Predictions and outcomes

What it is

A prediction is a forecast, classification, or decision recommendation produced from ontology and workflow state — always with evidence and always awaiting your review. It is advice with proof, not an answer from nowhere.

What it holds

• Kinds of forecasts: probabilities ("70% renewal risk"), categories ("at-risk / healthy"), numbers ("expected volume next quarter"), and scenarios ("if hiring freezes, then…"). Different questions need different kinds — scope the decision first.
• Evidence per figure: drivers that make business sense, inputs you recognize, a time range you chose. Skepticism at preview time is the job, not pessimism.
• Outcomes: what actually happened — occurred, did not occur, still unknown, ambiguous, or unresolvable. Outcomes are recorded, not assumed, and accuracy is measured from them.
• Review state: approved, corrected, or rejected — with reasons that future runs can learn from.

How it relates

• Predictions read ontology runtime and workflow outputs; weak state in, weak forecast out.
• Evidence chains connect each figure to source records through the run that produced it.
• Outcomes close the loop: the operational result (renewed, churned, resolved) feeds the next round — this is what turns one forecast into improving judgment.

What this means for you

• Scope target, time range, and audience before anything runs.
• Explain forecasts from evidence, never from scores.
• Record outcomes; a forecast without a measured result taught nothing.

Image: forecast anatomy — figure, drivers, evidence links, review actions, outcome state.

Next

• Produce one: Get your first forecast.
• Kinds, modes, budgets: Forecasts.