View Single Post
Old 04-28-2012, 06:07 PM  
signupdamnit
Confirmed User
 
signupdamnit's Avatar
 
Industry Role:
Join Date: Aug 2007
Posts: 6,697
SwirlsGirl, I did some number crunching for you and with the limited data I see it does seem as if April 2012 is a bit abnormal and cause for some concern based on the form submissions (but approval % seems fine and please keep reading!).

What I suggest is to take the last 3-5 years of monthly data and do something like I did below except use all the data you have. Enter it on a site like this http://easycalculation.com/statistic...-deviation.php. That will give you the mean (simple average) as well as the Standard Deviation among other things. With the standard deviation you then add and subtract it from the mean (see example below). That will show you the limits for data to be within one Standard Deviation. Typically in about 70% of cases on average in a normal distribution data should be within one deviation. To see what is REALLY abnormal (usually less than a 5% chance of it happening) go out two deviations.

Your April 2012 form submission % is outside of one Standard Deviation so it is a little abnormal. But it's also only a low sample size I have and April isn't finished yet.

Keep in mind that it would be better if somehow you could eliminate traffic from tier 3 (and maybe tier 2) traffic from all these stats as well as other bad traffic like the google bot. Ideally these stats would track uniques and not raws too. If a spider or a bunch or surfers from a Korean forum hit your join page 1000 times in the month then that can mess up your stats unless you eliminate it to prevent that.

Here's some calculations I did using only the last five years values for April that you gave :

Mean is simple average. Very easily skewed. Don't rely much on it.

Standard deviation is a measure of how much the data varies. It's still subject to skew but much less so than the Mean. The more data points you have the more reliable it tends to become.

"Normal" is just what I call the range of one Standard Deviation from the Mean. To get it take the Standard Deviation and add it to the mean. That's your top range. Now take the Standard Deviation again and subtract it from the original Mean. That is your bottom range. On average about 70% of all data should be between this normal range. But that can vary depending on the data and the sample size. It suggests something is off when outside this range but it's still probable that it can happen. Bad luck.

"Somewhat normal" I didn't do but it's just like what I call "Normal" but with a range of two Standard Deviations from the mean. Typically 95% of data points would be within this range. Being outside it suggests something is wrong but it doesn't mean impossible. I would be looking to see what is wrong pretty hard if I saw a month outside this.

Form Hits

2008 - 184
2009 - 899
2010 - 462
2011 - 275
2012 - 589

Avg Form hits: 482
Standard Deviation: 282

"Normal" (approx. 70% of time on average with large sample size and normal distribution) : 200 - 764



Form Submissions

2008 - 71
2009 - 240
2010 - 89
2011 - 101
2012 - 50

Avg Submissions: 110
Standard Deviation: 75

"Normal" : 35 - 185

Approvals

2008 - 55
2009 - 178
2010 - 80
2011 - 53
2012 - 37

Avg Approvals: 81
Standard Deviation: 57

"Normal" : 24 - 138



Form Submission %

2008 - 39%
2009 - 27%
2010 - 20%
2011 - 36.73%
2012 - 8.49%

Avg Form submission %: 26.24%
Standard Deviation: 13%

"Normal" : 13.24% - 39.24%


Approval %

2008 - 77%
2009 - 74%
2010 - 89%
2011 - 52%
2012 - 74%

Avg Approval %: 73.2%
Standard Deviation: 13%

"Normal" : 60.2% - 86.2%

This is of limited use because it's only five data points and because of what I said. So the Standard Deviation is going to be huge and almost useless. It would be better like I said if you did every month for the last three or five years. ;)

You might have someone make a Excel spreadsheet for you so you can plot this stuff and have a nice graph showing all the data and see where the "Normal" ranges are and maybe get stuff like the margin of error. I bet something like this already exists somewhere. I hope this is helpful. Good luck!

Last edited by signupdamnit; 04-28-2012 at 06:10 PM..
signupdamnit is offline   Share thread on Digg Share thread on Twitter Share thread on Reddit Share thread on Facebook Reply With Quote