While social influence scoring services like Klout, Kred and PeerIndex can provide valuable data for marketers, it’s important that you understand their limitations as well as their strengths. (Note: This is part 2 of a 2-part series. See part 1 here.)
One challenge is the fact that you can’t compare scores in an apples-to-apples way when looking at different geographies. Professionals in Germany are more likely to use Xing rather than LinkedIn for their professional networking needs, putting them at a disadvantage as Xing isn’t currently used by any of the social capital scoring services.
The Chinese government blocks most western social media sites, leading to the rise of local champions Sina Weibo (similar to Twitter) and Renren (similar to Facebook). Keep local website usage differences in mind when working to identify influencers across geographies: to the best of my knowledge the scoring systems aren’t set up to compensate for these differences.
When I brought up the nuances of geography with PeerIndex’s CEO Azeem Azhar, he indicated that this hasn’t yet been an issue in the campaigns they’ve managed. Influencer marketing is still in its infancy so, in his view, Twitter signals have been more than sufficient. He acknowledged this will probably change as the industry matures.
Perks: Partnerships, And Bait To Keep Influencers In The Game?
Figure 2: PeerPerks campaign for Ford B-Max: I’ve Won!
Once influencers to target have been selected, the next step in the process is to get samples into the hands of influencers, so as to influence the influencers to talk about the product, service or brand.
This could be done directly, in theory, but each social capital scoring service prefers to serve as an intermediary, supplying targeted influencers with perks on behalf of marketers through the rewards programs Klout Perks, PeerPerks and Kred Rewards.
Kred says their rewards are available world-wide. Kred also offers a reporting service, notable in part for pricing transparency. Klout seems to be focusing their program on the US (I’ve asked for clarification of this point) while PeerPerks are available world-wide.
The benefits of the rewards programs to the scoring services are clear: the rewards serve an incentive to get individuals to hand over the keys to their Facebook, Google+, LinkedIn and other profiles, profiles which wouldn’t otherwise be available for analysis.
Yet the lure of these same perks will likely contaminate the scores as “influencers” fine-tune their social participation behavior with an eye to upping, or at least maintaining, their Klout, PeerIndex and/or Kred scores, a phenomena known as the observer or measurement effect.
This may be one of the reasons, together with score anxiety, that Klout no longer offers their badge as a widget although both PeerIndex and Kred do currently offer widgets.
When I raised concern of the observer effect with Kred, Kred’s CEO Andrew Grill minimized the concern:
While it is true that those users that sign up for Kred and connect additional networks such as Facebook allow a broader view of their influence to be calculated, experience shows that we can find those who have real influence by tightly-defined communities based on a market segment or location.
With Kred, and unlike other services, it is not always a single score that excites marketers. Those with high Kred in relevant communities can be even more influential and valuable to brands in a specific market segment than someone who has a high global Kred score.
PeerIndex’s CEO Azeem Azhar told me the effect is inescapable, just as the first vanity searches in Google were an incentive for people to actively curate their personal reputation on the web with an eye to the SEO impact.
Azhar stressed that genuine behaviors are hard to achieve. He implied I won’t be seeing Al Gore retweet me on the topic of climate change or Mitt Romney retweet me on taxes anytime soon. I suspect he’s right.
In our conversation, Azhar also noted strong demand for Topical PeerIndex (TPI) scores, where all profiles scored in up to 8 major categories and up to 5 niche categories specific to each profile. The data is also freely available, within limits, through PeerIndex’s API.
Endorsements Within Scoring System
Figure 3: Klout’s +K and “Add Influencer” in action (grazie, Davide!)
Rather than trying to minimize data contamination by the observer effect, active involvement by the measured in the measurement process is actually encouraged by the three scoring systems.
Klout allows those who participate in their system to give points to others by endorsing them as an influencer and/or giving them a +K as an influencer on a specific topic.
Kred allows a user to award another user with +Kred in a specific community; the person who offers the award also receives points for their good deed.
