Stop guessing which forecast to trust.
Upload a dated series and let up to nine statistical and machine-learning models compete on data they never saw — then keep the one that actually earned it.
Why Forecast Studio
Most volume forecasts come from one method picked years ago, tuned by feel, and judged on how well it fits the past. Forecast Studio runs the models side by side and ranks them on held-out data they never saw — so the one you use is the one that actually predicts.
What you get
- Prophet, SARIMAX, XGBoost, neural nets and more
- Holiday calendars for 250 countries
- Anomaly detection by model consensus
- Weekly / monthly roll-ups, Excel export
Built for
- WFM forecasters and planners
- Contact centre and BPO analysts
- Anyone forecasting a dated series
How it works
Three steps
- 1
Load your history
Upload a CSV or Excel file, or paste a column of dates and values. Duplicate dates are summed and unreadable rows are counted, not silently dropped.
- 2
Let the models compete
Up to nine models — Prophet, SARIMAX, ARIMA, XGBoost, Random Forest, neural networks, exponential smoothing and more — are grid-searched and scored on a held-out split.
- 3
Keep the winner
Pick the best-ranked model, roll the horizon up to weeks or months, and export to CSV or a multi-sheet workbook.
FAQ
Common questions
What data do I need?
One dated series: contact volume, tickets, orders — anything counted or measured over time. Upload a CSV or Excel file, or paste it in.
Does it handle holidays?
Yes. Holiday calendars for 250 countries are fed to the models that accept them as a regressor.
What happens to spikes and outages in my history?
Anomalies are flagged only when two independent methods agree. Replacing them is opt-in, because a real spike can be signal.
Do I need to install anything?
No. It runs in the browser.
See Forecast Studio on your data
Tell us about your team and we'll set up a walkthrough — usually within 1–2 business days.