The scholar Andrew Lang once said,

"He uses statistics as a drunken man uses lampposts - for support rather than for illumination. "

Unfortunately, I end up using statistics like a drunken man uses toilets - incorrectly and with my head hanging over the toilet-bowl thinking, "How did I get here?"

Community Analytics 

With all the statistics available to online marketers, how do you decide what's useless and what's useful? Time to dredge up an old school metric called Signal to Noise ratio. SNR is a simple way to trend the health of the SEOmoz community.

Appropriated by the blogosphere from electrical engineers, SNR measures the quality of your user comments. It is relatively easy to compute - divide a user's MozPoints by their number of comments - and as will be revealed could quite potentially unlock the mysteries of the SEOmoz universe. I use 'potentially' because, like all well-intentioned drunken ideas, I have to admit I haven't really thought this all the way through. 

Daniel and Rae's F%^king Excellent Adventure

My first task is to calculate the SNR for SEOmoz' top 100 members (I excluded the staff members because they're paid to play on SEOmoz):


After I'd done so, it doesn't take long to realize DanielTynski is a hidden gem. At 102 points and just 16 comments, he may be a man of few words but when he speaks, we listen. Daniel boasts an amazing 6.38 SNR, more than twice that of the next best. Daniel must be the epitome of the SEOmoz community, right?

Mayhap, because on closer inspection he has also submitted 3 (excellent) YOUmoz entries - one which generated 41 points - 40% of his total MOZpoints. Are YOUmoz points distorting the SNR?

Yes! Why? Um, I'm not sure, is there such a thing as analytics goggles? When I was drinking, SNR was looking a lot sexier.

Anyway, as my brain stumbles out of its stupor, I adjusted the original set of results with a little normalization to produce a handy weighted SNR.


With weighted SNR, Rae has everyone beat, coming in at 2.4 SNR (just ahead of Daniel at a very respectable 2.2). This immediately teaches me one thing.

Swearing a lot is the key to racking up MozPoints!

Welcome to drunken analytics, solving mysteries of the universe one at a time.

And If You Don't Know, Now You Know 

So now you how the metric works. What else can the notorious S.N.R show?

Stuff.

SEOmoz' top 100 members (weighted SNR in brackets):

  • SNR = 1.55 (1.44)
  • Excluding staff = 1.35 (1.23)
  • SEOmoz Staff only = 2.08 (2.08)
  • Premium Members only = 1.49 (1.35)

Premium Members have a higher signal-to-noise ratio than regular members. That's a great reason to sign up to Premium, right Rand? 

SEOmoz top 10 (from SNR100, excluding staff):

  1. DanielTynski
  2. sugarrae
  3. shor
  4. Igor Mordkovich
  5. stever
  6. Bud-Caddell (who inspired this post)
  7. Nathania Johnson
  8. feedthebot
  9. ciaran
  10. WebGeek

SEOmoz top 10 (weighted; excludes staff)

  1. sugarrae
  2. DanielTynski
  3. shor
  4. stever
  5. WebGeek
  6. kimber
  7. microdesign
  8. mbarr
  9. willcritchlow
  10. Dr. Pete

The comments of the top 10 are typically interesting and add value to the conversation (except for that shor dude, he talks about toilets).

SNR Head vs. Tail (excludes staff):
  • Head, top 20 members: SNR = 1.43 (1.27)
  • Tail, top 500 members: SNR = 1.29 (1.18)

Weighted SNR of Top 500 members, including staff:

 

Note the concentration of high signal users in the top ranks and, more importantly, there are very few users below the lower control limit - the SEOmoz community is healthy! 

SNR is a simple metric. Some would call it silly (just not to my face). It can't adjust for articles and comments predating the MOZpoints system, nor does it include SEOmoz staff entries. Should we refrain from engaging in banter, me-too comments and taking the 'bad cop' stance because of these results? No, because the community would become narrow, single-minded and suffer for it.

What do these statistics actually reveal? A lot. Or maybe nothing. As they say, torture numbers and they'll confess to anything. It is up to you to decide what the key performance indicators for your community are. Once you have these KPIs, my best advice is to set up those nifty upper and lower control limits to filter out the statistical noise from the signal.

The moral of the story? We might never invent a hangover cure, but at least Avinash Kaushik has analytics covered.

To paraphrase:

Signal minus Noise -> Insights -> Action -> Happy Community -> Traffic, Customer Satisfaction, Money!

How do you deal with signal versus noise?

What metrics would you suggest to measure the health and quality of your blog or forum community?