to make sure the model give you accurate estimates or not. If you put overly rosy assumptions, you will get great results and fool yourself willingly.