Segment, segment, segment

Patterns of analysis: Segment, segment, segment

Typical day:

Calculate a rate ( sales / traffic )

Looks low, I thought it was higher wonder how has it been changing?

Plot a rate over time (trend pattern)

Oh, I see a huge drop this week, what up with that ?

Segment: create 2 separate charts, one with the trend of sales and other with trend of traffic

Oh the trend of traffic had a huge jump this week, so looks like we are getting loads more traffic that are not making a purchase.
Realize we’ve launched a mail marketing campaign this week sent to millions of users

That explains it, loads people got an email they came to see what’s up, not as many are interested and leave (bounce).

Method

When puzzling over why data is looking like it does, is a good idea to segment (a lot) it further to pin point exactly the data point, or variable that is influencing the top level change we are seeing.