PeerIndex, Influence and the “Magic Middle”: Insights from CEO Azeem Azhar

Tweet

Influence is a term that is tossed around a lot in social media circles and is done so quite lightly. People pursue and grow their influence, brands want to leverage influencers and a new breed of tools strives to define influence and identify influencers. But influence has always been a slippery concept, defying easy definitions and metrics. To learn more, I interviewed Azeem Azhar, founder and CEO of PeerIndex (PI), a technology company that applies a scientific approach to the problem of influence. (If you’re not familiar with PI, you can check out my public PeerIndex profile here.)

Ironically, Azeem starts by tossing the term “influence” into the trashcan. “In a broad generic sense we know what we mean by influence, but in a specific sense it truly breaks down.” Instead of the term “influencer,” he prefers to use “peer opinion leader.” Azeem notes that opinions matter, but only the opinions from people others find trustworthy within specific domains of interest. “Expertise and influence rarely cross boundaries,” he says, although Azeem notes there are exceptions to the rule--polymaths such as Paul Kedrosky who manage to earn high topic PI scores across a range of very diverse topics.

While there are those who seek to game influence and amass tens of thousands of followers in automated ways, none of that delivers true influence or opinion leadership. “Simplistic counting mechanisms that reward the inane and popular do not reflect the discrimination with which most people and businesses make decisions,” Azeem notes.

PeerIndex’s solution to this problem is to separate the social media signals that are “cheap” from the ones that are “expensive.” “Following someone on Twitter is cheap. Retweeting a tweet to all your followers is expensive because you impose a cost on them. Getting a smart, busy, respected person to retweet a tweet of yours is even harder and more expensive.”

I love Azeem’s use of the terms “cheap” and “expensive” in this context, because it reminds us that the essentials of relationship building are not altered in social media. A “cheap” automated DM that thanks users is not the same as an “expensive” personalized DM that conveys specific curiosity in a new follower. Building relationships is and should be difficult, and PI's approach reinforces the precept that there are no “cheap” shortcuts.

PeerIndex evaluates opinion leadership in ways that might seem obvious and in other ways that are not. Of course, it helps to get many people commenting and replying and to be added to Twitter lists, but much like Google’s search engine algorithm, quality matters—not every retweet from every Twitterer is equal.

Another key to building your opinion leadership is to narrow your topics. As noted, it is difficult to be an opinion leader across multiple topics. “Your activity density in that topic matters--it is harder to be a Peer Opinion Leader across multiple topics,” notes Azeem. In other words, if your goal is to build influence, focus on a topic and the people who matter in that topic domain.

PeerIndex’s model goes even deeper to find those with real influence, not just Twitterers with many connections and interactions. The company’s system values more highly people who say something other people will come to believe or agree with and finds the people who are viewed as knowledgeable by their peers. Here again, the concepts of “cheap” and “expensive” resonate—it’s “cheap” to auto follow many but “expensive” to earn trust and be seen as knowledgeable by others.

Azeem brings a refreshing humility to his discussion of PeerIndex and admits the technology is still early. For example, automated bots can earn high influence scores merely by systematically retweeting others. Azeem notes this is a problem, “but we do a better job of identifying bots than anyone else I have seen,” and he points to PI’s code which performs “bot identification and punishment.” Azeem also invites people to bring any problems they may see in PI to his attention: “Tell us about them - it’s the best way we can improve.”

Among the challenges that PeerIndex is working on is to find ways to assess “real world authority.” Today’s tools overemphasize social at the expense of other channels, a problem Azeem calls the “Shirky problem” after Clay Shirky, author and professor on digital culture. “Clay rarely tweets, but he builds amazing theses and publishes books that literally create industry sectors for the next ten years. Should he have a sucky PI score? No way! The data to take into account other, often more important, channels exists; we just have to get around to indexing it.”

In my discussion with Azeem, we turned repeatedly to how influence cannot be confined into a single score because influence is created within specific topic domains. He said to me, “I would trust you on the subject of peer influence marketing, but I wouldn’t trust you on the topic of neuro-oncology. Sorry!” (No apology necessary!) To overcome this problem and appropriately reflect opinion leadership within topic domains, Azeem says, “We already measure topic PI across more than 3000 topics, and that number is growing. We can also estimate it on any arbitrary topic–even one we haven’t seen before.”

Azeem admits the current PeerIndex site does not do an adequate job of reflecting the number of topic domains monitored or how specific opinion leadership accrues to individuals. Currently, the site provides an individual’s scores in specific topics based on frequency of discussion and resonance, but changes are coming to the PeerIndex interface.

As you might imagine, PeerIndex’s business model is based on assisting brands to find the right opinion leaders within their product categories. I found it interesting that when PI works with brands, they neither focus on overall PI scores nor ignore them completely either. “As we work with brands, we focus on topic PI scores as a first pass and then cut by overall PI score as a final sense check, not the other way round. The reason is that someone with a very high topic PI of 70 in Computer Security may be a specialist with a more moderate overall PI, such as 40. If we cut first on the overall PI score, we’d lose this key expert.”

As a former Forrester analyst, one of my great concerns with the focus on influence is that it can miss the point—the key to success isn’t to find the few with massive influence but the many who represent mass influence. Azeem agrees, which is why PeerIndex steers brands toward the “Magic Middle,” a term coined by Charlie Osmond, CEO of FreshNetworks.

Says Azeem, “The obvious top-twenty bloggers in a given topic resemble media outlets--inundated in tips and leads and with audiences so wide they are restricted in who they can engage with personally.” Rather than focus on those, PI “directs brands to the tier below the most obvious names--towards people who show strong affinity and influence and who have engaged networks in that topic. Engage with enough of these and you’ll unleash ‘mass influence.’”

I appreciate Azeem taking the time to share insights on influence—excuse me, I mean peer opinion leadership. I hope we’ll continue the dialog, because the topics of influence and influence measurement will remain essential ones for social media professionals. Also of note, I posed many of the same questions to Joe Fernandez, CEO of Klout, and I will soon share his perspectives on these fascinating topics.