When I discuss social media marketing I sometimes call it digital word-of-mouth. It’s apt. Digital word-of-mouth could someday become as powerful at growing brands as its off-line namesake. But hasn’t the Digital Age promised us the ability to accurately measure this type of brand influence? Now that word-of-mouth is in moving to “ones and zeroes,” can’t we quantify these conversations — perhaps even predict their future impact?
In a word: No. But that’s quickly changing, due in part to the work of Kevin Hillstrom and described in his new book Hashtag Analytics. One finding from Kevin’s scrupulously scientific approach to Twitter analytics is this less-than-scientific tidbit: Communities rely on love to grow and thrive.
What he means by “love” is something more akin to The Golden Rule than pledges of affection. Or, to quote his post on an analysis of the hashtag community surrounding the Martha Stewart brand: “If a community ‘loves’ those who participate in the community, especially newbies, the community thrives”
Considering his background in catalog analytics, it’s no surprise Hillstrom handles the new challenge of measuring social media with familiar tools. He analyzes online interactions the way he’d parse a customer list, applying mathematical models as a way to assign people digital profiles — similar to the methodology behind list segmentation systems like PRIZM by Claritas.
No, it’s no surprise he’d try this approach. What is surprising is its effectiveness.
Not that the transfer is easy. It’s one thing to look at a database of purchase behavior and demographic attributes. It’s quite another to look at lists of user “amplifications,” replies and re-tweets. Kevin just makes it look easy. He’s a master, and as I read his book I was in awe at how he tackled the challenge. I’m equally impressed with his conclusions, and the implied promise of future insights that this system holds for web marketing wonks like myself.
(In other words, I’m wowed by this book and its author. It’s therefore in goodhearted fun that I mashed up Mike Myers’ image with Kevin’s. No offense intented, bro. If nothing else, consider it a visual warning never to try updating your personal grooming style by rocking Joaquin Phoenix. Big mistake.)
When Twitter Becomes An Online Forum
Before I share a few of his findings, I should touch on methodology. But just a touch. You’ll always be able to learn more by buying his book, or by visiting his blog, or by following my own posts about my own forays into Hashtag Analytics, starting with the one I posted yesterday.
To begin, a word about why analyzing conversations around hashtags is important. The advantages of studying Twitter participants are clear:
- Twitter is hugely popular, providing a ton of data
- Twitter makes downloading this data easy and inexpensive
- Twitter has shown itself to be a channel capable of driving business for a brand
But there’s a problem: Unless a tweet contains a hashtag, it’s hard to measure sustained, multi-participant conversations. Hillstrom focused on tweets that all contained the popular hashtag #blogchat. This has the advantage of being a weekly conversation — a crucial requirement. For this analysis you need to measure at least a few recent weeks of behavior. It’s only when you’ve gathered this much data that you can begin attributing profiles to participants and predicting their future group status and level of influence.
Kevin relies on the hashtag’s ability to turn Twitter conversations into a loose online forum. You probably know another, more popular use of hashtags: Event-based topics, such as #CES or #SXSW. The jury is still out on what can be quantitatively surmised from this type of conversation, but the qualitative value of these event-based tweets is irrefutable.
Here is a sample of Tweets from May of last year, surrounding the hashtag #ungeeked. I captured these two tweets (plus many more, which can be seen when you click on the image) during my presentation at an Ungeeked Elite event. Jason, Sally Hogshead and Chuck Frey were three people in attendance.
I downloaded a batch of tweets from this event so I could study the effectiveness of my presentation. I had worked on the presentation with my employer at the time, HarQen, to explain what voice asset management is and why it matters.
A Useful Online Focus Group
That morning was the first time I had ever given the talk. When I got back to my computer, I made this screen capture and highlighted in yellow every comment with mention of me or what I was presenting. It became a useful online focus group.
If you look at the full thread that I captured you’ll see that when I played for the audience a voice excerpt of Martin Luther King Jr.’s most famous speech, it had the desired effect. His “I have a dream” pronouncement dramatized the impact of “original voice.” Another slide, showing Ben Franklin flying a kite in a rainstorm, also had some impact. One attendee tweeted that original voice has the power of electricity — but is only truly useful when captured and managed. This was exactly the take-away I was aiming for!
These discoveries were helpful, and one related finding from the tweets also jumped out at me: Chuck Frey is a terrific and vocal advocate. I, in fact, met many extraordinary people for the first time that morning, including Chuck, Jason and Sally. But here’s an important question: Is Chuck a more influential advocate for the brand than, say, Jason? Or Sally? For that I’d need quantitative information, derived over time.
A Quantitative Analysis of Hashtag Participation
He applies a process called logistic regression to a handful of data points — markers of weekly participant behavior. This distributes participants into one of eight digital profiles, of Hillstrom’s invention.
The graphic to the right shows a predictive model he created. It’s applied to the various profiles, listed in green. By counting the number of participants in each digital “bucket” and applying predictions based on past behavior of those in the bucket, he can forecast the longterm viability of the group.
A key conclusion from his research is that without acquiring and nurturing new participants, a hashtag community will soon shrink in both size and reach. Another is that in order to fully understand and benefit from these communities we need to look at them as full-blown ecosystems.
The Hashtag Ecosystem
I’ll leave you with this excerpt from the book:
Hashtag Analytics is designed to illustrate that there are different, non-traditional ways to evaluate how a social media ecosystem evolves and changes. Maybe the most important takeaway from Hashtag Analytics Is the concept of “ecosystem”.
In an ecosystem changes that happen today yield outcomes that may surprise one a few months from now. Models are created to explain what might happen in the future, and by using the models, individuals are able to mitigate potentially negative outcomes by making subtle changes today.
The same process can be applied to social media ecosystems. Too often, we focus on Influencers … we try to Identify the Influencer, then we market to the Influencer in an effort to achieve our own objectives. Our analysis suggests that manipulation may or may not work. Our analysis suggests that kindness always works.
What we learned in Hashtag Analytics is that “everybody” is an influencer in some way. This is an enormously liberating finding. The new participant who re-tweets content from an influencer can one day become an influencer if the influencer simply offers kindness in return for the retweet! We also learned that Influencers don’t necessarily maintain Influencer status over time.
I found Kevin’s work particularly exciting since I’ve been searching for some time for a reliable way to identify influencers through database analysis. He’s let me know that, at least by by some measures, these elusive souls do indeed exist.
The bigger news is that “Influential status” is ephemoral — here today and often gone tomorrow. Which leads to the biggest news of all, which is also a reassurance:
At least in Twitter, Influencers are made, not born.
Special thanks go out to Michael Czerwinski, data analytics extraordinaire, and his generous employer, BVK. You have no idea how helpful you were, Mike, in getting me to fully understand this complex analytic methodology.
- The science of the hashtag – Twitter Media [del.icio.us] (media.twitter.com)
- #HashtagsTopTwitterTrends (marketingpilgrim.com)
- Twitter hashtags for journalists (billbennett.co.nz)
- 5 Tools to Make the Most of Twitter Hashtags (oneforty.com)