Influence metrics are growing up. According to Adage, Razorfish is about to introduce “the SIM score, which stands for social influence marketing.” The new score is covered by Abbey Klaassen in What’s Your Brand’s Social Score?. Social media marketers have long sought (relatively speaking) a standard measure of social media ROI. I don’t know that this SIM score is it. Let’s have a quick look.
The SIM score is apparently a social media version of the Net Promoter mode. Adage puts it like this: “How likely is it that you would recommend our company to a friend or colleague?” (To get the score, subtract the “highly likelies,” or promoters, from the “unlikelies,” or detractors.)
And according to the article, the Razorfish SIM score seeks to capture the strength of social media as a medium for organically surfacing recommendations. Quoted in Adage, Shiv Singh, VP-global social media lead at Razorfish also recognizes what many social media marketers have long known: the conversation is out there (like it or not):
“Any mention of a brand, as long as it’s not negative, serves a brand-awareness purpose on the web because once it’s there, it stays there.”
The score comprises of a net measure of sentiment as captured in social media mentions. Again, from Adage:
Razorfish worked with TNS/Cymfony to capture social media content and the net sentiment of a brand: the positive and neutral conversations minus negative ones, divided by total conversations about the brand.
As most folks in the social media analytics space know, as I’m sure is familiar at Razorfish and Cymfony, social media do not make it easy to obtain sentiment and semantic metrics. There are several reasons for this, some of which are specific to the medium and some of which are behavioral:
- The 140 character limit on tweets puts significant pressure on context. Context is often left out of tweets where it can be assumed by the reader. Crawlers of course have difficulty recognizing the implicit references and context of tweets, so some if not many tweets are simply missed.
- Expressions in twitter are colloquial, if not also abbreviated, shortened, and clipped. Again, expressions often don’t explicitly reference topics and content (brands, industries, products included).
- People make recommendations in twitter shaped in part by who they follow and who’s following them. One can’t remove the act of recommending from the audience the recommendation is made to or in front of. People will often make recommendations not only to share their feelings about a product/brand, but also to publicly identify with that product or brand. References made in social media like twitter reflect on the twitterer. Tweets can show a person identifying with something or someone, attracting the attention of someone, showing gratitude to someone, showing affection for someone, and so on.
- In public social media like twitter, a recommendation may also serve the purpose of building a person’s credibility or reputation as an expert, influencer, trusted authority, and so on. Consider the difference in recommendations made by @Scobleizer and @GuyKawasaki and @jowyang. Each of these heavy users and influencers has his own way of watching for, filtering, selecting and then tweeting or retweeting. @guykawasaki has influence as a newswire, more than @jowyang, whose influence rests more on his personal and professional authority.
- Recommendations can come as answers to solicited or unsolicited requests for help or information.
- Recommendations may be made as a means of introduction on twitter — sometimes to get followed back, to get noticed, or simply to be helpful.
These are some of the ways in which recommendations might be distinguished in social media from recommendations made face to face or by other means (as measured by the Net Promoter method). In conversational media, the act of communicating is difficult to separate from the information communicated. Recommendations and the act of recommending can be measured differently, and have different meanings: the intention behind the act, the message or information provided, motives inferred by recipient to the act. (Person A tells person B to go see Harry Potter, hoping to get the question “Oh you saw it?! Was it good?” and instead Person B ignores Person A, wondering to herself “Why is A telling me to see Harry Potter? Don’t they know it’s not my kind of thing?”)
There are also ways in which recommendations may elude attempts to simplify sentiment captured from social media. There are also ways in which social media provide information about a brand’s “influence” that are not in what people say but in how they say it, to whom, and what happens when they do. Some of this is what we can call “envelope” information (tweet addressing: to whom, for whom, citing whom, or @name, @reply, RT).
The rest of it is in the distribution: reach, volume, velocity, acceleration. These are aspects of flow and are among the attributes captured by some social media analytics tools. In marketing speak:
- How quickly is brand retweeted?
- Who retweets?
- How deep down a social graph does the retweeting go?
- How far across a social network does the retweeting go?
- and so on
I know that these aspects of social media activity are difficult to track and measure. But it would be great if there were an industry-wide effort to define and codify some of the attributes of social networks, relationship-based communications, and common types of expression in order to better represent conversational activity in social media. The results would not only paint a more accurate picture of brand presence in social media, but would also match the real social mechanics and dynamcis of online conversations. It may take a while for algorithms and tools to emerge for this. In the meantime, I would supplement SIM scoring with insight from a good community manager.
