First, let's set the scene for what weblogs are all about in the business world. Weblogs are pushbutton publishing on the web, making it very easy for “amateur” web publishing sites to spring up over night on any topic. Often communities of these sites form, with sites referencing each other and creating a sort of buzz. This phenomenon leapt to general attention in the election campaign of 2004 with bloggers actually attending the Democratic and Republican conventions. But, it has been going on for years in the tech sector (e.g., weblogs, inc.) and has been more and more pushing its way into traditional product categories (frequently corporate sponsored) like yogurt (Stonyfield Farm), cars (GM), and appliances (Maytag). At the end of 2004, people reading blogs accounted for 27% of Americans online, and that number is growing.

These weblog communities should be understood as social networks of people who are engaging in word-of-mouth discussion about your product. The great thing about weblogs from an analytic perspective is that all of this data is publicly available, easily accessed, and all but preformatted for analysis. Further, many free services already exist for aggregating this data (pubsub, technorati, tagsurf, feedster, bloglines).

You might think that the free services are all you need, and indeed, some firms make a business out of essentially just watching these and similar services for you (e.g., Bacon's) or gathering all of the information in one dashboard display that you can use for decision making (e.g., Expansion+). The key value proposition of these aggregation services is to automate and systematize the mundane aspects of data collection so that you do not have to devote (expensive) people to it while still receiving quality information. The services expect to make their money off of volume, and you should expect the prices of these services to drop as the novelty wears off (they appear not to be working off of proprietary data that would justify premium prices). Successful firms in this arena will have both figured out how to automate large parts of the data gathering process while still providing very high value, relevant data feeds (quality control still requiring a human element). You should demand that the firms provide you the information in standardized formats such as RSS (definition) to facilitate your own switch should quality fall.

Getting the data, however, is probably only about 20–30% of the story in terms of distilling strategic insight (based on typical cost of technology relative to effort required to derive value; compares favorably with my own anecdotal observations). Gaining strategic insight involves understanding what drives adoption of your meme and analyzing the data in a framework that will show you how to do it better. So, let's say you spend $1000/month on this type of service (I've not competitively priced them), you should expect to spend at least another $3000 to $5000 analyzing. That's around $50k for a fairly small scale operation, and based on the amount of time I have seen owners of small firms spend on analytics, this number is real. Firms selling web analytics (analysis of data for strategic insight vs. just providing data) at the corporate level (think to a major auto company or telecom) like to get annual billings over the whole corporation north of the $250k level, typically encompassing several brands and products. They can get this price because companies get actionable recommendations that significantly impact the bottom line vs. just data.

So, there's really a big potential market in figuring out how to provide efficient and effective weblog analytics. My guess is that CooperKatz is trying to plug itself in at this level. So, how might PR firms or yourself for that matter derive strategic insight for how to spread your meme from blog data? Answering this question involves answering several others in order.

Who is in your blog word of mouth network?

This is obvious and absolutely fundamental. It is not just who is talking about you at the moment (available for 0 cost from technorati) but who talks to who, the volume of posting, the relationships that seem to have developed. In other words, everything you would need to network in the real world but translated into the web world (PR firms should be a natural at this if they understand the web). Automated approaches (google searches, technorati, pubsub, FOAF, attention.xml) can surface the raw data for this, but people are still needed to make sense of it.

The analysts doing this essentially need to live in the web. In his recent podcast on G'day World, Rubel indicates that as of about the middle of January 2005, he had two interns focused on this type of research. Andy King indicates that this level of effort is not unusual for the sort of virtual shoe leather that needs to be expended.

How does information flow in your word of mouth blog network?

You might think that this is part of the first level of analysis, but truth be told, it is another level that has to take place over time. Who starts conversations? Who reads them? Who tends to agree with whom, and who tends to disagree?

To some extent, capturing this information requires reading the posts, a very labor intensive effort requiring people with good language skills. Think smart college upperclassman or recent graduates. However, there are some automated ways to distill the message that I will touch on next.

What is the message flow in the network at any given point?

What are people talking about? There are really three points here. What are the distilled one line or even one word messages? What are the specific posts that seem to be getting play? How are people categorizing those posts? Let's consider each in turn. First of all, you want to get a sense of people's gut reaction to your product or news. One way to do that would be to read the post and then force yourself to write an extremely short summary with several tags attached. I, myself, do this for between 10 and 20 posts a day on my link blog del.icio.us WebCites, and discussed my general strategy for doing so here and here (others do it too, Sally Falkow, Richard MacManus, and there is even a free server software to help, reblog). More automated approaches might use latent semantic indexing (definition) or other meaning extraction technique. Search engine tools companies such as Grokker or Nav4 sell products that begin to address this type of analysis. You might just be able to buy the google search appliance (the mini is $5k) and point it at an internally reconstructed version of the blog space you are trying to analyze. All of these automated approaches are in the interest of leveraging and focusing the human analysis I started this discussion with.

That's all fine for an overview, and the big picture is important, but frequently it is one post with lots of lengthy analysis that can be pivotal. How do you identify those pivotal posts? First, it is not just a question of checking the online buzz at a service like technorati. Such a broad picture is a start but is inadequate for figuring the structure of the interactions that led to the post, likely your key for figuring out how to manage its effects. To get this more refined picture, you need to construct your own search grid over the social network you identified initially. It might be as simple as skimming the participant posts which you have collected in a feedreader. If the space is big enough, you might want to find a script that will go strip out URLs and catalog frequency of mention within this group over a period of time. The high frequency ones are the posts you should prioritize reading (mind you, the day after the Super Bowl for instance, you might find highly mentioned but irrelevant posts; the trick is to make the aspect of finding the posts easy so that you can quickly discard the irrelevant ones).

The next question is how are people perceiving the high mention posts. Recent innovations in social networking software have focused on providing people with the means of tagging (applying labels to) web content (see del.icio.us, furl, tagsurf, flickr, technorati, and pivotal posts by Lou Rosenfeld, Clay Shirky, and Liz Lawley). As I have noted elsewhere, the key point here is how people are applying tags to these solid web artifacts. You might view it as sort of meme votes, with the tag getting the highest number of mentions serving as the consensus meme. Various tools exist for cleaning up tagging systems so that you get an aggregate view of the overall idea vs. the overall idea broken into many idiosyncratic sub-ideas. An issue with this type of analysis is that as far as I can determine, only the digitally elite are using tags, except perhaps for flickr, the photo sharing service, but flickr's tags are not easily queried or exported making their analysis much more tedious.

How is the message flow evolving?

Finally, once you have successfully established a system for answering all of the preceding questions in a systematic, repeatable, efficient way (something that will require a fair amount of customization for each product/brand space), you can start looking at trends. What posts continue to hold sway in the influential group months down the line? How is the perception of these posts evolving based on how people are tagging them? Are your efforts to influence the flows having a noticeable effect? You can think of more questions.

So, again, why should I pay someone for this?

I noted earlier, that this type of analysis requires significant effort. Firms who provide analysis like to see corporate billings north of $250k and attempt to provide the value to justify the cost. In a small company, you might think of yourself as having to dedicate a smart, college graduate analyst to this kind of task. Depending on industry and region of the country, that could run you over $100k with benefits, certainly no less than $50k. The cost comes in setting up and maintaining the right analytic frame, not in spread sheet jockeying. You should automate as much of this latter as possible, attempting to reduce its cost to 0. If you find the value is not there, cease the exercise.