From Gary Angel’s Blog:
Recap on Dashboarding Social Media Webinar, Scott K. Wilder and I did our latest webinar on Social Media Dashboarding this past week and since we got quite a few questions, I thought I’d recap it (first), show some of the dashboard samples (next time), and answer the questions that I got (right after that).
We broke the webinar up into three major sections. In the first, we did the obligatory “why social matters” (I know – yawn), and talked about the heritage (PR and Brand Monitoring) of most of the tools used for social measurement. That heritage is important to understand, because it explains and highlights many of the weaknesses that an analyst coming to these tools should expect: poor aggregation, lack of automation, limited export capabilities, limited flexibility in setup, etc.
In the second section, we introduced my “3Cs” of social media dashboarding. The 3Cs, my own personal answer to Tom Shane, are Culling, Classification and Context.
Culling is the process of pulling the social content you want from the “river of news.” I call it culling because it’s very much an exercise in ongoing, patient weeding. Like my backyard garden, social networks tend to sprout the information equivalent of weeds much more often than they do fruits or flowers. Turning a social measurement tool on is a bit like connecting your garden hose to the city water main. Volume becomes the problem and weeding out the stuff you care about from the stuff you don’t a real challenge.
Social measurement tools don’t always provide a great set of tools for culling – mostly because they’ve been built on the assumption that the data is ultimately being read by people. With dashboarding, that’s often not the case.
So the first step in building a good social dashboarding system is a careful job constructing the information profiles you want. This is a lot harder than it looks –it’s challenging to tell the weeds from the flowers using simple keyword selection. Most of the information value of your reporting will be dependent on the types of classifications you make. That’s jumping ahead to the next “C”, but the two are closely related. Because of limitations in the reporting capabilities of most social measurement tools, the way you set up the software to pull the data will have a significant impact on the reporting you can do.
One other thing about “culling” – it’s a job that never really ends. Because conversations are so dynamic, a profile definition that gets you a clean set of conversational data around your company or product today may capture a boatload of junk tomorrow. This limits the potential for blind automation in Dashboarding with social – someone had better look at the data before it’s dumped into an Executive’s laptop.
After the culling process comes classification. Without classification, your dashboarding is limited to the basic metrics provided by social measurement tools: mentions broken down by source and influencer. With clever classification and trending, you can take these basic metrics and make them significantly more interesting to an Executive.
We identified five major dimensions of classification: topic, sentiment, source, influencer and impact.
Topic classifications are the most obvious and the most interesting. With clever use of topic classifications, you can measure a whole range of interesting things: from share of brand, to competitive share, to percent of support vs. marketing mentions by product/brand.
Classification by sentiment is increasingly supported by social measurement tools, but the level of that support is less than impressive. One of our listeners contributed that they only surface manual sentiment analysis to Executives, not automated sentiment analysis. I think that’s wise. Automated sentiment analysis as currently delivered just isn’t robust enough to defend.
Source (channel) classifications are interesting from a tactical perspective, but they can also highlight differences in campaign style and branding. Classification by source type is primarily useful when done when comparing brand/product or campaign mentions to each other. It’s in the differences in distribution of mentions by source type that most of the interest actually lies.
Influencer classification has two primary uses. At the tactical level, it can be used to identify people or pubs you need to talk to. That’s not necessarily the function for an Executive Dashboard but as an actionable component to a Marketing Report it can be useful. Another opportunity is to show how influencers are shifting in terms of topic and or sentiment. I’ll show examples of dashboards that capture each of these.
Impact is a different sort of classification – I mean it to cover reporting that measures how social mentions or activities translate into other measured channels (especially the web site). One of the core functions of dashboarding is to create a framework for how to think about a channel and its success. Impact classifications embody the answers you give what counts as a success when it comes to social media.
After classification comes the final “C” – Context. All dashboarding is ultimately an exercise in context. When you report on conversion and satisfaction, each provides context for the other. When you report on competitive mentions vs. brand mentions, you’re providing context for understanding true change. Social media is by no means unique or different when it comes to this final “C”. In fact, I think many social media metrics are inherently more contextual and easier to understand than, say, the web analytics metrics we often need to dashboard. But as I pointed out in the webinar, there’s less established practice to steal from in Social Media dashboarding so you have to work a little bit harder than you otherwise might.
Scott and I identified four major areas where social media metrics flow up into Executive Dashboarding and provide valuable context: branding, competitive landscaping, marketing evaluation, and trust & satisfaction. In each of these areas, careful measurement of properly classified social metrics provides interesting and novel insight.
All of this is a bit abstract, of course, and in my next post I’ll show some of the dashboards from the webinar, explain why I chose them, and how they illustrate some of the concepts I’ve talked about here.