Now it’s time to measure the problem against the data to set the desired performance. Basically, how much do you need to change the situation, relative to the current benchmark to reach the desired level of improvement?
Sam, a director of Admissions at a mid-size university is in charge of retaining students. We have determined that he has:
- A major problem that data analytics can help resolve
- Enough data records to accomplish this task
His goal is to restore retention back to its previous level without increasing institutional costs. A retention rate increase of 2% to 2.5% would meet the goal.
In data mining terms, a moderate improvement is generally in the range of 10% to 100%. Sam’s need is in this range, at 25%. He needs a moderate performance increase.
Performance is the result of data mining effort, not a precursor to it. It helps us aim in the right direction, but we don’t know if the arrow will hit the bullseye until after we completed the analysis.
Basically, none of us have a crystal ball that shows the outcome in advance. That is why we are doing analysis, to look at the facts and uncover possible causes of the problem. The process will not necessarily confirm our assumptions.
That said, even without a guarantee, you can use our experience as a guide. IMA can assist with this if you are unsure how to measure or determine a moderate improvement. Incremental-to-moderate improvements are reasonable to expect with data mining. But don’t expect data mining to produce a miracle.
