Note: This is a continuing excerpt from what I teach my social media marketing classes on introduction to metrics and measurement…
Influence Scores: Interactivity, Performance or Nothing?
At the time of this writing, the growing influence of influence measures has generated considerable controversy. There are stories of otherwise qualified candidates allegedly turned down for jobs because their Klout™ or PeerIndex scores were less than 90th percentile level. Mark Schaefer (2012, p. 1) begins his book on influence measures with a story of a marketing expert who was turned down for a job because of a Klout score of 45. After a “tweeting rampage” to game the system, his Klout score reached 70 and he started received unsolicited job inquiries. Some applicants with high Klout scores have begun including the scores on resumes. This raises a fairness issue: Should people be evaluated for jobs and special marketing offers based on an ambiguous score generated by a secret algorithm?
There have also been outcries when Klout adjusts its algorithm. In October 2011, a change in the Klout algorithm resulted in many participants with high scores seeing the measure drop ten to twenty points literally overnight. This caused an online uproar and a call for people to opt out of Klout. However there seems to have been little long-term change except that some organizations began to follow at least one additional influence measure besides Klout.
A big issue is: What is it that these services actually measure? What is really happening when someone “Likes” or “retweets” someone else? Should anyone have confidence in secret algorithms measuring occurrences that may or may not signal influence? For example, it seems hard to believe that Justin Bieber had more clout or influence than anyone else in the world, including President Obama; yet until Klout made a revision Justin was the only 100 by Klout’s measurement. Defenders of Klout are quick to point out that Klout and the other influence measures are only measuring online influence. But that doesn’t explain why some celebrities like Oprah Winfrey have high scores despite following few people and being relatively inactive. Clearly there is a celebrity effect from offline activities.
Ferenc Huszár, data scientist at PeerIndex, pointed out to me that influence is the type of latent social measure that generally requires multiple items and weightings driven by the data for measurement. Ferenc compared influence measures to IQ measures; IQ measures are similarly attacked for imprecision and imprecise definition, but have proven to be predictive. To further the comparison he noted that the top IQ scorer, Christopher Michael Langan, a horse farmer and former nightclub bouncer, seems to be the same sort of anomaly for IQ scoring as Justin Bieber is for influence measurement.[i]
A basic issue is what kind of “influence” these services measure. Klout, PeerIndex, or Kred influence seems to be a very weak sort of “influence.” Getting someone to click a link or like, retweet, or favorite a message is a long way from getting someone to buy something or contribute to a cause. One SMM expert even advises people wanting to be more effective on social media to avoid asking followers to actually do anything[ii] (Schaefer 2012, p. 121). An example: I posted a very favorable video review of the Tao of Twitter by Mark Schaefer. In order to measure the impact of the recommendation, I sent out five tweets to the video which had a private Amazon link on it and another five tweets simply recommending the book with the Amazon link in the tweet. With high influence scores and 50,000 Twitter followers, I was disappointed with only eight visits to the Amazon site and two purchases of the book resulting from the campaign. Of course some may have bought it another time or gone to another online service but eight visits and two purchases did not make me feel very influential! According to results cited in ROI (Schaefer 2012, p. 119–120) these results were not bad relatively: an author of a different social media book arranged for recommendations and tweets from a social media superstar with 1.2 million followers and then a services of influential SM experts with an aggregate of 1.5 million followers and as a combined result sold only one book. Influence to persuade others to follow, retweet, like or favorite may not directly lead to sales or donations.
Schaefer (2102) defines the online influence measured by these services as the ability to create content that is moved through the network and creates reactions. When Justin Bieber tweets, his multitude of followers pay attention and react to his content. This is very specific type of influence, but is still important. As noted in earlier chapters, there are many success stories in social media marketing. For example ticketing site Everbrite.com notes that each Facebook share of an event is worth $2.52 in average sales, a LinkedIn share $.90, and a Tweet $.43 (Rowan and Cheshire 2012). Nicolas Christakis, the author of the book that only had one sale through the high-influence/high-follower promotions, found greater success with a social media promoter who had a smaller, more focused following, leading him to the conclusion that “a small group of attentive follower may be more easily influenced than millions following a famous name” (Rowan and Cheshire 2012). Klout, PeerIndex, Kred, and likely other influence-measuring firms are studying and reporting influence in niches, such as topics, interests, and hobbies.
