Avoid Relying Solely On “Canned” Scores For Influencer Targeting -- But Don’t Write Them Off Either

Influencer marketing is on my mind these days. In addition to working on a report about how interactive marketers should collaborate with different resources to execute influencer marketing, I’m also speaking about the topic at Forrester’s Consumer Forum in Chicago later this month.

Talking with marketers, agencies, and service providers, everyone (yes, it’s been everyone) has voiced opinions about “canned,” algorithm-based influence scores available through providers like Klout and PeerIndex. Detractors say that black-box influence scores focus too much on reach and not enough on context or topics, and that influencer identification is too complex to boil down using an opaque calculation. For example, Charlie Sheen may look like a valuable influencer based on his high Klout score, but a marketer of diapers would probably prefer to tap a mommy blogger with a lower top-line score to advocate its brand.

The supporters’ rebuttal: why would you use these scores in a vacuum in the first place?  The score providers themselves dissuade marketers from looking only at an individual’s top-line number with no filters for topic or brand relevancy – and those filters are available. Count me in this camp.

So how should interactive marketers regard off-the-shelf influence scores?  Keep in mind that:

  • These scoring systems are evolving. Companies like Klout and PeerIndex are in their early stages and certainly do have their limitations. But their capabilities to mine scores for topic, category, and brand influence continue to improve.
  • Off-the-shelf influence scores can be helpful inputs. The scoring providers have made strong advancements in how they understand and define influence. The scores can serve as components of a larger identification process and also help marketers looking to scale their influencer programs. 
  • There is no “easy button.”  Influencer marketing is complicated and requires collaboration with different roles within your organization.  Don’t rely solely on off-the-shelf scores as a short-cut for identifying and segmenting influencers.  Instead, work closely with your customer intelligence, PR, and other colleagues to discover and target influencers.

What do you think about off-the-shelf influencer scores?  Do you think they add value to influencer marketing in their current state, or are they more hype than practical substance at this point?  Chime in with comments and let us know.


Hi Ari Thanks for the shout

Hi Ari

Thanks for the shout out - and you are right this is a really exciting area of development. Even last week we saw Jodee Rich and his crew roll out Kred, which is another take on this area of influence.

It's important to understand the scope of your project to figure out what you are doing:

A. The fewer, more detailed people you need to identify, the more manual work you will need to do. If your goal is to hit 5 or 10 people, then you'd better double-check exactly which outcome you are seeking and go manual. If you are looking for dozens, hundreds, thousands or millions of names, then you will need to use a statistically based platform like PeerIndex - manual analysis doesn't scale.

B. Run split tests using both PeerIndex and Klout and work out which data set performs better for your particular campaign. Split testing is very common in analytically-driven marketing. In these early days, it is not going to be clear whose particular take on influence or opinion-leadership is going to provide you with the more suitable identification. So instead, split your campaign in two and see who will perform better. (We already work with clients who wish to use PeerIndex data together with either their own models or a 3rd party vendor.)

C. Understand that just as influence is not something that varies day to day, like a yo-yo, so working with an influencer is not as simple as send a tweet with a message. Instead, plan on a pro-active campaign, identify the influencers early and work with them with clear defined messages.

Thanks for your insights

Thanks for your comments

Azeem -- thanks for your comments on the blog post. The notions of being able to scale and testing different influence data sets are great points. Thanks for contributing.


Hype more than practical substance

Hi Ari,

It is too early to rely on any influence ranking services out there, so far it is all hype than substance. Azeem makes a really good point by saying "understand that just as influence is not something that varies day to day, like a yo-yo, so working with an influencer is not as simple as send a tweet with a message."

Klout's model seems to be very flawed as they are just measuring levels of interaction and selling it as "clout". I have wrote two blog posts on this issue accompanied with a few examples that you might be interested in.