This is definitely one of the best tutorial videos I ever watched. I would request to publish more such videos. One thought Hans, the "Year" variable you were putting in the Regression, I might avoid it. "Year" is not a continuous variable and might give a wrong interpretation.
I think the only missing step would be to apply clustered errors to account for serial correlation between observations from before and after treatment.
Great video - clear explanation! Why would adding the year cause this to be statistically significant? Are there approaches I can take to ensure all appropriate variables are in the model formula?