Features
Evidence grades: no number without a rung
A report about a future nobody has seen is only as honest as its numbers. Each one says where it came from and how much to trust it.
The problem
Idea reports often contain confident numbers: a market size, a conversion rate, a price customers will pay. Few of them say where those numbers came from. A figure someone measured and a figure someone guessed look the same on the page, so a reader cannot tell which parts of the conclusion are solid and which are hope.
How it works
In a Startup Intelligence report, a number cannot appear without a rung. The product's own code refuses to store a figure that has no rung and no note on its source. The ladder has seven rungs, grouped into three tiers:
| Rung | Name | Tier | What it means |
|---|---|---|---|
| P1 | Measurement | measured | Someone measured it directly |
| P2 | Analog | modelled | A comparable venture's figure, with the difference stated |
| P3 | Reference class | modelled | A base rate from a group of similar cases |
| P4 | Structural derivation | modelled | Worked out from other figures by arithmetic |
| P5 | Constraint bound | modelled | A limit the figure cannot cross |
| P6 | Elicitation | asserted | An estimate someone gave when asked |
| P7 | Founder belief | asserted | Your own estimate, or a figure nobody has sourced yet |
The rung is printed beside the number, not in a footnote. A weakly evidenced figure is also given a wider range: the code keeps a belief from being stated more precisely than a sourced figure for the same thing. Where no figure can be given at all, the report marks the gap for you to fill in instead of inventing a value.
The report also grades its own evidence as a whole, and says plainly that this grade is about the inputs, not about your idea. A good idea can carry a weak evidence grade simply because nobody has sourced its numbers yet.
Who uses it
Anyone who has to decide how far to trust a report: the founder deciding what to test, an incubator comparing ideas, a consultant preparing a first look for a client.
Limits
Today no outside source is consulted during a run, so the base figures come from the AI model's own knowledge and sit on the lowest rungs. The ladder makes that visible rather than hiding it. Linking each assumption to published data you can open and check is planned, and until it exists the honest reading of most figures is "asserted".
Related features
Questions
Does the report give one overall score?
No. It gives no single score. Each figure carries its own rung, and the report as a whole states how much of its evidence is measured, modelled or asserted.
Where do the figures come from today?
From the AI model's own knowledge. No outside source is consulted during a run yet, so the report files those figures on the lowest rungs and says so. Linking each assumption to published data is planned.
