How it decides
Four steps, in order, none of which is a judgement call. The arithmetic is deterministic and lives in Python; the research is done by agents that are forbidden from stating a number they did not measure or cite.
Measure, and record what could not be measured
Adapters call public APIs and parse published datasets. Every result is stored with its source, its retrieval timestamp and its evidence tier. A failure is not dropped — it is written down as an unknown with a reason, because an unknown that lowers confidence is useful and an unknown that silently becomes an average is dangerous.
Gate before scoring
Deal-killers are categorical, and a weighted average will happily average them away. Eight knockout checks run first. A gate whose input is missing does not quietly pass — it is reported as undecidable, which is a different and more honest state than satisfied.
Score, then discount for age and for tier
Confidence is the tier weight of every measurement discounted by how far past its source's refresh window it is, weighted by how much the factor matters. Low confidence widens the ± band. A domain with nothing measured drops out of the score entirely, and the page says so rather than letting the mean quietly cover a smaller model.
Ask what would change the answer
Each factor is re-scored at its best and worst plausible resolution, one at a time, through the same scorer. Anything that moves the result across a verdict threshold on its own is a factor that has not yet been decided by evidence — and it goes to the top of the verification queue.
Every number says how well it is known
A score of 82 built on satellite measurements is a different asset from a score of 82 built on news articles. Most tools present them identically.
- Tier A
- A live scientific API returned the number. Reproducible today and next year.
- Tier B
- An official published dataset or government file was parsed.
- Tier C
- Researched on the web and synthesized, with links. Real evidence, weaker footing.
- Tier D
- Modeled, estimated, or supplied as an assumption. Labeled loudly as such.
- Unknown
- Could not be measured. Recorded with a reason — never quietly averaged in.
Most of the picture is still missing, and every page says so. Across these analyses, most factors are unmeasured. That is why every score carries a ± band, why the decision on all seven sites is currently not proven, and why two sites whose bands overlap are reported as not separable rather than ranked.
What this cannot do
Stated plainly, because a tool that oversells itself gets trusted once.
It is not due diligence
It tells you which twenty sites out of two thousand deserve a real investigation, and what to ask when you get there. The verification queue is the handoff, not the answer.
It cannot price a deal
Profitability needs the tariff, the land price and the incentive package. Where those are unmeasured the system refuses the profitability question outright instead of implying it from the score.
It cannot predict a utility or a county board
It reports published queue data and historical timelines. Whether a utility will agree to serve 500 MW is a negotiation, not a measurement.
Land ownership is largely invisible
Cadastral data is openly published in about a dozen countries. Elsewhere the system resolves to administrative boundaries and says so rather than inventing parcels.
Coverage is uneven by design
Climate and terrain are strong everywhere. Grid capacity, land price and local politics are weak nearly everywhere. The tiers make that visible per factor rather than averaging it into a single number.
Drawing an area does not analyze it
The map computes geometry in your browser and hands you a command. Running it calls live data sources, which costs somebody rate limit — so it happens on your machine or in a tracked request, never silently from this page.