Build a forecast

Time to complete: ~10 min (sample) or ~20 min (gold, after ingest)

Time series forecasting from gold or the AirPassengers sample.

Try the sample first

Quick start: sample walks through AirPassengers end-to-end.

Before you start

Path	You need
Your data	Gold with Forecast enabled, 30+ rows, valid time + value columns
Sample	Signed-in account only

Open the wizard

1. Sidebar → Data Intelligence → Forecast (/forecast).
2. Start a new Build (opens the training wizard).

Train from gold

1. Data source → select gold artifact (Forecast-enabled, status ready).
2. Forecast interval — hourly through quarterly; match how your data was collected.
3. Analysis step — review seasonality, trend, algorithm hint.
4. Name the forecast → Train.
5. Open from list when training completes.

Train from sample

1. Data source → Sample dataset → AirPassengers.
2. Set Interval → Monthly → Train.

After training

1. Forecast detail page → Deploy → Generate API key.
2. Monitor on asset page and Usage.

Next steps

• Deployment
• API integration
• User journeys

Troubleshooting

Analysis: invalid data
Republish gold with a clear datetime column and numeric values.

No gold in the list?

Confirm Forecast was enabled at publish and the artifact is ready (not draft or failed).

Wrong interval?

Monthly data trained as hourly produces poor forecasts. Match interval to how your series was collected.