1 · Independent model passes
The subject is scored by more than one frontier model, each working from the same evidence packet. No single model's opinion becomes the score.
Methodology
A DICE score is not one model's guess.It is a reconciled consensus.
- Several independent frontier models score the same evidence packet.
- Disagreement between models is treated as uncertainty, not averaged away.
- Every report states the Data Confidence behind its own numbers.
- Re-runs are deterministic — a score moves when your product moves.
2 · Consensus reconciliation
Where the models agree, confidence rises. Where they disagree, the spread is treated as uncertainty and the score is pulled toward the conservative read rather than the flattering one.
3 · Strategic Intelligence layer
Raw model output is checked against market structure — category dynamics, comparable pricing, distribution reality, and the questions a real investor asks in the first meeting.
4 · Data Confidence rating
Every report states how much evidence it was built from. Public URL, live site content, category context, and any private founder context each move the rating. We grade the evidence instead of pretending it is complete.
5 · Deterministic re-runs
Scoring runs at temperature zero against a stable seed, so the same product scored twice returns the same numbers. A change in your score means a change in your product — not a change in the model's mood.
Standing principles
One scale, always
Every DICE score — subject and competitor alike — is a whole number from 0 to 100. Risk is stated as 0-100 where lower is safer. No mixed scales, ever.
Readiness, never outcome
DICE scores how prepared a product is to be evaluated. It does not predict whether a company will succeed, and no DICE score should ever be read as a prediction of return.
Evidence over assertion
Traction numbers are never invented. If something is estimated, the report says it is estimated. If evidence is thin, Data Confidence says so on the face of the report.
Private data stays private
Context a founder supplies is used to sharpen that founder's own report. It is not published, not sold, and not shown to other users.
See it applied to your product
Run a Scoreboard and read the Data Confidence rating at the top of your report.
Run my ScoreboardDICE is decision support, not investment advice. See our disclaimer.