Doesn’t this create more potential for observer effect data contamination? PeerIndex is the current exception, they rely instead on the more natural ebb and flow of endorsement signals from the external data sources like Twitter.
This opens up a philosophical question: when marketing makes use of research and statistics, shouldn’t it be subject to the full rigor of the scientific method, i.e. methodology disclosure, peer review, etc? I would argue that if one is to base business decisions on data, then yes. We shouldn’t be using dubious data. Caveat emptor.
Score Reporting Period Length And Refresh Frequency
Klout uses Twitter, Facebook and Google+ engagement statistics from the last 90 days. Klout says the Klout Score is updated daily and that Influencers and topics update weekly. Yet it appears that it takes several days to process data from services like Facebook and Google+ once these accounts are added.
PeerIndex says they update scores multiple times a week and that the scores reflect 120 days of activity (source: CEO Azeem Azhar). Newly-added social accounts require a few days to process, although PeerIndex’s data is currently stuck in time as they roll out a new infrastructure which should be in place soon.
Kred says they analyze tweets from the last 1,000 days (almost 3 years). One issue with Klout’s approach (and potentially that of PeerIndex) is that a small amount of time off-line, such as a week in the countryside where the digital divide is all too present, will negatively impact an individual’s score.
Can A Social Capital Score Be Based On Just 13 Tweets?
Among the various complaints leveled at some of those in the business of measuring Internet traffic, e.g. Nielsen & comScore, is that their sample sizes have been small, perhaps much too small.
Data quality will indeed depend in part on the quality of sampling used by scoring services. Klout is known to source data, including the Twitter firehose, i.e. the complete stream of tweets, from GNIP and they say they analyze 2.7+ billion pieces of content and connections a day.
Kred tells me they also accesses the full Twitter firehose, with a data mine going back to 2008. PeerIndex tells me while they don’t have a relationship to access the entire Twitter firehose, they do have whitelisted API access which allows them to access over 1 billion status messages a week.
[Update September 5, 2012: In a follow-up email, PeerIndex clarified that the actual number they look at “varies according to a variety of business and operational reasons”.]
At the moment it appears that my own PeerIndex score was computed on a total of 13 tweets (Tweets analyzed in the Channels dashboard) but I was assured that, due to architecture issues, this isn’t really representative of what goes into my score. By September 2012 PeerIndex aims to have a more robust architecture in place which will allow greater and faster innovation. Reminds me of Google’s famous caffeine project.
Social Engagement Isn’t So Simple
In 2006 Jakob Nielsen stated that about 90% of people are lurkers — they read, they observe but they don’t contribute online. That number has undoubtedly decreased with the rise of Facebook which does everything in its power to get people to click as often as possible (assuming they actually see an update), but the general substance is still true: motivating someone to (intelligently) comment a post or to forward a link takes work.
Some engagement is more valuable than other engagement. A shared post/retweet is clearly worth more than a simple click on an endorsement button (like, +1, favorite), as is a positive comment. Yet an endorsement click is much more appreciated than a negative comment!
Engaging Influencers? Consider The Impact Of Those Left Out
Both Google and Microsoft have given away cellular phones at events they’ve organized, knowing full well that word of mouth discussion, hopefully positive, would ensue online. That Microsoft has had lots of difficulty getting traction in the phone market reminds us that word of mouth marketing can’t solve underlying problems like a product which might not be competitive for any of a series of reasons.
Use of influencers in social media marketing campaigns also needs to take into account the reaction of those who consider themselves to be influencers yet are not included in a campaign. Nobody likes to be left out, especially when their exclusion may be known to their peer group and they believe, rightly or wrongly, they should have been included.
Some of those with hurt feelings will use the soap boxes available to them to vent their frustration. A prominent Techcrunch blog post deriding the value of Klout notes the author was excluded from the Microsoft phone promotion.
