What to do if ARIMA and ETS predictions are bad?

[JohnLight]

I have a bunch of data for multiple states, and most of them look good, but there's some that look like the picture. How do I use forecasting tools for this data if ETS and ARIMA predictions look this bad? The configuration for all the prediction tools are the same between states, and all the rest look good, but these few aren't working.

[aatalai]

Hey @JohnLight have you tried using ARIMA with covariates?

Alternatively you can try changing the hyper parameters to get some improvements here is some tips for ARIMA  

and ETS

Hope it helps