How to Read Predictions

Purpose

Understand what each prediction means so you can prioritize attention and act appropriately.

What a prediction typically includes

• Title: short name of the potential upcoming issue.
• Probability scores:
• AI probability: an AI-based confidence that this issue is likely to occur next.
• Statistical probability: a data-based value derived from historical frequency (no AI interpretation).
• Actions:
• Help: opens matching guidance suggestions
• Analysis: opens historical error analysis
• Ignore: available only if your role allows ignoring messages

Prediction cards

Many deployments show prediction cards with a visual hierarchy:

• the highest-probability item is highlighted (often a "primary" card)
• additional predictions can be selected from a chart and shown as secondary cards

Card colors and exact behavior can vary by deployment.

Example (names and layout may vary by deployment):

Error prediction chart

A bar chart may appear below the cards to compare potential upcoming issues.

Depending on your deployment, the chart can show:

• statistical probability vs. AI recommendation (often using different bar colors)
• the currently selected prediction (highlighted)
• exact probability values on hover or tap

Best practices

• Treat predictions as guidance, not an automatic decision.
• Use the Analysis view to validate whether the prediction matches historical behavior.
• Use Help when you want to look for existing guidance before drafting a new procedure.

Troubleshooting

• A prediction does not make sense -> Filters/machine selection mismatch or data patterns changed -> Verify machine and time range, then re-check -> Escalate with the prediction title and selected machine
• Probabilities look inconsistent -> Different probability types are shown (AI vs statistical) -> Compare both values and use Analysis for context -> Escalate if values look obviously wrong for known machine behavior