Demand Forecasting - Error in predictions

I’ve been trying to do daily demand forecasting using the H2O AutoML package for using 4 years of daily sales data as the training set. However, the daily estimates always seems to find a range around which they oscillate. The following are the features I use:

1. Date
2. Demand of same weekday in the previous week
3. Demand of same weekday in the second previous week
4. Median demand of the same weekday in the previous month
5. Median demand of the same weekday in the previous quarter
6. Median demand of previous month
7. Median demand of previous quarter
8. 95% demand quantile of previous month
9. 95% demand quantile of previous quarter
10. 5% demand quantile of previous month
11. 5% demand quantile of previous quarter

Can anyone please suggest where I’m going wrong?