A data management department at a pharmaceutical company has collected a set of data on database lock times (defined as the number of days from last-patient-visit to database lock) between two groups within the department. A quick calculation of the average database lock times shows that:

* Group A = 76.6 days
* Group B = 58.4 days

and therefore the immediate conclusion is that Group B is performing much better than Group A at locking databases.

Unlocking the Power of Data

The raw data for the database lock times (Days) for Groups A and B are as follow:

• Group A: 15, 20, 22, 23, 23, 24, 45, 126, 175, 293
• Group B: 21, 28, 32, 42, 61, 67, 76, 82, 84, 91

The main question here is what we mean by "better". For example, as long as the databases are locked within 25 days, it is considered to be good. Based on our data, we find that the proportions of studies that have been locked within 25 days to be:

• Group A = 6 out of 10 studies
• Group B = 1 out of 10 studies

So in this instance Group A is better at locking databases than B. Also we have found that the reason for the 3 studies in Group A that have taken over 100 days to lock is due to the low priority of these studies - they are not considered important enough to allocate sufficient resource and effort to lock these studies quickly.

Enhancing Knowledge and Performance

Often we collect a large amount of data, and we only skim over the surface and come up with the results that may lead to the wrong conclusion and decision. In this example, we have learnt that while Group B may be seen as "better" than Group A at first, it really depends on:

• what is mean by "better"
• what are the real underlying factors that affect the performance when we examine the data more closely