I recently tried this in a Google tutorial, it was very nice, but I was surprised to learn in the same tutorial that Google Cloud AutoML provides better results than BigQuery ML.
You can do your own forecasting in SQL with Python or R. Depending which database you use, there will probably be some kind of integration. Postgres has PL/R and PL/Python.
Check out the statistical functions baked into Snowflake SQL. We've got a PL/SQL script that forecasts sales volume with seasonality, using a modified ARIMA approach. We're also turning this approach loose on our own data pipeline to trigger alerts on abnormal row counts.
