Abstract
Marketing mix model is a widely used statistical model to measure the effects of online and offline advertisements.However,there are three difficulties in measuring the effect of advertisements by marketing mix model ：Spurious causality,long-term effect,and evaluation of future advertisements.
First,there is a spurious causality between sales and advertisements because there are more sales and advertisements during selling seasons. To avoid this problem, we recommend that companies change the pattern of advertisements.
Next, it is difficult to measure the long-term effect of advertisements since marketing mix modeling measures only the short-term effect.It cannot simply extend the term due to overfitting. To measure the long-term effect,another analysis is required.
Finally, even if we measure the effect of past advertisements, we cannot evaluate future advertisements,so it is difficult to optimize the future budget.To optimize it,we can assume that
the past and future advertisements are the same.
