Posted on September 24, 2011 by
Spread the word...

Last week I posted an article highlighting how you could game Klout. In my post and the subsequent commentary I argued that:

  • Klout’s model of scoring seems to be based predominantly on levels of network activity which is a rather primitive approach to “measuring influence”.
  • Klout scores seem to fluctuate on daily basis, influence and/or authority can not be that volatile.
  • Most importantly, there should be a clear distinction between, “noise”, “activity” and “influence”. If someone is very active and noisy, it does not necessarily mean that they are influential. Joe Fernandez should pay close attention to the following word and its definition. (Troll: “A troll is someone who posts inflammatory, extraneous, or off-topic messages in an online community, such as an online discussion forum, chat room, or blog, with the primary intent of provoking readers into an emotional response or of otherwise disrupting normal on-topic discussion.”)
  • A troll is a nuisance, not an influencer.

In this post I will highlight a few other shockingly inaccurate cases which demonstrate how ineffective and flawed Klout scores are. I am sure Klout’s PR team would try to claim that my previous revelation was an isolated case. And talking of PR, notice in all their corporate comms Klout uses, in fact, overuses the word “science”, “scientific” and “science team”, here are a few examples. Frank Luntz anyone?

Introducing @mi

Say hello to @mi, a lonely Twitter profile which doesn’t have any followers, it hasn’t been listed and it doesn’t even follow any one, see screenshot:

@Mi

Go on, have a guess & be very generous, what is @mi’s Klout score?

http://klout.com/#/mi

This is absolutely ridiculous! How can an account with two tweets, no other profile signals and a default “egg” avatar score 73 out of 100 for “health food”? Let’s compare this to Dr Weil’s account who has 237,426 followers and has been listed 4,423 times and is influential about “health” and “diet”, see below.

http://klout.com/#/DrWeil

Is this how science works? Should advertisers use Klout (as it is) as a scientific barometer? You decide…

Who is more influential on “social media”, Evan Williams or Joe Fernandez?

I would be very interested to find out who you think is more influential on social media, Evan Williams the co-founder of Twitter or Joe Fernandez co-founder of Klout. I am not sure about you but I think Evan Williams is much more influential when it comes to social media as compared to Joe Fernandez.

Poll – Evan Williams vs. Joe Fernandez – Please vote.

Please use the above poll to vote, I will publish the results once the the poll is over.

For now, let’s resort to Klout’s version of the story and find out who is more influential. Actually before that, let’s take a look at their profile stats:

Joe Fernandez:
  • 4,832 Tweets
  • Following 654
  • Followers 8,098
  • Listed 792
Evan Williams
  • 6,333 Tweets
  • Following 1,346
  • Followers 1,395,276
  • Listed 17,337

Joe Fernandez’s Klout

As you can see in the screenshot below, Joe scores 72 out of 100 and is influential about social media based on Klout’s assessment.

Evan Williams’s Klout

Evan Williams who co-founded Twitter, one of the biggest social media networks, is less influential about social media as compared to Joe Fernandez. If you take a look at http://klout.com/#/ev/topics you will find that social media is the forth topic in the list.

William Evan’s Klout

Now if I was an advertiser who didn’t know who Joe or Evan are and therefore relied on using Klout as a “measure of influence” I will probably target Joe instead of Evan because he has a much higher score.  Would I be right though? Would my money be well spent? Would my targeting be accurate? For me the answer is a big no.

Doing it Las Vegas style

On 30th September 2010, Adage.com reported that Palms Hotel & Casino in Las Vegas will be using Klout to “allow high-ranking influencers to experience Palms’ impressive set of amenities in hopes that these influencers will want to communicate their positive experience to their followers.”

The Scenario

“You’ve just walked up to the front desk at the Palms Hotel and Casino in Las Vegas and given your name. The clerk pulls up your information and right there next to your reservation is a number that will determine what kind of treatment you’re going to get. Will you be invited to hang out in the exclusive Hugh Hefner Villa? Maybe. It all depends on that number, your Klout Score — a ranking that will follow you around, whether you know it or not, and tell the world your worth as a consumer … and eventually, as a friend.”

Based on the above scenario, if @mi or @_borgeCollective or any other spammer/scammer turns up at Palms Hotel with a Klout score of 80, would it be right to give them a “different (am assuming better) treatment than the another customer who has a Klout score of let’s say 50?

Wouldn’t that expose businesses to a new kind of threat? A threat where people who manipulate Klout get better deals and treatment at the expense of truly loyal customers?

