Why Evidence Confidence Matters in Digital-Business Valuation
A valuation is only as strong as the evidence under it
Every valuation is built from facts. Some facts are confirmed, some are supplied by the owner, some are observed publicly, and some are inferred. Most valuations blur these together and report one confident number. The difference between a credible estimate and an invented one is whether the evidence underneath it is labeled.
Evidence states
ValuFai attaches an evidence state to every material input:
- Verified — independently confirmed.
- Owner-supplied — provided by the user, marked as such.
- Public source — observed from public information.
- Model-inferred — derived by the model, clearly labeled.
- Unavailable / unknown — honestly not known.
Why labels matter
Owner-supplied revenue and verified revenue are not the same evidence. A pricing page fetched publicly is not the same as a stated belief about a market. When the model cannot tell the difference, the user cannot either — and neither can a buyer, a lender, or an investor trying to rely on the result.
Confidence is bounded by evidence
ValuFai reports confidence on a 0–100 scale, and the scale has a hard rule: unknown information cannot increase confidence. Missing data widens the range and lowers confidence; it does not produce a tighter, more optimistic number. This inverts the normal temptation, which is to fill gaps with assumptions and present the result as precision.
Unknown is a legitimate state
Not knowing is not a failure of a valuation; it is a fact about the valuation. When information is unavailable, the honest outputs are a wider range and a lower confidence score — which is exactly what a serious reader needs in order to price the uncertainty themselves.
The practical effect
Two portfolios with identical revenue can carry very different confidence scores because of what is proven versus what is assumed. That is not noise; it is information. Confidence tells a buyer what to verify and a founder what to fix — which is more useful than a single digit pretending to know everything.
Last updated: 2026-08-08.