First I hoodwinked Mom into blogging - she now has two blogs and a Flickr Stream (she’s even using Twitter).

Now my sister Ali has started her own blog!

Here’s the list of links for family, friends and anyone else who is interested:

• Hays Travelogue
• The Date Book (It’s not what you think)
• Mom’s Flickr Stream
• Ali Marie’s S.I.N. (Service Industry Network)

I am totally looking forward to Ali’s anecdotes and insights from the front lines of the service industry!

Now if I can just get Nate (and Family), Dad (and Dee) and Grammie (and John) to join the game…

I updated to wordpress version 2.1 last night. Hopefully this blog is now faster and more stable. If you notice any problems, please let me know by leaving a comment or sending me an email (civy at instantcognition.com).

Avinash posted today on “Leveraging Statistical Control Limits” in web analytics as a way to cut out the noise in your KPIs and therefore being able to focus on the true outliers - data points that are truly worthy and in need of your valuable brain power.

In his post, Avinash challenged his readers to provide an Excel solution since there are no OOTB (Out Of The Box) solutions for introducing constants, statistical or otherwise, into charts - Excel tries to box us into using just vertical and horizontal axes scale controls.

Well, I can’t pass up a challenge like that right?

What do you think of this?

So what’s going on here?

1. Named Ranges - Named ranges are your friend. In this case, I’ve got named ranges for each important series of data, especially my key metric - page views per visit. That way, I can insert a new row of data and my constants (Standard Deviation, Mean, Upper Control Limit, and Lower Control Limit) will update automatically. BTW, the chart uses the named ranges as well so that it will update automatically every time a new row of data is added.
2. The chart is a combination of lines and stacked bars. The stacked bars create the gray biorhythm (normal variation) area. Line charts are used for everything else.
3. For the top 5 outliers (either upper or lower) I’m dynamically adding a data label so that I know, at a glance, the specific values for them.
4. I’ve eliminated all horizontal lines other than my own constants because they are just distracting

You can download the spreadsheet and play with it if you want.

A couple of caveats:

• Named ranges won’t do the dishes for you. In other words, when you insert a new row, you’ll need to copy down the formulae from the previous row
• Excel doesn’t always do a good job of placing data labels so the outlier labels may need to be massaged after an update
• My spreadsheet only uses 1 standard deviation. While 3 might be a six-sigma best practice, three standard deviations would have put the lower control at 0.1 and common sense tells me that it shouldn’t be possible to have an average page views per visit value lower than 1.

I’ve deliberately not gone into great detail on the inner-mechanics of the spreadsheet (not that their profound or anything) because that would make for an impossibly long post - if you have a question please post it in the comments or email me directly (civy at instantcognition.com).

Update: I’ve had to (temporarily) disable Snap on my site - it was giving my site search gran mal seizures when it tried to spider the site.
I finally got my invitation to Snap - among other things they have a service called Preview Anywhere ™ which, like Ask.com offers roll-over previews of the destination site.

Try it now by rolling over this link to Snap.com.

It can be configured to preview either external links only or all links.

Further customizations allow you to disallow certain links by using a special class. Conversely, you can set it to not preview any links other than those with a different special class.

There is a wordpress plugin for it, once you’ve got your invite.

I have noticed that it takes some time to get image previews for some of the more obscure links on my blog, but I imagine that will improve with time as they grow out their image database.

I wonder if I can figure out a way to track any searches initiated from the rollover? Something to think about…

Personally, I like this functionality, but let me know what you think about it on this site (or others, TechCrunch is using it) by leaving a comment…

Jeremiah Owyang posted today on a Round Table he attended with the folks of Factiva and some others (bloggers, corporate program managers, PR consultants and social media practitioners) to discuss what should be measured in the Social Web. They came up with a nice list which they voted against and ranked according to the most votes.

