Short answer
Predictive analytics uses historical and real-time data to estimate what will happen — failures, demand, cash, risk or churn. In a decision intelligence platform, predictive models feed prescriptive logic: each prediction is compared with thresholds and options, turned into a recommended action with its cost of waiting, routed to the owner and measured against the outcome so the model improves.
From descriptive to decision automation
| Level | Question | Output |
|---|---|---|
| Descriptive | What happened? | Reports and dashboards |
| Diagnostic | Why did it happen? | Drill-downs, root cause |
| Predictive | What will happen? | Forecasts, probabilities, remaining useful life |
| Prescriptive | What should we do? | Ranked options with trade-offs |
| Decision automation | Do it, safely | Approved actions written back; outcomes measured |
Predictive modeling techniques we use
| Technique | Typical decisions |
|---|---|
| Time-series forecasting | Production, demand, cash, staffing |
| Survival analysis & remaining useful life | Maintenance timing, asset replacement |
| Anomaly detection | Quality drift, fraud, sensor faults |
| Classification & propensity | Churn, conversion, claim risk |
| Optimization | Schedules, routes, inventory, pricing |
| Monte Carlo simulation | Cost of waiting, scenario ranges |
| Causal & uplift modeling | Which action actually changes the outcome |
How a prediction becomes a decision
01PredictProbability of failure in 30 days: 0.71
02CompareAbove the 0.6 threshold
03OptionsFix now, fix at slowdown, run
04CostMonte Carlo cost of waiting
05DecideOwner approves
06LearnOutcome labels the model
Keeping predictions honest
- CalibrateA 70% prediction should be right about 70% of the time.
- Show uncertaintyRanges, not single numbers, on every card.
- Monitor driftAlert when data or accuracy shifts.
- AbstainWhen context is missing, ask rather than guess.
Key takeaways
- Predictions matter only when they change decisions.
- Prescriptive logic turns forecasts into ranked actions.
- Outcomes retrain the models — the loop compounds.
Frequently asked questions
What is the difference between predictive and prescriptive analytics?
Predictive analytics estimates what will happen; prescriptive analytics recommends what to do about it. Decision intelligence adds execution and learning from outcomes.
What is predictive modeling?
Building statistical or machine-learning models that estimate future outcomes — such as failures, demand or risk — from historical and real-time data.
Is predictive maintenance the same as decision intelligence?
No. Predictive maintenance estimates failure; decision intelligence decides what to do given production, spares, cost and authority, then executes and measures.