In Italy, some excluded from a Trenitalia (Italian National Railways) open house hijacked the designated Twitter hashtag, #meetFS, with acerbic tweets, many of which appeared to have been driven by frustration at having been excluded rather than representing mature observations regarding Trenitalia’s admittedly inconsistent service or interaction with the event’s participants.
Viewing Scores, Without Giving Up Your Data
So they want my data and want to post on my profile, do they?
Each of the major “influence” scoring services wants your data, particularly data it can only access with consent. They generally won’t show you scores if you don’t give them the keys to your social media homes, i.e. login with credentials from Twitter, Facebook, or perhaps Google+. Ostensibly this procedure serves to save you from creating yet another account, yet by using social media network credentials, you must allow the scoring services to access data from your profile, and if you’re not attentive, they may well spam all your contacts with updates telling the world to use said scoring service!
Klout makes one undocumented exception to their login to view data policy: search engine crawlers have free reign. Search for the Klout profile in Google or Bing and view the cache version, or use a browser plugin to change (spoof) your user agent to Googlebot then open the user’s profile in Klout, no login required. The Klout score URLs are of the form http://klout.com/<account name> where account name is the user’s twitter account.
PeerIndex will display a user’s basic score using the URL http://www.peerindex.com/<account name>; no login is required.
Kred uses a similar URL structure, http://kred.com/<account name> although you have to dismiss a login nag overlay to view the data.
Privacy Considerations? Klout Missing Opt-Out Option
Some people, including noted communications professionals, don’t want to participate in the influence scoring business. Klout offers individuals the option to delete their profile. PeerIndex offers options to hide or delete a profile (a link magically appeared in the settings page as I researched this article). Kred only offer an option to hide a person’s profile – marketers should insure they don’t enter in contact with those disgruntled by the whole idea of the influence game. Both PeerIndex and Kred incredulously suggest that a person opt out of their services by changing twitter profile settings to private; deflecting responsibility to provide a solution to a problem they created.
Social Capital Scores May Not Pass Scientific Rigor. But Can They Be Useful Nonetheless?
As long as one understands the limitations of each scoring service, the scores might be useful as a rough signal, together with other measures, to identify influencers. Should they be used by themselves? Probably not, but I suppose it depends on what is at stake.
Do understand that the services will undervalue those who don’t register with their services (a specific consideration for HR professionals). It would certainly be, in my view, ill conceived to use influencer scores by themselves in hiring and for making similar decisions. Darrel Huff’s 1950s classic, How to lie with statistics, is recommended reading for all.
Klout, PeerIndex & Kred: Where The Data Might Come From
The following table lists the data sources used by Klout, PeerIndex and Kred to compute their social capital scores. In general, services only consider Twitter data unless an individual has registered with the scoring service.
Even then, data retrieved from closed social networks like Facebook may be incomplete, ignoring activity within groups and/or pages, for example. In addition to the social sites listed, Klout says that support for bit.ly, BranchOut, Disqus, livefyre, posterous and Yelp is “coming soon”.KloutPeerIndexKredFacebook■■■Twitter√√√Google+■XXLinkedIn■■XFoursquare■XXQuoraX■XWikipedia■XXYouTubeX¹XXInstagramX¹XXFlickrX¹XXLast.fmX¹XXWordPress.comX¹■XBloggerX¹■XSelf-hosted blogsXXXForumsXXXTumblrX¹■XScaleLogarithmic, 0-100Logarithmic, 0-100Normalized, 0-1000Number of profiles scored100 million192 million (source: CEO Azeem Azhar); at 50+ million, FAQ is out of date120 million (source: Kred CEO Andrew Grill)Number of active profiles, i.e. individuals registered with service?not publicnot publicReward availabilityUS?World-wideWorld-wideTopics for segmentation?2000/8000200Bloghttp://corp.klout.com/bloghttp://blog.peerindex.com/http://blog.kred.com/Data Coverage LegendData is publicly accessible√Data is available for users who explicitly give permission to scoring service. Data may still be incomplete.■Not currently considered¹Klout allows users to associate these profiles, but the data isn’t yet used.X
Comments