Shiv Singh
July 15, 2009 at 7:09 pmAdrian,
Thank you for your thoughtful comments on Fluent and the SIM Score. It is certainly not perfect but it is a starting point. One difference to the Net Promoter score – while that it is focused on the likelihood to recommend, with the SIM Score we think of it in terms of the brand health or the pulse or the blood pressure of the brand in the social web. It is less tied to recommendations because as you highlighted that's a lot trickier.
Shiv
gravity7
July 17, 2009 at 10:27 amShiv,
Thanks for your comment. I have to confess that I had not finished the report when I wrote this and was responding in part to its reception. The report is superbly written and offers a very clear rendition of the state of affairs.
Social context and user-centered social browsing and engagement is clearly the direction we're headed in, as much on ever-better profile-based social media like Facebook as well as interconnected and aggregated conversation-based social media (twitter, friendfeed, etc).
I like the challenge of measuring influence and of categorizing influeners, as is done by folks like klout.net. Social media analytics is slowly developing categories that are more behavioral and even psychological than the market segmentation or user adoption categories they started from. Which is smart — we can now get more granular about users than Forrester's initial social technographics.
I think we all recognize that some twitterers, for example, are soap boxers, some are announcers, some are em-cees, some are critics, some pundits, and so on… that in retweeting behavior there is retweeting for personal gain, retweeting as a favor, retweeting because the content is good, retweeting because the event is socially ranked…
In short, the social behaviors we impute to users, influencers and their audiences in particular, are motivated in part by the user's relation to the medium. (Which I like to simplify into Self-oriented, Other-oriented, and Relational.) That this can create distortions, bias, and many of the other “social effects” that characterize social media sociality (see shirky, danah boyd, et al…).
Just as the insights provided by focus groups are corrupted to some degree by interpersonal, group-think, and social group dynamics; just as surveys provide biased results because how we talk about ourselves and how we describe what we do (self-reporting) is not transparently how we are motivated; similarly, online interactions are not a direct reflection of user interests but are often a reflection also of mediated interpersonal, social, and public perceptions, expectations, interpretations, and intentions…
So I think it would be nifty to get even more granular, as the data and metrics develop by which to capture and weight, with characteristics of social branding in conversational media. I've suggested a personas 2.0 approach, but I think we need conversational models too (differences between inviting, giving, quoting, requesting, recommending…).
I know this is more than is needed at this stage — but the marketing grail is still in relationships, not with one but with trusted friend circles. Social has been the grail since ancient tribes organized their economies, power structures, marriages etc around kinship and alliances….
cheers,
a
pamelarutledge
July 17, 2009 at 1:29 pmAdrian,
I like how you are teasing out the unseen behavioral implications of Twitter conversations. As added behavioral complications to measuring trust and influence, you might also include:
– how the communication style mutates as two or more people get to know one another (in the Twitter context) and adopt an increasingly familiar and shorthand style.
-the implied understanding in a continued conversation. Content indicating endorsement can be completely separated from endorsee as the exchange continues.
I am also interested in how this influences perceptions of interpersonal connection and the attendent impact on trust–more illusive and eminently less practical than brand monitoring!
Pam
gravity7
July 17, 2009 at 2:35 pmpamela,
love your comment!
On first point, yes, and add to the conversational familiarity (shorthand, tweeting vernaculars) social behaviors. For example a status-seeking retweeter who retweets a pundit less after getting to know him/her: that personal brand equity attracts more social tweeting behaviors when the celebrity/branded individual is inaccessible. And that once friends, two individuals would perceive some retweeting behaviors as disingenuous.
This social effect occurs, I think, because one can use other people, including their reputation and celebrity, as content. Tweeting others can make them as well as their tweets the content and reference of one's own tweets. This could certainly rub some people the wrong way.
Totally agree that the closer people are, the more likely they build a reference base of inside jokes, quotes, and other coded references… Again on twitter the ambiguity of intent raised by the social AND public audiences addressed can create social effects…
Danah Boyd calls these “invisible audiences” — the psychologized version of the invisible audience would augment the presence aspect with a perceptual relation: projected vs internalized audiences, perhaps?
a
RF Health Club Consulting
October 15, 2010 at 3:07 pmGreat information! I’ve been looking for something like this for a while now. Thanks!
privacy protection virus
December 21, 2011 at 9:56 amI really like your writing style. Thanks for this beautiful article!