I have the following information: EPC (earnings per click) for the previous 12 months across an entire portfolio of websites are:

$$ \begin{pmatrix} Jun 18 & 0.11 \\ Jul 18 & 0.12 \\ Aug 18 & 0.15 \\ Sep 18 & 0.16 \\ Oct 18 & 0.15 \\ Nov 18 & 0.19 \\ Dec 18 & 0.24 \\ Jan 19 & 0.15 \\ Feb 19 & 0.17 \\ Mar 19 & 0.15 \\ Apr 18 & 0.14 \\ May 19 & 0.17 \end{pmatrix} $$

For one specific website in that portfolio, the EPC over the past 4 months are

$$ \begin{pmatrix} Feb 19 & 0.89 \\ Mar 19 & 0.97 \\ Apr 18 & 0.93 \\ May 19 & 1.14 \end{pmatrix} $$

Making assumptions that no other external factors can affect EPC and that the industry is cyclic (i.e the increase in December will happen again), how do I go about predicting the EPC for November/December this year?

I initially calculated Z scores for the yearly EPC and was going to use this to calculate the probability that the EPC will increase at the same rate, however I don't think that is possible here.

I am effectively trying to do a simple linear regression model but with my dependent variable being my independent variable?

How could I go about trying to figure out a rough EPC prediction based on this?

The reason I don't want to open the regression model to what it should be (i.e multi variable regression model with clicks and earnings) is because a) I am making this in Excel so I'm not sure how to write it up and b) because I have to present it tomorrow and I'm not sure if I will be confident in it in time.

Edit: I changed my tactic slightly. What I did was calculate the average EPC across the entire year. I then worked out how much of an increase the EPC in November and December had, against the average yearly EPC.

From this, I can see that there is an increase in EPC around 16& in November and 34% in December.

The average EPC for the listed website has been around 1.012.

So assuming this is line with the yearly average, I can predict that this EPC will increase by 16% in November and 34% in December. Can this work? I hate using average's so much, but is there another way?

I know, 100%, that this is a really bad model to predict EPC on. But like I mentioned, with the time I have I'm not able to do it using the multi linear regression model and so need a patch up job.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.