14 days. Nexus' micro-mesh forecasting system can predict beyond 14 days; however, we typically restrict our forecasts to 14 days or less because forecast confidence beyond that becomes very low, and we don't want our customers making bad decisions. Forecasts beyond 14 days enter a new category of modeling such as sub-seasonal, seasonal, and climate forecasting, all of which is not our core focus.

Yes, there is a real-time customer-facing validation module that overlays forecasts against publicly available observations or the customer's own sensors. The Nexus Platform stores the forecasts, not just the observations, so you can replay what the forecast said at the moment a decision was made, not just what the weather ended up doing. You have a record of what you decided on and why.

Everything is probabilistic. Nexus doesn't just tell you it's more accurate; it tells you when the models are not high probability, so you know to wait before committing to a decision.

API-first. Everything visible in the UI is available through API endpoints.

Raw parameter data, and risk-based output: the joint probabilistic exceedance of multiple thresholds simultaneously, delivered as a single score.

Potentially! We have API integrations with several partners.

Each customer's data is fully siloed in its own ML module and is never shared across accounts. The Nexus-produced base machine learning training set is common to all accounts as a starting point only; anything the customer contributes, and the output forecast from that training data set stays theirs.