Folder:
111 Epic
File:
111.10.05 Epic - Averages are dangerous

Averages are dangerous

The "average" of a group of numbers (also called the "mean") is an easy way to understand the data. But it is also dangerous because it can hide important information.

Averages hide layers of information

If you had 3 products selling over time but only looked at the average of the sales of the products, you would not see the important detail about the movement of the product sales over time.

  • Formerly low performing segment is now the best performer.
  • The Average of all segments tells you very little.
  • Formerly high performing segments are now poor performers.

Averages which aren't based on sufficient data points can be misleading

Let's say we are comparing the profit margin of two landing page variations.

Landing page 1 profit margin: $24
Landing page 2 profit margin: $32

At first glance you'd want to promote Landing page 2 because the profit margin is significantly higher. But if the number of conversions for that page are significantly lower, the data is not valid.

Landing page 1 conversions: 99
Landing page 2 conversions:  3

You clearly can't trust the second landing pages results until it has more conversions.

Averages of dispersed data can be misleading

The average of 2 and 98 (50) is the same as the average of 49 and 51.

  • https://towardsdatascience.com/never-cross-a-river-4-feet-deep-on-average-d1a8d1ec345c

Source:
  • Me