The basic issue here is that when you make your calculation tabs and columns presentation ready as a report should be, they don't do a very good job of performing the calculations in a logical, auditable fashion. You end up having to hide rows, or make very complicated calculations to go from line to line to fit the big picture metrics in one table. Likewise when you do too many calculations in your reports, you'll have that secondary report that will need the same calculation done to it and before you know if you have slightly different numbers when they should be exactly the same number.
Following along the post below about keeping each tab pure, a common issue I see is people who use their calculations tabs for displaying and printing reports/outputs, or who try to do place calculations into a report to get the report they want.
I understand why this happens. The modeler will provide a nice chart or graph to the boss for a presentation, but it may show slightly different numbers than the boss has already represented to the client. There is no ethical issue here to adjust the numbers up or down, or maybe you're moving to even more conservative numbers. So the modeler goes into the report, makes a change directly on the table and voila, everyone is happy.
Unfortunately, the detail table that's on the back up slide doesn't get changed, or if you do remember, it doesn't get changed in the right direction because there is a lot of moving variables. So when the client wants to dive deeper, it's an embarrassing look. How to prevent this is to keep the tabs pure. If the boss wanted a more conservative number, go ahead and make a quick change to the underlying calculation tab and then it will populate through to all the tables, and all the downstream calculations that feed off it.
The other error that occurs from changing the report directly is that if you've done this "on the fly" then when you refresh the analysis 6 months later you forget that you've 'hardcoded' that one cell in the report and when you refresh with new data, that one cell doesn't change. No big deal if it's noticed, but if it's not then it's just as embarrassing.
No comments:
Post a Comment