LJ 2009-01-08 14:14:00
I hate volatile numbers. Bane of my bloody existence, they really are…
Say I have a company, which generates Â£100 in revenue, and has expenses of Â£90. I make a tidy Â£10 profit, which is what everything is going to be measured on.
Now let’s say my revenue dips by 5%, and my costs increase by 5%. Those are small margins, right..?
Well now my revenue is Â£95, and my costs are Â£94.50. My profit is consequently 50p, down 95% on what it used to be. All because of the cumulative effect of two relatively small variations in numbers. It works the other way too, of course – to double my profit in the original scenario, I would only have to cut costs by 11% or increase revenue by 10%. So it doesn’t favour movements in one direction over another..
The problem is that this sort of instability is a bastard when it crops up in analysis. The accuracy of your results pretty much goes out of the window, because all it takes is one blip and everything goes to crap.
So, for example, when measuring the efficiency of the fraud project I’ve been on for the last couple of months.. The calculations involve various points where a single anomalous result can throw an entire month’s performance down the toilet, it’s not at all good.. And as the initiative sees more success, the level of fraud that’s even attempted drops as fraudsters start to give up, which means fewer cases, which means less statistical validity in the analysis due to poor sample size.
Sometimes the numbers just don’t want to let you win..