Laura Downing, one of our founding partners, knows that I have a serious thing against averages. Every time someone says "the average went up", or the "average went down" she looks over to see my face contort in horror. So what's the big deal with averages?

In short, averages hide the story behind the data.

Take the case of rainfall. If we took all the NOAA Weather Service data recorded at each station in the country for the past few years, added it all up, and then divided by the number of weather stations, we would have the "average" US rainfall.

It may look something like this- a nice steady chart with an average around 62 inches of rain per year.

"Great," an observer might say, "we know the average, let's move on."

"Not so fast!" I retort. How do you explain the difference between these two images? They are both in the US but one is clearly getting more rain than the other.

And this is the heart of my distaste for averages. Sure, they are a valuable tool for getting a baseline perspective. But, they can also distort the story behind a measure.

In an ideal world, averages AND distributions should both be analyzed. So what is a distribution? Professional statisticians have all sorts of fancy algorithms and models, but I am simply talking about comparing each of the pieces of the average against each other. Only after understanding the underlying pieces can we understand what "the average" actually represents.

As an example, Wall Street salaries have been in the news quite a bit recently. What if you were to ask one of these firms to provide their "average salary"? They would do some quick calculations and be happy to report an average of something like $100,000. But the newspapers say they make millions, what happened?

Again, the story was lost in the average.

A simple average would say the firms combined salaries equaled $1.873 million. Averaged across 17 people, this equates to a measly $110,000, clearly very low for the risk and stress these poor bankers endure.

However, a closer examination might show the following information. There were 4 interns who made $2,500, then 2 couriers making $18,000, 2 landscapers making $21,000, 5 receptionists making $24,000, 3 Junior Partners making $55,000, and one Senior Partner making a whopping $1.5 million per year.

Distributions are simply a way of looking at the underlying data to see if the average is a good representation, or if the average is actually hiding a major outlier (the Senior Partner's Salary, in this case).

A quick glance at the chart above confirms "averaged salary" is neither a helpful nor informative measure.

Rather than saying the firm's averages salary is $100k per year (which is clearly false), a more accurate analysis might say that 13 of the employees make under $24k, while one single employee makes over a million bucks a year.

So, do I really hate averages? No. But they can be very deceiving. It is our job as data analysts and decision makers, to:

1. Understand the underlying data 2. Ensure "the average" is not oversimplifying a situation 3. And if the average is an oversimplification, we need to drill down and show an additional level of detail.