Pick any measure with dates: revenue, cases, headcount. The forecast fits seasonality when there is enough history, a trend when there is not, and always states the error it made on the most recent months it was not shown.
A forecast without its error is a guess with a chart.
The last months are held out, forecast, and compared. That error travels with the number into every answer and report.
A level shift — a closure, a relaunch, an acquisition — is detected, and the model trains on the business you run now.
One forecast per product, region or customer when you ask for it; the totals reconcile.
Four periods for a trend, two full seasons (24 months) for seasonality. With less, the forecast says so instead of pretending.
Yes — a watcher can watch a projection and fire when it drops below a target.
Yes. Ask for one forecast per product, region or customer and you get one for each, and the totals reconcile with the overall forecast.
A level shift such as a closure, a relaunch or an acquisition is detected, and the model trains on the business you run now rather than averaging across the break.