below 2, it cannot exceed 3. If a high-risk use case lacks risk classification, it cannot exceed 2. If cost visibility is missing, it cannot exceed 2. A fatal weakness sets a ceiling that no amount of strength elsewhere can lift.
| Score | What it means on a face |
|---|---|
| 0 | Not implemented, or not even visible |
| 1 | Basic awareness; ad hoc practice |
| 2 | Partially implemented, limited consistency |
| 3 | Managed and repeatable |
| 4 | Optimized and measured |
| 5 | Intelligent, governed, continuously improved |
| Overall | Band |
|---|---|
| 0.0–0.9 | Blind Spend: cost-to-value largely unmanaged |
| 1.0–1.9 | Visible: basic visibility and awareness exist |
| 2.0–2.9 | Controlled: some repeatable controls and metrics |
| 3.0–3.9 | Managed: cost, quality, governance managed across major use cases |
| 4.0–4.5 | Optimized: patterns actively measured and improved |
| 4.6–5.0 | Intelligent ValueOps: dynamic, governed, continuously improved |
The output is a profile, not a grade: six face scores, five dimension checks, the capped overall, its band, and the evidence label. Its real value is that it points: the lowest face, especially one that triggers a cap, is where improvement starts. Chapters 16 and 17 of The Value Cube work through the score and the maturity climb in full.