ROI under scanner

Now lets assume that you have somehow reached a consensus and have finalized your business metric. Good now start tracking it. But as soon as you start getting the data, you realize that you are getting traffic from different sources. The money spent on each of the source is different and the tracked metric has a different value. In other words, each source has a different ROI.

The next buzzword is "optimization". You want to optimize the money spent on advertising and increase your ROI. For this you need to stop going to some sources and promote other sources till they saturate, but the problem is to know where to draw the line. What is the ROI, below which you don't want to advertise and above which you want to promote?

It is a general practice to use average as the beacon below which you press the panic button. Several businesses use average as the sacred no. below which to demote and above which to promote. But is average really a good number? Let me use an adage, I heard from a statistics professor of mine - "Average is a bad thing, remember the guy who drowned in a river that on average was only 4 feet deep."

The point is that average can be skewed BIG TIME due to outliers. Lets take an example. If you are riding a bus with commoners, the average of the bus passengers could be a few thousand Rs. a month but if at the next stop, the Ambani Duo board the bus for some unexplained reason, the average salary of the bus passengers can go through the roof. This is because the 2 newcomers earn more than few crores every month. These passengers are considered outliers and they skew the averages from 1000's to 100000's.

A better option is to keep an eye on the median, mode and average before deciding where to draw the line. You can also use frequency table in case you are stats savvy and that will give you a much better perspective.