TNET methodology
TNET does not use a rule such as “rain means thirty percent less speed.” The system evaluates measured network context, the structure of the atmosphere, time and plausible orbital geometry before producing a practical range and connection score.
The product
A TNET forecast is an authored estimate for a selected place and time. It is designed to answer whether the available evidence points to a stronger, typical or more challenging connection window—not to promise an exact speed for every terminal.
The central estimate is accompanied by a range because live satellite allocation, beam load, terminal hardware, installation and local obstructions are not fully visible to an independent public system. Those unknowns are part of real-world satellite internet, not replaced with invented precision.
Evidence profile
No individual parameter decides the forecast. Each one is interpreted within a wider evidence profile and the selected time window.
Empirical reference
Quality-controlled Starlink measurements establish the observed performance context for a location and time. Download, upload and latency are treated separately, and sparse areas use broader evidence rather than unstable local claims.
Atmospheric structure
TNET evaluates precipitation, cloud and moisture structure, temperature layers, freezing conditions, wind and convective context at the selected time and across the surrounding area.
Candidate geometry
Public orbital records are propagated for the location and interval to describe plausible satellite elevations, ranges and geometry changes. Serving satellites and beam assignments are not claimed.
Propagation context
Atmospheric and candidate-path evidence is translated into radio-sensitivity context. A physical signal indicator is never presented as a simplistic fixed percentage of lost throughput.
Confidence reflects evidence coverage, freshness and internal consistency for the selected result. It is separate from connection quality: a modest outlook may have strong evidence, while a favorable outlook can have limited support.
TNET publishes the meaning and provenance of evidence, but not private feature engineering, model weights, thresholds, calibration logic, data-selection rules or internal scoring methods. These form part of the TNET forecasting system and commercial know-how.
Read the evidence, then choose the window.
Return to the forecast and inspect the scientific profile for a place and date.
