You know, the technology in the online world is advancing at a ferocious speed. There’s always a new fangled web ap that’s being introduced that some PR folk will experiment with looking to fit into their day to day roles as practitioners. For example, take a look at Kevin Dugan’s 10 Flickr hacks, Steve Rubel’s 10 Technorati hacks and, boastingly, my own 8 moblog hacks (because I couldn’t think of ten). All very useful… maybe.
But something I’ve been interested in a while now and something I’ve been doing at my new place of work (which is going swimmingly well by the way) is online brand monitoring. I believe there are huge benefits in discovering what people are saying about a particular brand by way of detailed analysis.
Like most things, finding out useful data comes at a price; both in the monetary and time sense of the words. There is no quick fix.
Listening to what is being said online about a global brand will take a whole lot of human resources because, as of yet, there’s no tool that can measure or read sentiment. And that’s where the meat and potatoes of brand monitoring lie.
This is a great post by Mike Manuel from November last year on what he calls (and what I’ve adopted) “mining and monitoring the conversation gap”. Mike describes two types of methods for listening to online conversations:
The first is ‘conversational mining’ which means digging deep into past conversations over a certain period of time. Say, for example, the last six months. As Mike mentions, conversational mining is good for “keeping tabs on big picture trends and sentiment shifts but it’s lousy at capturing real-time issues as they typically bubble up online.” What I would add to that is, conversational mining is a great first start for an organisation that hasn’t typically done any kind of online monitoring in the past. It gives them an idea of the current situation and positioning of their brand and from that they can identify their objectives for any online PR activity.
Depending on the brand (obviously there’s a lot of online discussion going on about global tech brands) conversational mining is very time intensive. No simple graph from Ice Rocket or BlogPulse will do. It’s about digging down and find out where the negative commentary (if there is any) coming from; it’s about identifying issues that have been discussed in the past which don’t necessarily have to come from blogs either. Message boards and forums are still primary drivers of conversation on certain products.
Brand monitoring doesn’t necessarily mean keeping tabs on just blogs.
For some brands, UGC sites like the immensely popular Digg should be searched for past articles relating to their product(s). You’d be quite surprised at the plain derogatory or false remarks from some Digg users.
For me personally, I believe conversational mining is only effective when the conversation sentiment is measured. Knowing the topic of conversation and measuring it as simply as a ticking check boxes for positive, negative and neutral tones. But, as mentioned above, there’s no tool to do this yet. There are, however, some advancements into curing this bane of brand monitoring. Development into sites like Opinmind would help make the process a whole lot easier. (Are there any other sentiment measuring sites?)
The second method of listening is ‘conversation monitoring’ which pretty much means listening to what’s being said in real time. Again, this takes time but the hard part is figuring out what to do if or when it does happen. If your brand is being discussed does it mean you should necessarily engage in the conversation? Or should you continue working in the shadows?
Conversational monitoring is beneficial when you’re already carrying out some sort of PR activity online. That way you have a benchmark from which you can measure the success of your efforts.
One thing is certain though. If the process of brand monitoring is to be made easier and done quicker, then more time needs to be spent in R&D specific for this area. Maybe it’s only a matter of time.
technorati tags: branding, brand+monitoring, conversation+mining, conversation+monitoring, stephen+davies
Stephen is a communications consultant based out of the UK. You can connect with him on Twitter or check out his LinkedIn profile. | Email Stephen
Stumble it
Digg it
Deli.icio.us
Tweet this
Michael
Thanks for this interesting post, Stephen. No doubt: Content analysis has always been a key instrument in PR evaluation – and I agree there’s no reason why these efforts should be limited to offline media or web 1.0. As you pointed out (and as Opinmind shows) this might be automated to a certain extent in near future – and chances are high that there will be algorithms delivering high quality results, such as taking into account the number of page impressions of all content analysed. (To answer your question: No, I don’t know any other sentiment measuring sites by now.)
However, although categorising conversations into positive, negative and neutral sentiments might be a reasonable tool for (quick and dirty) PR evaluation, it does not give any starting-point for setting up a (online) PR strategy. It will (in my humble opinion) stay vital to analyse web content “by foot” in order to find patterns and frames that will give us more insights about the way people talk about our brands than these mere categories “friendly” and “hostile”. There might be powerful software assisting us (such as AtlasTI), but there will always be some manual work left for us. Which is not too bad though, I find…
Kevin Dugan
Stephen – Good stuff. Personally I think folks get paralyzed by the great unknown. They have no idea what is being said or how much and where.