In an interview posted on YouTube, Azeem Azhar, founder and CEO of PeerIndex, said that marketers should focus on influencers in the “mighty middle” of PeerIndex scores, 35 to 65 (Remember from Table 5.3 that although 35 to 65 is the middle third of PeerIndex scores, which run from 1–100, the distribution of scores is skewed toward lower scores, so SM participants scoring 35–65 are actually in the 75th–98th percentile of all Twitter users). Ries (2012, p. 15) cites Kami Huyse, CEO of Zoetica Media, who believes that those with the very highest scores may be jaded and prefers to target influencers in the “magic middle…who are more serious, consistent, and eager to engage.”[iii] Morgan Brown (2012) compared two marketing campaigns by his firm and concluded that it is better to market to the Klout masses than the very top influencers, suggesting that the very top influencers get too much attention already.[iv] These comments suggest that when an organization attempts an influencer campaign, it should experiment with different levels of influence levels.
Example: The Author’s Influence Measures (August 2012)
I have been active on social media, especially Twitter and a blog, for several years. As noted elsewhere my online community helped me design my social media marketing courses. In Table 5.4 my influence ratings on five of the leading influence measuring services is compared to three social media personalities who I consider my “mentors” or aspirant group in social media. All four had TweetGrader scores of “100” higher than 99.5 percent of Twitter participants according to that algorithm, while the four ranged from the 93rd to 99th percentiles in the Twitalyzer influence measure. The four have many followers and are active users which, as noted, are rewarded in those measures. Recalling from Table 5.4 that a Klout score of 53, PeerIndex score of 45, and Kred score of 670 placed a participant in the 90th percentile for the respective measures, all four participants are at or above the 90th percentile in those rating schemes.
Table 5.4 Comparative Influence Scores—Author vs. three “mentors”
|The Author||70||72||805||100 (2nd)||97|
There are divergences in relative scores for different measuring services. Even though each service is attempting to measure social media influence primarily by analyzing interactions and communities on Twitter and Facebook, the relative scores are not consistent. Mentor 3 has a Klout Score that puts him below the 95th percentile of participants, while TweetGrader indicates 99th percentile. Similarly there are discrepancies in ranking the four SM users: Mentor1, who has written two popular books about social media and has an award-winning blog, has a higher Klout score than me, while I have a slightly higher PeerIndex score; among the four compared in the table, I was third on Klout, Kred, and Twitalyzer, but second on TweetGrader and first by the PeerIndex measure. Another free site which only uses Twitter data, TweetLevel™, ranked the four participants in the same order as Klout and Kred, but the top three all had high scores in the 80s. This probably reflects the fact that Twitter is the dominant SM platform for these SM participants.
Readers are encouraged to check out their own social media influence measures on the services discussed here and some of the others listed in Appendix B of Schaefer (2012), as well other new services that have been established by the time this is being read. For personal scores it is useful to compare scores with a referent group of people of similar interests or an aspirant set of social media participants. As shown in Table 5.4, I gauge my scores against those of three prominent social media marketing people. For organizations it is logical to look at ranking compared to competitors as well as “best practice” organizations in social media.
Marketing To and Through Those with “Influence”
As noted in Chapter 6, an organization employing a word-of-mouth marketing strategy may seek to leverage individuals with influence. Companies can provide product or services to individuals whose Klout, PeerIndex, Kred, or other influence score indicates an attractive level of influence. Klout and PeerIndex have public rewards systems under which individuals with a given score level and area of influence can claim free benefits. The services also provide private outreach services for companies who seek to privately contact influencers who are influential on given topics or interests for special offers or gifts.
Influencers receiving free products and services have an ethical and legal obligation to disclose the benefits when writing, blogging, or even tweeting about them. Organizations running influencer campaigns should make that responsibility clear to their targetted influencers. Of course, typical standard disclosure language would not easily fit Twitter’s 140-character limit for an entire message!