Questioning Klout’s ethics

The cases I highlighted above and in my previous post clearly prove how unreliable Klout is. I am sure Klout already recognizes that their scoring system is not accurate and needs a lot of work. I have a few questions and I would love to hear answers from all stakeholders:

  • Is it ethical to sell and promote a premature product knowing that it does not deliver accurate results?
  • After knowing that Klout is inaccurate, what do Subway, HP, WWF, Pepsi, Audi, Nudie Jeans think of Klout? (Note: All links point to Klout Perks for relevant brands.)

Conclusion

Everything I have found leads me to believe that Klout is far from accurate and it shouldn’t be marketed because it can be easily games, manipulated and there are fundamental errors in the way it tries to measure influence. If Klout want’s run trials with a certain brands then those brands need to know the real deal about Klout instead of the spin and PR talk.

To restore confidence, Klout should clarify things instead of trying to simply get away with “our algorithms are being constantly improved” responses. Furthermore, I propose the following two features that might help Klout in its journey towards unattainable perfection.

  • Introduce a -K score – give weight to votes originating from influencers who are influential in closely related topics. (Yes, it is slightly tricky and prone to abuse but I am sure you science team can get it right.).
  • Integrate Radian6 or something similar and look for metrics outside the currently connected social platforms specially popularity and sentiment metrics from forums, blog posts, articles etc. For example, @mi and @_borgCollective do not exist outside the “connected networks”. If a profile doesn’t have a wider social footprint then it should be treated totally differently.

Looking at you, yes you, I can tell you love sharing, that is why I have made this easy for you to share this post on Twitter

Leave a Comment

Your email address will not be made public or shared. Inappropriate and irrelevant comments will be removed.

  1. it shouldn’t be marketed .But it does say Beta on site but people are easy to jump on things.Do i believe a program can tell me what my influence to people is NO. it is clearly obvious on my followers on twitter and the hits on my blog with that said i really care nothing about my klout score is i look at the real demographics are. Keep up the great work ….

    • Yousaf says:

      Yes, ideally it shouldn’t be marketed but Klout has to make money and implement real world campaigns to better itself. Therefore, it is okay to market it encouraging big brands to trial it. My only concern is, do these brands/advertisers know that Klout isn’t accurate? Are they being informed? Or are they being told a different story?

  2. [...] Google: The Small Business Serial Killer.UPDATE: Made some more shocking discoveries about Klout, check my new post. Tweet A little about Yousaf … Yousaf is the Group Head of Search & Social Marketing at [...]

  3. [...] exemple, @_BorgCollective et @Mi ont respectivement des scores de 57 et de 73 sur Klout. Le cas de @Mi est intéressant, Klout lui [...]

  4. Hey Yousaf,

    I am not sure why @mi is getting so much interaction but check out twitter search (http://twitter.com/#!/search/%40mi). There is obviously more going on here then your article suggests.

    Joe

    • Yousaf says:

      Hi Joe,

      Yes, even the guy who owned @mi had to change his username from @mi to @ldnmi because he was being mentioned by mistake.

      Yes, there is more going on, Klout simply isn’t intelligent enough to detect what is going on. If you look at http://klout.com/#/ldnmi which is his new account you are giving him 32 out 100 which could be realisitic-ish. However, Klout fails to understand that the username has been changed.

      Peerindex on the other hand does a good job as it consolidates those accounts, so when you look at http://www.peerindex.com/mi or http://www.peerindex.com/ldnmi you get the same score and profile.

      It is just a matter of doing things intelligently, some people might change their Twitter user name so your system has to cater for that and take that into account. Let’s assume Ashton Kutcher decides that @aplusk isn’t for him anymore, he changes it to @ashtonkutcher. You are going to rate his new account from 0 out of 100, all over again?

      If I change my Twitter username from @ysekand to @ysekander, Klout doesn’t know that therefore I will loose my “social capital” or my “influence” and would have to start all over again to get a decent kscore.

      What are your thoughts regarding Evan Williams issue? Is it fair or accurate that you have a higher score than him in the social media sector?

      • Dave says:

        Focusing on a 1-in-100,000-type edge case really tells us nothing about Klout’s algorithm. But you not pointing out that people us @mi to mean “at my” tells us something about your bias. Why not explain the situation? Why hide the facts?

        The whole point of a good algo is to handle as large a percentage of common cases as possible. No algorithm should be weighted to handle 1-in-100,000-type edge cases.

        Klout has many failings, but emphasizing @mi as a “shocking case study” is simply really, really misleading.

  5. People change their twitter usernames every day and we are able to quickly reassign the influence they had previously earned. This is a unique case where the old acct is still driving huge amounts of engagement.

