Predictive Planning (Auto Predict)
Time-series ML that produces a statistical baseline forecast from your historical data — the starting point analysts refine, not a black box that replaces them.
Predictive PlanningWhat it does
Auto Predict runs against any time-series in your EPM cube — revenue, expense, headcount, demand — and produces a statistical forecast using classical time-series methods (ARIMA, exponential smoothing, etc.). The model is selected automatically per series based on best fit.
The forecast lands as a baseline in your planning model. Analysts adjust based on judgment (a new product launch, a known customer loss) — they don't start from zero.
Why it matters
Most baseline forecasts are still last-year-times-growth. ML-based time-series gets the shape right (seasonality, trend changes) in a way human-built baselines rarely do — and it does it for thousands of series at once.
Where the value shows up
- Less time building the baseline; more time on judgment overlays.
- Better rolling forecasts — the cost of refreshing drops to near zero.
- Variance explanations get easier because you have a model-explainable baseline to compare actuals against.
What Auto Predict does not do
- It doesn't know about your plans (a launch, a divestiture, a price change). Overlay those manually.
- It struggles with short or volatile series. Less than ~18 data points and it falls back to a simple mean.
- It's not a replacement for driver-based planning — it's a complement. Use it for the components where the past is the best predictor of the future.
Action checklist
Tap each step as you complete it.