Conversational mining helps them get a read on the level of activity and anecdotal input on the type of converations: good, bad or ugly.
From there they need to be prepared to spend some upfront time scouring the conversations and social media sites like the ones you mention as well as Flickr, YouTube and MySpace. You are right that it’s beyond blogs now and Google is no longer enough to get an accurate, quick read of activity.
New tools are needed to make this initial process easier and to make monitoring easier in the long term. Now you are relegated to creating a ton of feeds and an internal process for tracking everything.
It’s still hand intensive and we need to make it more head intensive. Hopefully a whip-smart software startup is monitoring this conversation and has an idea they’re working on that will answer our prayers.
There are enterprise-sized solutions like Brandpulse, but smaller companies need a cost-effective solution.
Stephen
Micheal: Good points and, like you said, chances are we will see a development (fingers crossed) of designed algorithms delivering better results. Hmm, not too sure about the manual work though … surely it’s the PR pro’s job to make sense of the results rather than finding them?
Kevin: “It’s still hand intensive and we need to make it more head intensive. Hopefully a whip-smart software startup is monitoring this conversation and has an idea they’re working on that will answer our prayers.” Amen.
Kami Huyse
Awesome stuff, I espeically like the Opinmind resource. The Neilson Buzzmetrics BrandPulse offering sounds interesting, but I don’t know much it costs andI agree with Kevin that there is a business opportunity here for someone.
Stephen
Hey Kami,
I’m not too sure on how much it costs either. I have used it but didn’t deal with the subscription.
Matthew Hurst
The problem of sentiment mining is an active research area. Havea look here for a partial bibliography of this research.
Automated sentiment mining is also a standard feature of many solutions in this space (BuzzMetrics and Umbria both offer this). Solutions like Opinmind, while interesting, are not as sophisticated and concentrate on precision rather than recall (i.e. improve the chances of a result being correct rather than ensuring that the coverage is complete).
You are right that monitoring takes time and effort, but that time and effort doesn’t have to be spent by you. It can be invested in technology. This is the investment that we are making at BuzzMetrics. Note that even with this automated solution, there is still a central role for a person to interpret and manage the analysis.
I’m glad to see that you take into account other forms of social media – if you are interested in free tools, you might want to check out BoardTracker, one of the best message board search engines.
Note: I work for Nielsen BuzzMetrics.
Stephen
Thanks for that Mathew. And thanks for the link to the research paper. I agree, Opinmind isn’t the correct solution for, shall we say, a major brand to use, but I feel the technology behind it has a lot of potential.
But do you think recall is better than precision? Or did you mean both is best?
I agree also – no matter how advanced the technology becomes, a human being will always have to interpret and make sense of it all. And, of course, from a PR point of view, someone will have to make a strategic decision from where to go from there.
P.S. I’m a fan of your blog.
David Phillips
Stephen,
I think there are some interesting things that can be added.
First is the extent of the literature (you may find my bibliography helpful – about 500 references http://www.managementclarity.com/big_bibliography.htm.
The use of latent semantic analysis is probably the nest bet for finding out concepts in extensive corpora. I have an engine used in research that identifies concepts.
It was created to prove the Relationship Value Model (the research will be published next year both as a research paper and in John Leddingham’s new book)
The advantage with this approach is that it can also bring topics to the fore and create ’summaries’ by subjects. Thus in, for example a political or industry sector corpus it will identify the people and their interests and the ‘conversation’ around the subject.
If you would like a demo, let me know.
There is the issue of the ‘long tail’ and once again, this approach helps.
I have to say I am delighted that you are doing this work. I was under the impression that it is rare in consultancy.
Stephen
Thanks for the link, David. I’ll take a look.
Yes a demo would be great some time too.
P.S. Have you ever considered a Wikipedia page for the Relationship Value Model? It could bring extra participation.
David Phillips
I shall be publishing empirical evidence for the Relationship Value Model next year (chapter in John Leddingham’s new book and a research paper). It may then be an idea to add it to Wikipedia. The corpus is HUGE and the methodology will be replicable which should mean that there will be a basis for a further developed research base for Public Relations.
It should also bring PR out from under the yoke of Marketing once and for all
Alexandra Pullin
Good to see you back and blogging Stephen!