My editor let me know the first copies of Social Media Analytics have arrived at the main office in Manhattan ( last Wednesday night) and review copies are just being sent out now.  I’m having a book signing in Manhattan at the offices of @PeekYou (whose advisory board I’m part of) on August 18th, from 6-8 pm – information is on the sidebar of every page on this site).  So, I’m excited – and can’t wait to touch my first printed copy.  Wow!   A year of work, now materialized.

Anyway, I meant to write about something I observed this week – eliminating noise in Social Media Monitoring and Analytics and came up an interesting idea that reminded me of what I saw in Collective Intellect CI:Insight earlier this year, when I still had access.  Collective Intellect has a visual interface that allows an analyst to cluster information using the semantic filter builder that I feel is one of finest interfaces I’ve seen.  NetBase, follows with similar functionality but it’s not visual (yet).  The semantic filter builder is captured in the movie below between 0:57 and 1:45 seconds into the video with additional dimensions which are custom created for clients (similar to the intent information OpenAmplify provides)  at 2:12 to 2:14 into the video.

Using Social Media Analtics to Identify Customers Concerned with Pricing from Collective Intellect on Vimeo.

The fundamental point was to disambiguate the choices we face when working with these systems – but it also shows me that there are back doors part of the same functionality, via Word Clouds, in platforms that support it such as Radian6.

I did a simple query that included (“New York City” OR NYC) and I wanted to see how much noise I’d get by sculpting my query and then regenerating the results – till I got what I wanted.

Got rid of “miss?—Can” , “owner”, “tomorrow”,  etc, each time getting closer to signals I wanted.

And actually, learning about patterns was helpful as there was all kinds of things I don’t know about my city – but you could do the same for anything, and query, and just keep sculpting till you have what you want.

It isn’t nearly what Collective Intellect is offering, but it is a way to segment a query, sculpt it, that might be better than looking only at a river of news, or verbatim results, which is much harder for the average mind to work with.

I was also thinking of a recent post by Gary Angel on Analyst vs. Implementer and another on Two Tiered Segmentation and decided that Social Media Analytics, similar to Web Analytics, could support the same segmentation that exists for Databases, as Gary Angel points out, but it would need a lot of tagging and metadata around each social mention which is currently not usually done – in other words, it would be very manual and tedious to replicate what tagging schemes exist in site Analytics into Social Media, but might be worth while, if were done (and someone was willing to pay for it – which is usually not the case).

For example, I rewrote what was provided in the second post against a possible use case to point out that a whole range of typical meta-data items that should be captured: 

• Functional Taxonomy: (examine mentions to decide what someone is doing – supported using OpenAmplify)
• Taxonomy: The hierarchical level a user it at (ie: stage 1, stage 2, stage 3, stage 4 of a particular journey or issue)
• Product Taxonomy: The product/family the mention concerns (e.g. TVs/LCD/ModelX, drug Y, disease Z, etc)
• Topic Taxonomy: A topic coding of what the mention or content pertains to (e.g. International Affairs/Middle East/Egypt/Revolution, shopping, illness, disease, etc – this take a lot of work to do manually, but may be the best way)
• Audience: The visitor falls into what segment?  (e.g. All, engineers, consumers, health-care providers, professionals, etc. – text mining of profiles may provide much of this information)
• Sales-Stage: The place in the sales stage the content is direct to (e.g. Early, Middle, Late)
• Page Components: The modules an online mention contains (e.g. videos, images, reviews, etc.)
• Component Classification: The value or status of the online mention (e.g. Overall Review Rating is High or Low, Price is Discounted or List, Availability is Out-of-Stock)
• Content Cardinality: The percentage on topic a mention is for a subject of the query
• Content Source: The publisher, source, author of the content (e.g. Columnist X, Database Y, blogger Z)
• Publish Date & Days since Changed: The recency and freshness of the content

This is a very partial list, my adaption of what was in Gary Angel’s post – but my point being we have two options to get this data tagged

1. Do it our selves manually (ie: adding tagging in Radian6, etc)  very time consuming and tedious, esp for large projects
2. Use robots to do it via platforms like Crimson Hexagon / Glide Intelligence, at el.

So, in theory, you could train Crimson Hexagon to pull data out and segment it along the lines of the above, once you translate what those  categorizations mean in terms of online listening and map it to the actual verbatim languages – but its a lot of work, no matter how you cut it… it probably can’t be done well in a MarCom environment because no one today would pay for this kind of work and there is hardly any in those situation that has right DNA to do it – but I’ll argue this is how we get value out of Social Media that becomes actionable in a new, and much more timely way.

I don’t want to over-complicate things, as people often want quick results with data (I do the same thing sometimes) but I think the “rough’em tough’em” approach to social media mostly fails where it could succeed, sometimes, in web Analytics and search Analytics.   The difference is that most of the data can be easily tagged and worked with if you set up is right with Web Analytics (though the setup might take a lot of work) and Search Analytics – but that’s almost impossible to do well, off the bat with Social Media Analytics.

Some data is structured, more or less, such as “likes” and “votes”, but the rest largely isn’t – anyone that thinks the “Wild West”, roll up your sleeves, quick and dirty approach works well in Social Media Analytics, such as you could get away with it using site and search Analytics,  is probably a PR based person that really value or understand data Analytics.

Kevin Hillstrom has an interesting post, or set of 2 posts on why not to hire him  (mine that data), but I think it’s actually meant to illustrate why they should (hire him); Kevin also wrote a testimonial for my book and I’m honored.