There are three aspects that may lead to an invalid forecasting: data, model and context. I think that data has an important role here, but it is also relevant to work with an accurate model. What are your thoughts about this? What factor mainly affects forecasted results?