A downside to using the maximal margin hyperplane is that it is extremely sensitive to outliers and may have a tendency to overfit the training data. In this case, it may be beneficial to pick a hyperplane that misclassifies some of the training observations using the support vector classifier (AKA soft margin classifier). Instead of making sure that all the training observations fall on the correct sides of the margin and hyperplane, we can allow for some to be on the incorrect side of the margin, and even the hyperplane. This allows for more accurate classifications for future observations. This increases the bias of our method since it is not as flexible as we allow misclassifications, but it has a lower variance when using new observations.