Advice on dealing with Cold Start forecasting [D]

Hey everyone. I am dealing with a problem statement at work where the goal is to predict the sales of a new innovation to be launched. Note that the data that I actually have is just product attributes, the month level invoice data for each region, weather data and some macroeconomic data. Currently there is a methodology in place where we are using a xgboost regressor trained on the historical data for this prediction, but no one in the team is very confident in the methodology. Given this, does anyone have advice on how we could go about this problem? Please do let me know if you've dealt wi