    For the most part what you are pointing out are edge cases which we appreciate and will continue to improve upon.

    • Yousaf says:

      These could very well be edge cases and you continuing to improve Klout is indeed commendable.

      I have two questions though:

      1. What transparency initiatives* have you put in place?

      2. If you don’t have any transparency initiatives in place and you are aware of Klout’s rough edges then why are you marketing it so aggressively?

      Hint: http://trust.salesforce.com/trust/status/

  6. What is the point of gaming Klout? Is there any way of turning a high Klout score in to money or something else tangible?

    • Yousaf says:

      Hi Bill,

      Yes, there is money involved in it. Quickest example would be the Las Vegas scenario I have mentioned above.

      Brands use Klout for WOM & social media marketing, these marketing strategies will be huge in the next couple of years.

  7. was doing little searching ,What’s online influence really worth? For Klout, $8.5M.here is brief post
    .”Klout, a startup that helps people measure how influential they are across a variety of social media sites, announced today that it has scored $8.5 million in a round of funding led by Kleiner Perkins.

    The investment is part of Kleiner’s $250M sFund for social startups, and the firm’s Bing Gordon will be joining Klout’s board of directors.”
    this is now more interesting.

  8. Jon Zanoff says:

    I use klout scores as ONE of many factors when reviewing potentially interesting people to follow. Is measuring social influence a mature and precise science? Absolutely not. In the absence of any other readily available metrics, does klout add value for my purposes? Absolutely. If your goal is to show the accuracy of the data, I’d be interested in hearing about a statistically meaningful sample size. eg a RANDOM selection of a thousand users. Thanks for the thought-provoking post.

    • Yousaf says:

      Hi Jon,

      I have used Klout as one of many factors as well. In fact, if you read my previous post (http://www.rocketmill.co.uk/here-is-how-you-can-game-klout) I have mentioned that I have built a tool that looks into Peerindex, Klout and Followerwonk and then gives me a mean score which is more realistic, to some extent.

      Regarding a statistical review of a large sample, I think that is what Klout should do. That should be one of their transparency initiatives. They should carry out research, publish their methodology along with the results.

      That way me and you along with potential advertisers can look at the data and decide for ourselves. At the moment all we hear from Klout is just marketing speak – big hat, no cattle.

  9. Zeus says:

    Great post – I have noticed for example that it places an illogical amount if weight on twitter versus all other networks.

  10. Yousaf says:

    Hey Zeus,

    Thank you. Yes, any influential person who scores between 60 to 70, if their Twitter activity dips for a few weeks then their score drop like a stone which proves your point.

  11. Linda says:

    You’re WAY off on one thing Yousaf. The Twitter activity doesn’t need to dip for a few weeks for the Klout score to drop.

    When my Klout score began to climb 1 point ever 2 or 3 days I decided to test a theory. I’ve been following a certain behavior for 2 weeks, actively thanking FollowBack teams and any #FF mentions. I was sure to include the original list of people in my thank you. (Most were crap accounts that I would NEVER give the time of day to and possibly even spam/block them if they became too annoying. Believe me, they are annoying!)

    My faithful replies were nothing big, not much conversation at all, and simply said thanks. Yesterday I purposely changed my habits with them and only replied to a few compared to normal over the 2 weeks. I even stopped including all the people the original tweeter included along with my user name in the #WhateverHashTag of choice was for that day or tweet.

    Last night just before midnight I was a 73.33. I woke up this morning to a 73.13 and by 4:00pm today I am at a 69.88.

    My plan was to continue this until I was up there with YouTube’s 100 score, but when I saw I was an influencer of one of these Twitter accounts set up specifically to just tweet mentions to get followers for people, that was it! I sure as hell don’t want anyone I respect in my social media circles to check out my Klout page and see I have nothing but a bunch of fake profiles listed as people I influence.

    Now the plan is to try and keep my score balanced where it is now, but I really feel like this is a no win situation. The amount of mentions I got for these fakers and the number of replies I did are enormous compared to my normal activity. I would need to have no responsibilities to keep up with it all and do the amount of tweeting it really takes to increase my score without a trillion followers. I’m not even at 3000 followers yet and did I mention when I replied with those “thank yous” I included my user name as well so I was mentioning myself too?

    I’m out of breath just thinking about what it takes to “game” Klout, even as simple as it is.

    …and it is simple to do. It takes time, not influence, or brains for that matter.

    • Yousaf says:

      Hi Linda,

      Thanks for sharing your experience, very insightful indeed. I really like the sound of mention oneself, that is something I had not thought about. I am going to experiment that right away and see what happens.

