The multi-period control problem analyzed assumes the data are generated by the simple regression model with an unknown slope coefficient. There is a tradeoff between stabilization and experimentation to learn more about the unknown coefficient. When parameter uncertainty is large, experimentation becomes an important consideration.