The idea is to scan the array from left to right: when you are considering v_i you should count how many elements v_j (with j < i) are there such that v_i < v_j (in other words you should count how many elements, greater than the current element, you have already seen).

Using a BIT, you can keep how many times each possible value has appeared in the array. In this way you can easily retrieve how many numbers are there in some values range. So, if you are cosidering a generic element v_i, first you should increment the answer opportunely, then you should update the BIT, because you have seen v_i one more time than before.