Here’s the list:

1. Participation & Engagement
2. Influential Ideas: Memes and their intensity over time
3. Relevance
4. Sentiment/Tone/Opinion/Favorability/Emotion
5. Content
6. Relationships & Connections
7. Analytics & Activity
8. Community Activity or Call to Action
9. Reach
10. Conversational Index/Engagement Tied With Demographic/Who

Because Jeremiah did not provide definitions, I’m going to take a jump out into the blue and start playing word association…Participation & Engagement sounds like something to be measured while the Conversational Index sounds like a KPI that could be used to affect that measure - for blogs. As defined by Stowe Boyd the Conversational Index is:

[# of posts] / ([# of comments] + [# of trackbacks])

An index score that is less than one indicates an engaging blog while a score of 1 or more indicates an unengaging blog.

While I agree with the principal of this metric, it seems to me that this score is too easily gamed - and maybe not even on purpose.

In blogging, one of the best practices I’ve heard is to interlink between your posts (Jeremiah does this a lot) which is very helpful to the readers but since platforms like wordpress will treat that intralink as a trackback, the total # of trackbacks will be polluted with your own linking and therefore susceptible to gaming.

Next, when we’re looking at comments, this is a conversation where (hopefully) the author is responding to the comments from her readers. Now, this could be gamed if rather than responding to comments normally, the author starts to break her comments into smaller bits in order to drive a better CI score. On the flip side, bloggers that tend to write shorter posts more often as opposed to bloggers that write longer posts less frequently will be penalized simply because of their writing style.

So what to do? How can this excellent kernel of an idea be taken to something that might be acceptable industry-wide as a standard? I don’t have the perfect answer but here are some thoughts:

1. Intra-blog links need to be excluded from the trackback count, shouldn’t be that hard to do from a technology POV
2. Normalization: Because the CI is at the mercy of the writing style used, a way would have to be found to normalize the score across a spectrum of writing styles from very short/very frequent to very long/very infrequent
3. Author comments - I really don’t have any sort of clue on this one which either makes me incredibly daft or makes it very dangerous to the integrity of the CI score. My WAG is that some sort of algorithm would be needed to determine if the author’s commenting style is a deliberate attempt to positively impact the index.

I hope I don’t sound like I am beating up on the Conversation Index because I really like the idea, so let’s move on…

Memes and their ‘intensity’ over time seems to me to be very much like Reach or at least a special kind of reach. Lacking a definition of intensity, I guess that it means something like how quickly it spreads and how far it spreads and then the question is; how does it pace over time? Isn’t this like Reach?

Relevance: Relevant to What? To Whom? Either this is very atomic in nature (in which case measurement is going to be meaningless) or it means measuring to how many people something is relevant to - wait isn’t that Reach?

Sentiment/Tone/Opinion/Favorability/Emotion - Looks Like A Duck, Sounds Like A Duck so it must be Voice of Customer data. The one difference here being that it doesn’t necessarily require a survey or contact us, rather it can be directly scored (although somewhat subjectively) from blog comments, trackbacked-posts, etc.

Content - Umm… not enough to go on here

Relationships & Connections: Again I ask ‘isn’t this Reach?’ Maybe it’s reach in the context of a LinkedIn but it still looks like Reach.

Reach: ‘Nuff Said

Conversation Index/Engagement - See Engagement/Participation

Demographic/Who: Who is always important whether its from a demographic, technographic or behavioral standpoint. The CMO or advertiser or product manager will always want/need to know who is viewing/engaging with content.

Of course, that’s just a critique of the list that the round table came up with and not necessarily what Jeremiah would want to know from some one in the business of measurement.

But that will have to wait, this post is already making me twitchy from it’s length so I’ll have to do a follow up post on what I think should be measured in the social web - don’t let me forget.

Update

We’ve been having a pretty lively discussion about engagement over at Eric T. Peterson’s Blog, his posts on the topic are:

How Do You Calculate Engagement? Part 1

Frank Faubert Writes In

How Do You Calculate Engagement? Part 2

How Do You Calculate Engagement? Part 3 

Additionally, Eric has started a Web 2.0 Working Group over on Google Groups discussing methods, etc. for measuring Web 2.0