How scoring works
Consistent criteria, visible uncertainty and a clear separation between AI hypotheses and public-source research.
Five weighted dimensions
Customer problem and demand, differentiation and switching, revenue logic, customer acquisition, and execution feasibility are assessed using the same rubric for free and paid users. The server calculates the weighted total. The dimension cards show the applicable maximum points and reasoning.
Evidence, not enthusiasm
Each dimension receives a rating from 0 to 4. Eligibility is checked first. Clearly unlawful purposes or impossible core premises receive no score. Missing information or uncertain eligibility puts scoring on hold and prompts clarification. Hypothesis-only ratings are capped at 2 out of 4 per dimension, so an assessment based entirely on hypotheses cannot exceed 50/100. Areas assessed measures coverage, not validation. Specific explanations and described measurements receive different evidence labels. A claimed measurement is still a user claim, not an independently audited fact.
Interpretation
A score is not a success probability, investment recommendation or validated market ranking. A total score is assigned only after eligibility and all five dimensions can be assessed. Do not compare scores across different countries, inputs or rubric versions as a league table.
Market and language
Connection country is only a suggestion; choose the market you intend to serve. Language is separate. For the US, select a state or DC, or specify multiple operating states, and optionally a city or county. Detailed research seeks relevant federal, state and local official sources and flags coverage gaps. Free scoring does not look up current laws. Customer locations and business activities may affect which rules apply; a selected state alone cannot establish compliance. Unsupported markets receive general hypotheses only.
Public-source research
New detailed reports for US/Korean markets request Google Search-grounded research. The complete search response is displayed with associated search suggestions and source attribution. The retrieval date is shown; a source's publication or effective date is not assumed to equal that date. Sources may still be incomplete, outdated or inapplicable. Follow the links and confirm their relevance. A search does not establish compliance.
Keep the layers separate
The detailed planning sections and follow-up analyses are AI hypotheses, separate from the attributed public-source section. Follow-up questions and execution plans do not run a fresh web search. Research does not silently rewrite the original score. Public-source responses are shown only in the requester's current session and are not archived with the report or publicly shared. Saved detailed reports offer one complimentary source refresh per account per UTC day; your saved AI analysis remains available without another credit.
Limits and failure handling
Free usage has rate limits to control cost and abuse. Successful results are saved and identical requests reuse them. If a detailed report's required search or generation fails, the analysis credit or free report trial is restored. Provider outages and quotas may temporarily prevent analysis.