From average to segmentation and distribution.
You are asked to analyze huge amount of data to find paths and come up with an idea on how to boost conversions. And BOOM: you are calculating an average, you will end up with average findings! We all like averages because they are very simple to understand and to calculate. So we start to talk about Average Conversion Rate, Average Time on Site. We want to find that one number explaining and representing the complexity of the performance.
Let's assume that your users spend on average 20 seconds on your homepage. What does that mean? Is it good? Is it bad?
Averages sucks, yes it does. There are different corners from which you can read your data and take actions out of it.
1. Segment or die
Segmentation is crucial and most of the times underestimated. If your Average Time on Site is 20 seconds, you can for example check if there are differences amongst your channels: are users coming from Paid Search spending more time on the homepage compared to the ones coming from your Native Ads?
2. Fall in love with distributions
Distributions do the opposite of averages. While averages want to find the one number explaining the phenomena, distributions dissect the data and transform them in something more digestable. You can create clusters like the following:
amount of users spending 0-10 seconds on your site
amount of users spending 11-15 seconds on your site
amount of users spending 15-20 seconds on your site
etc...
If you figure out that 60% of your users spent 0-10 seconds on your side, can you get now how dangerous is to just look at the 20 seconds average?
Σχόλια