September 15

Metric of the Month September: Brand Measurement

by Emile Daigle

Hello and welcome back to Molecular’s Metric of the Month. In this series, we hope to bring valuable insights into how different measurement strategies can help start or build better client relationships. After all, measured results are proof-positive that the work we do has real impact and over time builds trust.
metric of the month - social media
This month’s post will be discussing various aspects of social media and the analytic methods that can help us to derive value from the never-ending stream of user generated content (UGC). Since we’re talking about brand measurement in particular, we’ll be focusing on how to derive information about companies, products, sponsorships, and any other brand. With the advent of low barrier-to-entry online publishing, the rate at which new material floods into the internet has grown explosively. No longer must one brave the moral perils of USENET to get their words out; freely available blogging sites have made authors of us all. As a matter of fact, last month I set up my mom’s first blog, and my extended family has almost forgiven me for it.

First let’s briefly go over a few classes of UGC and touch on potential business intelligence we can look for within them. Forums are community-driven sites that most often share a core subject like photography, muscle cars, woodworking, and other enthusiast-enabling foci that will constantly have members at all points along the learning curve. Quality of posts may vary wildly, but most communities self-moderate and high-value material can be buoyed up using engagement metrics such as read rates and reply rates. In other words, a more engaging post represents higher value content. Through attribution of high engagement posts to authors, we can then identify key contributors within the subject domain of the community for further engagement. Having such a domain specific cluster of UGC also increases the relevancy of any trending we may do such as a timelines of brand or product mentions, but we’ll get to that later.

We all know blogs, we all read them (a self-asserting truth!), and many of us write them. Shortly after the explosion of blogging, a few clever people applied traditional web metrics to them to derive some useful things like readership and reader loyalty, but those belong in another post. After that came more medium-specific metrics like subscription rates and pingbacks (a protocol giving authors notification when someone links to their post). These in turn began to establish the interconnectivity of it all and to identify the Who’s Who of the blogosphere. Finally, by aggregating every blog that can be found, it is possible to glean more interesting factoids like:
• how many people are talking about a certain subject or brand
• how the number of discussions changes over time
This is the birth of brand measurement.

Then along came microblogs, an intersection of instant messenger “away messages” and blogging mashed together and cut into bite small enough to author and read via SMS. Blogs had taken the technological barriers down, and microblogs lowered the commitment of time and composition of posting. Microbloggers began posting an order of magnitude more frequently than “traditional” bloggers. What’s it all mean? A motivated analyst can tap into a nearly real-time stream of hundreds of thousands posts, as long as you can keep the infamous fail whale at bay. Messages are generally restricted to a single sentence, making them easier to analyze. As open source alternatives to Twitter work out the details of how to unify distributed services, I keep my fingers crossed that some high-octane APIs will come out to bring hundreds of thousands of one-liners into the analyst’s workshop.

When we talk about brand measurement, we take a brand-centric view of all the information available to us; our goal is to extract information about how the brand is discussed. As mentioned earlier, the most common measurement is looking at the volume of content mentioning our brand over time. It’s a pretty straight-forward metric showing us how popular we are, more or less. When correlated with a time line of product releases and marketing campaigns, we get a larger picture of the impact of those marketing efforts. Going further, we can use demographic data attached to the UGC to measure impact in specific regions and markets, helping to fine tune market spend or campaign strategies. The next step is to analyze our brand’s discussion volume with regards to that of its competitors to establish the brand’s share of voice, where we begin to get a sense of a brand’s overall health.

RC Cola could use a megaphone

Sentiment analysis is another valuable brand health indicator. When your customer service can’t determine why support call volume for your new product has doubled within days of its launch, a sentiment analysis of your support forums or helpdesk email base may be just the key. Essentially, generating this metric involves searching for UGC about your product or brand and looking for sentimental association like positive or negative. I won’t get too deep into the mechanics, the idea is to build a dictionary of “positive” terms like “rules”, “rocks”, etc and a dictionary of “negative” terms, and computing the lingual “distance” between your brand and sentiment terms. If your brand is mentioned in a paragraph chock-a-block full of negative words, chances are you have found a negative post! Looking through these negative posts with another search term like your new product, and then peeking at which of the results have engagement, then it is highly likely you have a problem with the product. You can see not only how “popular” your brand is, but how positive the discussion volume is over time. If before we could only tell if more or less people were talking about us, now we can see what percentage of all discussions is positive, and how that changes over time with respect to marketing events. Brilliant! A rolling average of this sentiment is another indicator of overall brand health (and a prime candidate for that dashboard you’re wire-framing…)

Unfortunately most of the tools to perform this level of work are not free; there is a high cost associated with warehousing and processing all of this content. I’ve worked with several of the key players in the past month to find a best fit for our business needs, but I’ll stop short of making recommendations here. If you find a vendor that interests you, ask for a demo and evaluation, and let me know how it goes.

So what’s the point of all this? Brand measurement gives us a 30,000 foot view of our brand health, with plenty of opportunity for deep dives. By evaluating your brand’s health over time, you can measure the impact of marketing initiatives and product releases. Furthermore, by splicing and dicing your UGC data you can evaluate the impact of campaigns region by region. Since people write about everything, not just what they see online, you can gauge reception of offline marketing efforts as well and what an impression is left on today’s authors – just about everyone. Engaging a client often means taking on stewardship of their image through branding and merchandising, and by taking a step back and looking at the distilled insights of hundreds of thousands of contributors, you can demonstrate to your client that people are watching, listening, and talking.

Did You Know?

Further Reading

  • “General Architecture for Text Engineering”
  • http://gate.ac.uk/
  • An open source platform for language processing and information extraction
Creative Commons License
The Metric of the Month September: Brand Measurement by Molecular Voices, unless otherwise expressly stated, is licensed under a Creative Commons Attribution 3.0 Unported License.

posted to: Data & Analytics No Comments »

Add a comment
Similar Posts:
Technorati Profile

Categories

Browse posts by month

Browse by author

Subscribing Comment Policy

We're always looking for rockstars

Come take a look at careers with Molecular