In this section
Part of Celiq's semantic layer platform. Connect your warehouse, model your data once, query it everywhere.
Forecasting
Predict any metric into the future using Meta Prophet — a production-grade time series model that accounts for weekly and yearly seasonality.
Ask Orion to forecast
Orion detects forecast intent automatically. Try questions like:
- "What will revenue be next month?"
- "Forecast total orders for the next 30 days"
- "Where will MRR be at end of quarter?"
- "Predict churn rate for the next 2 weeks"
Orion responds with:
- A forecast chart (historical + dashed forecast + confidence band)
- Plain-English summary of the prediction
- Trend direction and percentage
- Confidence level (high / medium / low)
- "Add forecast tile to mosaic" button
Forecast tile in Mosaic
Add a self-refreshing forecast tile to any mosaic:
- Open a Mosaic and enter edit mode
- Click Add tile
- Select 📈 Forecast tile
- Choose your reveal, enter the metric name and date field
- Set the forecast horizon (7 / 14 / 30 / 60 / 90 days)
- Click Add tile
The tile refreshes with every mosaic refresh, pulling the latest 90 days of warehouse data and re-running the model.
How the model works
| Setting | Value |
|---|---|
| Model | Meta Prophet |
| History required | 14 days minimum (90 days recommended) |
| Weekly seasonality | Always on |
| Yearly seasonality | On when 180+ days of history available |
| Confidence interval | 80% |
| Trend detection | up if forecast mean > recent mean by 2%+ |
| Level | Coefficient of variation |
|---|---|
| High | < 0.3 |
| Medium | 0.3 – 0.6 |
| Low | > 0.6 |
Limitations
- Needs at least 14 data points (days of history)
- Forecasts assume patterns similar to historical data continue
- Works best with daily-grain data; weekly and monthly also supported
- Not suitable for metrics with highly irregular, one-off events