Metrics for Measuring QA productivity

I work for a small software firm that makes mathematical models.
One of the company goals was to reduce the cost of repetitive QA
during this year
I have been given this task to measure the above mentioned goal.
Since we dont have have documented definitions of 'cost' and
'repetitive QA' I am finding this task difficult.
Everybody in the group feels that this goal has been achieved; however
its a soft feel. I cant produce any hard numbers to quantify the
feeling.
Can anybody suggest some kind of a scientific approach for me to
develop a metric that can be presented to the top management?
Many thanks in advance.
Sincerely,

Devang