      Just out of curiosity, have you done similar experiments with Peerindex or Followerwonk?

      • Linda says:

        Thanks Yousaf. I haven’t done any similar experiments yet, but I’ll keep you posted. How do you like Peerindex and Followerwonk? What do you find valuable?

  12. Michelle says:

    Hi, Yousaf. I’m not a spammer or a scammer, just an early adopter who likes short user names. I signed up for Twitter in February 2007 with the user name @mi.

    There are two reasons @mi gets many @ replies. First, because “mi” means “my” in Spanish-speaking countries, hence many people saying they’re “@mi casa”, etc. Second, there is a Nigerian rapper called Mister Incredible (or M.I), and many people are under the mistaken impression that @mi is his Twitter account. (Many of them have followed me over to my new account too.)

    Don’t get me wrong. I’m not saying the Klout score on my old account should be as high as it is, or that Klout scores are at all accurate, but you missed part of the story.

    • Yousaf says:

      Hi Michelle,

      Thanks for your comment. I think you have misunderstood me, I have not implied that you are a spammer or scammer. What I have pointed out is the fact that potentially spammers/scammers could abuse Klout. I am sorry if my statement has caused confusion/misunderstanding.

      Regarding your account, yes I know what is going on to some extent. Must admit I didn’t know the Spanish part of it but I do understand why you had to change your username.

    • “@mi” being linked sounds like a bug in Twitter, not in Klout. I don’t know why Twitter can’t detect things like that (e.g. by comparing @-tag rate to follower count) and rename the affected accounts with a double @.

  13. [...] existe des choses vraiment très étranges comme celles rapportés par Yousaf Sekander (merci a Benoit Descary pour l’info) qui témoignent de l’incapacité de Klout de [...]

  14. [...] already seeing free systems emerge for establishing who one should pay attention to (for example, Klout, which is flawed, but at least they’re giving it a go.)  And of course increasingly intelligent [...]

  15. Good points. You’re right, Klout isn’t the best. But what is? Peerindex? MyWebCareer? Something else?

    The value of a Klout score, in my opinion, is more about the trend than the score itself. A positive trend seams to mean that the person is more likely to engage you rather that just sit their passively while a negative trend often means that person isn’t going to bring a lot to a new contact.

    • Yousaf says:

      You can look at it that way and I am in no way saying that Klout is totally useless but I have a problem with their way of promoting it. It is not what they say it is, that’s all.

      • Fair enough. You’re definitely right, their promotion is misleading, but educating folks on what can be useful with it is to me the key (I guess that’s the old teacher in me coming out).

  16. more great research Yousaf, i think it may work by collecting data, so the longer your account has been running, the more accurate your K score is. Again this is something that needs to be raised to Klout, again great research buddy

    • Yousaf says:

      Not really, Klout seems to fluctuate every single day, it takes daily level of activity into account rather than long term level of activity/authority. That is why it is incredibly unreliable.

  17. Oli Conner says:

    Excellent post which has confirmed some of suspicions about Klout. I’ve always wondered why these services don’t connect to bit.ly, surely the amount of clicks you receive is an indication of influence…

    • Yousaf says:

      Thanks for your comment Oli. They are missing out on a lot of useful signals. You can not measure someone’s influence by simply listening to connected networks, you have to look for “external” social footprint and sentiments – at the moment none of these “clout” services are doing that.

  18. think klout getting mad at all there Criticism the been receiving what do you think ?
    http://corp.klout.com/blog/2011/09/keeping-you-updated-on-the-score/
    Here at Klout, we process terabytes of data every day to help you understand and leverage your influence. We have our own internal metrics to verify that every network is being processed correctly and scoring runs smoothly. We know that processing this data correctly is part of what makes people trust Klout.

    This weekend we experienced a bug with our pipeline for processing data for LinkedIn, and this resulted in a score drop for about 0.001% of our scored population which lasted approximately eight hours on Sunday. We know that even though the number of affected users isn’t large, it’s a big deal to anyone who is relying on the score and we take it very seriously.

    We’re working to ensure we keep you, our users, updated about any issues we experience and work to resolve them quickly. We’ve setup a dedicated @KloutStatus twitter account, so you can follow the updates as they happen

    We’re always working to improve our processes and we have big improvements coming very soon. If you think you can help, we’re hiring.

    This entry was posted on Monday, September 26th, 2011

  19. [...] outil complètement artificiel que j’aurai pu maquiller pour faire vrai). La preuve avec More shocking case studies – how to game Klout des exemples de comptes qui ont de très bons scores sur Klout en étant très très nuls en [...]

