Klout Sucks and Why You Should Use It
The Twitter influence measurement service Klout was a huge topic of debate at this year’s SXSW conference. Its founder Joe Fernandez was on various panels touting how wonderful it is that the startup has found a way to automate the influence measurement process – something many feel is the “holy grail” of social media marketing. Mr. Fernandez also discussed “Klout Perks” – a service that allows brands to target influencers by sending them tangible items in hopes they will talk about them throughout their social circles leading to increased sales. Unfortunately, Klout is nothing but snake oil when it comes to determining influence. That said, it’s an immensiely valuable service that should be incorporated into your reporting mix. Confused? Let me explain.
No matter what Joe Fernandez says, there is currently no tool that can accurately automate the process of identifying influencers on specific topics throughout Twitter. The only way to do this is to immerse yourself into the conversational context data manually. In fact, my own Klout profile says I’m most influential on the topics of “Social Media,” “Marketing,” “New York Yankees,” “Media,” “Family,” and “Public Relations.” Now I can see where some of those terms may come in to play as I do social media and digital marketing for a living, but I rarely ever tweet about topics related to my family or families in general. Perhaps a better example is a friend of mine that ranks as an influencer in “football” when in fact she probably can’t even name more than five teams in the NFL!
Now Klout claims they get their content analysis data on influencers from semantically analyzing tweets for engagement and reach to see which topics one is most influential in, but clearly this is a flawed method. I’m not saying it’s 100% inaccurate, but there is bad data coming back here. In fact, I recently spoke to a Klout employee on how I was personally targeted for two of their perks, one for Disney’s “Tangled” film and the other for an NBA promotion; these are two topics that I really couldn’t care less about, yet somehow I received invitations for their swag. When I mentioned this to the employee, I was told “sometimes people slip through the cracks” as if that’s an acceptable answer. If I was a brand paying good money to target influencers and there were people slipping through the cracks, I’d be very upset, especially with the value of some of these perks brands are sending out.
I also asked Joe Fernandez how Klout measures ROI on their perk program after he made the statement that some brands were so thrilled with the results that they came back for another go and was given no clear answer. It appears, from my take on his reply, that brands are using this money as “experimental marketing” and waiting to see if there’s any long tail effect. With no required action by those who receive perks, it’s impossible to measure any sort of ROI. (Klout users that receive perks are not required to say or do anything. The hope is that they will enjoy the perk so much they talk about it on their own.)
All this being said, I find there is value to be had in using the Klout score in social media monitoring as it aids in identifying those Twitter users that “get it” versus those that are speaking to a disengaged audience. While I firmly believe that brands should listen to and treat all customers throughout social media as if they have 100 Klout, I’m a realist and know that sometimes that’s impossible, especially if you’re a brand with thousands of mentions per day. When it comes down to reputation management and someone with a Klout of 50 is talking about how much they hate your brand, you’re going to want to put them on a higher priority list than someone with a Klout of 8. This score gives marketers some data to go back to CMOs with that satiates their need for numbers and data clustering.
I don’t want to come off like a Klout hater – I’d love to see the service thrive, but only if it’s doing it right. I had many conversations around this topic this week and can say that there are many out there that loathe Klout and everything it stands for. I just want any tool out there that can help me make my job easier. I don’t think we’re anywhere close to finding a way to automate the influencer identification process as there are too many variables that go into what makes one influential, but of everything out there, Klout at least seems to be heading down the right path, even if that path is thousands of miles long.
Sentiment, Schmentiment
One thing I experience with clients involved in social media conversation mining is an obsession with sentiment. While I agree wholeheartedly that how the brand is perceived in the social space is important, relying on automated sentiment as that gauge is extremely misleading and a bad idea.
The problem I find with the gamut of conversation mining tools, both free and paid, is how they handle automatic sentiment analysis. Some of the paid services will boast accuracy rates of over 80%, but I can’t think of one instance where any of them even came close to that.
Sarcasm
Despite what these listening services say, none can automatically track sentiment accurately in cases of sarcasm. “Gotta love the vomit taste in my mouth after drinking (brand x) soda” is going to show up as a positive mention, despite anyone with a half a brain knowing this is not the case. Some systems are smarter than others and can detect the inclusion of negative terms in the same sentence, but even in that case, more likely than not they mark the sentiment as “neutral.” In the aforementioned case, there is nothing positive or neutral about that statement; a vomit-like taste is never appealing!
Context
Another problem I experience with clients is how automatic sentiments picks up their brand mentions in blogs and news outlets. I have found on many occasions that the mere mention of a brand in a news article or blog post that also happens to contain a negative word anywhere within the context of that story will instantly register as a negative hit. Just because the overall tone of an article may be negative and happens to mention your brand, doesn’t always mean your brand is what is attracting the negative attention. Again, some software is smarter and limits it down to the paragraph or sentence level, but even so this is not always accurate.
The only way to really judge sentiment on your brand is to do the research. If your brand receives a manageable amount of mentions a day, registering sentiment manually is simple. For those brands that receive large amounts of mentions, where reading through every one is not feasible, automatic sentiment is a good starting point, but that’s it. Watching how automatic sentiment changes over time can be a good indicator of red flags that require your attention, but certainly should be no more than that.
One thing I recommend to clients is to stop living and dying by the sentiment score. By taking sentiment as a tool of measurement off the table from the beginning, it avoids so many problems down the road.
When you are sick, you go to the doctor for a diagnosis. Sure, you may use WebMD as a guide, but ultimately you need the human touch to accurately judge your health. How is this any different for the health of your brand? It’s not.
Follow Me