Contextual Influence Measures
The influence measures discussed so far are widely known since they are open to the public and in most cases depend on individuals registering, so they can track multiple social media sites for a participant. Ries (2012, p. 43) differentiates between this previous set of influence measures whom she refers to as “personal influence measurement tools” and a separate group of measures that are discussed in this section which she calls “contextual influence measurement tools.” These contextual influence measurement tools will not generate as much buzz in social media sites as the public measures, but may be of keen interest to marketers.
For contextual analysis a powerful search engine that allows researchers to focus on a complex topic or idea is a starting point of the analysis. Ries (2012, p. 44) mentions three firms that have developed such search capability along with a scheme to measure participants contribution to the conversation: MBlast™, TRAACKR™, and SpotInfluence™. Using these search tools an organization can identify participants with influence within specific topics and interests the firm identifies. These tools may also track conversations and actions of these key subject-influencers in real time. These contextual tools can help an organization track and guide key online conversations and key participants while the conversation is occurring.
Appinions™ (http://www.appinions.com) aims to revolutionize influence measurement by going several steps further in evaluating the online (and offline) world. Like the other contextual analysis tools, Appinions effectively integrates online and offline accomplishment by including Internet references, such as newspaper or other articles or citations, into their online analysis. The focus of Appinions, though, is on opinions and opinion-shapers. Based on a decade of research at Cornell University on semantic patterns of opinions and sentiments in text, Appinions scores authors and publishers (Ries 2012, p. 52) on opinion influence within a topic based on:
1. Preference—Who is picking up opinions?
2. Imitation—Are other people duplicating those opinions?
3. Trendsetters—Topic or discussion originators.
The ability to monitor opinion-makers in an organization’s market or key topics in real time has obvious implications for WOM marketing. Organizations can actively monitor the conversations and decide when to join in. Another interesting capability of Appinions research is to note what other topics are of interest to those key opinion shapers. Observed joint interest could shape marketing approaches.
In a discussion of current trends in influence measurement, Mark Schaefer, author of ROI, stated that Appinions will have a major impact on influence measurement[v]. I believe Mark’s statement could be an understatement: If the semantic analysis of opinions proves accurate, this form of influence measurement, monitoring, listening, and reacting should have a huge impact on marketing in the years ahead.
Summary of Influence Measuring
Influence measurement is rapidly evolving. There will certainly be new influence measurement services and new approaches to measurement when you read this chapter, and still other approaches six months hence. Two interesting ideas to take away from this discussion are to explore (1) marketing to the “magic middle” influencers as measured by the personal influence measures and (2) the use of contextual influence measurement tools to impact the online conversations of interest real time. These measures are controversial and raise privacy concerns but they seem likely to be employed even more by organizations going forward. According to Ries (2012, p. 5) the four major uses organizations make of influence measures today are to:
1. Prioritize customer service by influence of the customer.
2. Provide promotions and product samples for influencers.
3. Court influencers with ongoing public relations efforts to get the organization’s message out.
4. Manage the brand reputation and long-term plan.
There is some controversy attached to these uses of influence. The ethical and legal issue of promotions and samples for influencers and the need for disclose was already discussed. There is also a fairness or discrimination issue from special offers and gifts to those with high influence scores. How will customers react when they realize that they are pushed to the end of the queue for customer service because they don’t Tweet very often? Organizations must plan to address the fairness issue as they make such use of influence scores and discover new applications.
This is part #4 of an excerpt about “Metrics” from an early draft of a text for teaching Social Media Marketing. Please do not copy without the approval of Flatworld Knowledge and Gary Schirr. I welcome thoughts on omissions, additions and corrections!!!
[ii] Schaefer, Mark W. (2012) Return on Influence: The Revolutionary Power of Klout, Social Scoring, and Influence Marketing, New York, McGraw-Hill, 215 pages.
[iii] Ries, Tonia (2012) “Guide to Influence Measurement Tools” research report from Realtime Reports, March 2012.
[v] Schaefer, Mark W. (2012) Return on Influence: The Revolutionary Power of Klout, Social Scoring, and Influence Marketing, New York, McGraw-Hill, 215 pages.