  20. Jure says:

    Hi Yousaf,

    first thank you for stoping by another day and commenting on my post. Indeed i read both of your post and I am posting by the end of the week something even more shocking that you did. Klout simple needs to admit one thing, they dont measure influence as they have no bases of doing so. Measuring RT actions and reply is not really a indicator of influence. Let’s me give an example, people using RSS feeds as primary feed to twitter is not a curator of contant in first place, second just pressing like button on Facebook either, or many of people do RT just coz they have an idea more people i RT more the same people will RT me. As i said in my post Klout doesn’t measure any after effects and doesn’t look into content itself for example bad review etc….
    Joe Fernandez can speak as much as he wants but one things is clear he contradicts himself in many statements as well. I publicly ask several times to see the proof of their Standards. You can’t just call yourself a Standard and not disclose any studies or data that support your claim. I did today another post based on Klout comments, and its shocking to know that Klout access protected tweets to determine the topics. Will follow up on your post as would like to see the vote results.

  21. Miranda M says:

    Yousef, I usually enjoy your posts and this is no exception. FYI, a few weeks ago, after a night of debauchery in SF involving a black cowboy hat and a Pope bust, some friends and I found ourselves ranking on Klout as Pope experts :) Was easy to game.

  22. [...] just read this post on gaming klout, and this follow up that includes more examples of gaming klout. And my heart [...]

  23. ToddySM says:

    Yousaf, I had the same concerns with Klout last year (see my post: blog.toddysm.com/2010/09/do-you-believe-your-klout-score.html). One more thing that I discovered last week is that although Klout claims that it measures your social network influence the only thing it measures is your activity on Twitter only. I was on vacation and used Facebook actively but no Twitter and my Klout score went down several points.

  24. [...] that happen? Link building, branding, social clout (the real kind, not the kind you can game). I’ll explain more [...]

  25. [...] Presidents Remarks. One More Thing! Google Class Act. Jonathan Mac Image. Gaming Klout 100 #1 Gaming Klout 101 #2 Microsoft courting Yahoo again? Cloning Stem Cells. Pirate Bay lives on. Delta Free Wifi. Cell [...]

  26. alsy77 says:

    Great piece – thanks for taking the time to reply in such detail. I’ll be following from now on :)

  27. Vincent says:

    Same, I took the piss out of religion on a few posts and i was an expert on Jesus. Second to that , when I first checked Klout out my score wasn’t very good. I simple had a few long twitter chats with a couple followers (about 5) and boom my numbers started going up.

  28. This makes complete sense as what never did make sense was someone who has 800 followers listed 18 times and just auto-tweets.

    They have almost the same Klout score as I do with 37,500. followers and 900+ listed and our score is one point difference. That never added up until now. Great detective work. :)

  29. A few days ago, Klout thought I was influential about atheism. No clue why. I deleted that influence reference. Today, it thinks I’m influential about diapers. ???

    I searched my account. Approx 2-1/2 weeks ago, I used the word diaper, once, and a few days later, I RT’d a post with the word diaper. Only 2 uses and now I’m influential?

    To me, Klout is a general reference tool, and fun to check out once in a while. But as my primary use for the @CraftBeersBrews account, is about Craft Beer (and Klout is accurately acknowledging that, as well as properly listing my Beer related followers and those I follow) if it’s THIS easy to disrupt your “influence”, and people take Klout for some level of fact? Then it means you have to tread carefully on what you post, in order to ensure your brand is viewed properly, and you should never skew from that.

  30. [...] apparently very easy to fool or manipulate your score on tools like Klout; (check out this great blog post  for more details). It appears that any significant activity (or in-activity) even over a very [...]

  31. [...] }); /* Auto-Scroll */ }); Elevate Local: SEO CompanySEOPPCSocial MediaWeb DesignAbout UsContactBlogMore shocking case studies – how to game KloutLast week I posted an article highlighting how you could game Klout. In my post and the subsequent [...]

  32. [...] We don’t have to wonder, because the @mi Twitter account already exists, with only two tweets, no followers and no following. And yet it had a Klout score of 73 (very high) before the Kloutpocalypse – and now has a score of 55 (still quite high.) [...]

  33. Tom Weiss says:

    When I tweet, I end my tweet with @myusername so Klout will count it as a mention. hahaha.

  34. [...] More shocking case studies – how to game Klout 1 Upvotes Discuss Flag Submitted 1 min ago Yousaf Social elevatelocal.co.uk Comments [...]