I just did a test for a company for a job interview. I had information regarding Company Madeup, for who I was doing some analysis. I found out that, over the course of 2018, Company Madeup's website had made money for Shop Madeup. I created a table with this information, which had the following headers:
Year - Month - Total Clicks Sent to Shop - Total Commission made by Company
From here, I worked out how much commission was earned per click (i.e total commission/total clicks), for each month.
Even though the data was for 2018, I wanted to be a bit more proactive and use my analysis to predict something for 2019. What I wanted to do was to predict earning per click (EPC) for 2019, based on trends seen in 2018.
The way I started this was, after calculating all the earnings per click (EPC) for 2018, I created a distribution model (as I now have a mean and standard deviation).
I then worked out Z scores for each of the EPC value. This would let me see how good the EPC for that month is, in relation to the rest of the year. The aim was that using the Z score, I could calculate the probability that EPC in the next year would increase by the same amount. I never got round to this but I want to ask a few questions on it as I have an interview tomorrow to go through my result, so hopefully can talk about it there.
Is that sound reasoning or have I completely missed a step? Is there an alternative way to do this?
As a follow up also: I calculated my Z score for one month to be 2.46, which translates to a probability of .69%. Does this mean that probability that EPC in 2019 will beat the EPC in 2018 is 0.69%?