In case anyone still needs convincing: influencers are a big deal. They have more online reach and credibility than the rest of us, and that makes them an attractive resource for brands exploring ways to maximize the impact of their digital marketing spend. In the noisy world of social media, these tech-savvy sages provide an efficient delivery vehicle for targeted messaging that can cut through the chaos.
So, naturally, “influencer marketing” has become a thing. Agencies are adding it to their list of capabilities, brand marketers are insisting that it be pulled into the mix, and technology vendors are trying to replicate the success of Klout, Kred and similar platforms that claim to be able to quantify an individual’s influence based on proprietary algorithms. (We’d argue that they’re not quite there.) And then there’s Google search data on the subject, which shows a significant year-over-year increase in our collective level of interest (see chart).
Despite the hype, the interwebs are frustratingly short on details about how the whole thing works. Go ahead, try looking it up. You’ll quickly discover that it involves getting “influencers” to help with brand marketing. Then you’ll realize that little else has been written.
That’s a problem. Consider this our tiny contribution to the solution.
We’ll flesh out the end-to-end process in later posts, but here we’re going to focus on that critical first step: influencer identification. While it isn’t the sexiest part of influencer marketing, for obvious reasons it needs to be done right. For less-obvious reasons – immaturity of the practice, social media’s snake oil salesmen, general lack of understanding, etc. – most of the influencer identification projects we’ve seen leave something to be desired, generally due to one of two subpar approaches:
- Google it. This can eventually get you what you’re looking for – provided you have a bit of experience and an average amount of common sense. Armed with only Google (or another search engine that somehow thinks it can compete in this space), it’s not terribly difficult to identify, say, a handful of influential people on the subject of food trucks in San Francisco. But that doesn’t mean it’s fast, which means it isn’t cheap. Nor is it comprehensive. Or fun – manually combing through the Internet sucks. Most egregiously, it ignores the ever-growing number of big data technologies that simplify and improve the process.
- Social influencer tools. The more elegant of the two basic solutions, influencer tools do the Google-ing for you and then run the results through proprietary algorithms. The algorithms then spit out a list of influencers based on your broadly defined search criteria. Some of these tools are undoubtedly better than others, and some are downright impressive. None of them, however, are flawless. Users are forced to sacrifice quality for the speediness and ease that comes with automation.
Needless to say, both methods have major disadvantages. Google searches can work, but they’re grossly inefficient. Social influencer tools are less time consuming (and much sexier), but they don’t produce great results. Marketers are left with two flawed approaches to the critical step in an increasingly important process.
Recognizing an opportunity, we pulled together some of our best thinking to develop a better solution for identifying influencers. We tested a number of methodologies and ultimately decided on a hybrid approach that works well.
We start by using lots of different tools – because, as mentioned, none of them are perfect – and then aggregating and cross-referencing their results. This weeds out the worst results in a couple of quick steps and typically highlights a few names that are likely to appear on a final list of influencers. It also allows us to better quantify influence, which is important because it not only provides ranking, but also tells us something about the magnitude of difference between these folks. Finally, we run a systematic, manual quality check to make sure the individuals on our lists are both influential and contextually relevant (Bill O’Reilly is highly influential among Fox News viewers, not those who prefer MSNBC.)
There’s nothing particularly novel about our approach. We actually borrowed it from the political world and super-smart statisticians like Nate Silver, who aggregates polling results and generally produces the most reliable election predictions on the planet. It works for us because, like political polls, all social media influencer tools are slightly biased or miscalibrated in one way or another.
Aggregating and sanity-checking the results minimizes the impact of the really bad data points and allows us to focus on the ones that truly matter. It’s not the cheapest or fastest approach, but it isn’t the most expensive or slowest, either. Most importantly, it consistently leads us to the right influencers and allows us to confidently move on to the fun part.
Could this approach work for you? Have you seen other innovative ways to identify influencers in the social media landscape? Let us know what you think in the comments.