An aside - Pushing Intelligence into the Software

I digress from this discussion on the three factors needed - to touch on a topic that came to mind due to recent events at work.

The concept that I discovered was happening, or could happen, is that intelligence can be moved out of the database and into the software applications that are above.

Let me try and explain this a bit better.

In the surrounding applications there are many calculations done as a part of the routine operations of the organization. It is possible that application designers will refrain from storing interim results, or potentially final results, under the theory that they can always be recalculated by the application.

Saving database size may not be the issue, but releaving the database engine of the additional burden may be a deciding factor.

I must point out that I do not have direct evidence that this is happening, but it would not surprise me that application developers could be taking this approach, and might be within your organzation.

The danger from this condition is that interim results will not be immediately available to the data analyst to create the information from the data. Instead the interim results must be re-created. This is a task that may not be onerous, but does require additional resources and complications during the data analysis steps.

A second concern is that the calculations are being done twice, in two different environments, and there is a possibility of differences in results occurring.

"But that can't happen", you say. The computer will always get the same answer.

That is what I thought too, until I discovered that scientific and accounting rules for rounding are differet.

The scientific (and correct) approach is to complete all calculations and round to the desired accuracy at the end.

The accounting approach is to round to the nearest cent (two decimals) after each calculation.

These two rules can produce differences.

When dealing with auditors this can be a problem.