There is much talk about measuring influence lately. Most of the discussions surrounding it are based around people’s activity on social media services. Klout is a service that has emerged as a current leader when it comes to trying to determine influence for people. They do a pretty good job of calculating this based on many characteristics of user activity on Twitter and now Facebook. Unfortunately this only tells a small portion of the influence story for people.
Yesterday I wrote a blog post and over time I watched some activity surrounding it on Twitter. There was one tweet from Jesse Newhart that also got retweeted by 3 other folks, all of which have some significant influence according to Klout.
So what’s the problem here? None of this activity will affect my Klout score. While the secret sauce to Klout’s algorithm isn’t public, we know that they factor retweets in the scoring. But a tweet needs to have the originator’s @ referenced within it for any inclusion in the scoring. So as far as Klout is concerned these actions around my blog post are invisible. Making some assumptions about the tweet example above, Jesse Newhart will gain some influence scoring whereas I (the creator of the content that Jesse shared) would gain none. I gave this some thought and it elicited all kinds of issues around user activity that Klout is missing as well as trying to determine how a service like them could go about including it.
While Twitter is a great service and definitely has its place as a factor in calculating influence, there are so many other factors that need consideration. Blogging is a social media activity that arguably should garner a larger level of influence over a tweet if measured in a similar fashion. So I started to ask myself what metrics could be used to calculate blogger scoring and how could a service like Klout incorporate this information into their scoring?
So I started to do a brain dump on some ways that you could aggregate some data points from third party metrics to try and create an influence score for blogging. There are quite a few ways that this could be done.
- RSS Subscribers – Feedburner
- Blog Traffic – Quantcast, Google Analytics
- Comment Activity – Disqus, Echo, Intense Debate
- Social Media Activity – Postrank, Sharethis, Bitly, Tweetmeme, Facebook
The numbers generated around the above activities provided by the third party services could be used to provide a blogger influence score. For Klout to implement this they would need to somehow allow their users to claim blogs and associate them with their accounts or perhaps the data could be inherited by partnering with some of the above services.
Today I saw a blog post on the Top 10 Most Influential Independent PR Bloggers on Twitter. This top 10 list ranked PR bloggers calculated only using their Klout scores. I’m curious how these rankings would have been affected by using the blog activity metrics I list above? This list is a prime example of why only using a Klout score for ranking is troublesome and I’ve seen similar ones.
I would love to see a blogger score metric used as a calculation in a person’s influence. In fact I personally would put more weight behind this value in an overall influence score. Other areas of the Klout summary could benefit from this data as well including the topic summary to paint a better picture about a users knowledge. We are still in the early stages of measuring and determining influence. Using data to help calculate it is tricky, but I don’t feel the full story is being told and I’m looking forward to watching tools and services evolve in this area.
UPDATE: After reading the comment from Mahendra Palsule below I read his posts and testing out Peerindex. I started to type a reply but it grew to become more useful as an update
I went ahead and read both (post 1 & post 2) of your blog posts that provide insight into their service. I then also read about how they calculate their scores. Unfortunately as you’ve stated, although they let you add your blogs, they don’t currently seem to factor any of its metrics into their calculations.
Also, aside from their different methodology of ranking influence from Klout, I don’t find their user summary pages to be anywhere near as useful or accurate. When comparing my topics on Peerindex to Klout, I found Klout’s to not only be more accurate but also much more targeted than the broad ones used by Peerindex (Why is Robert Scoble listed as a topic for me?). I’m not sure how Klout calculates this but it appears that it may be based on the frequency of keywords used in the lists I’ve been added to if this service that creates a tag cloud I found a while back is any indication. In any case, it’s more accurate.
Also, the descriptions of how the 3 scores (Authority, Activity, and Audience) are calculated are a bit cryptic. For instance “Too much irrelevant activity can hurt your overall activity score” what is considered irrelevant? Also, “At PeerIndex, we look at the content your share, including the links you recommend, and score it against other members of a given topic.” It doesn’t look like they’re tracking more than a small number of high level topics, what about those that don’t fall into this category?
On Klout they provide some simple insight about some factors on how your score is being calculated with their use of descriptive achievement badges. They also provide even more details regarding people with their Klout classification. So while I don’t deny that Peerindex may take more data into account, ultimately it’s the way you present that data to users that will determine its usefulness. I will continue to follow Peerindex as well to see how they improve the service.
I wish you had included PeerIndex in this post. PeerIndex is a service competing with Klout, that does take into account your blog/website. I have done a couple of posts on PeerIndex/Klout. Besides Twitter and Facebook, both these are now starting to use LinkedIn as well.
Liked the suggestions for incorporating other activities and using them for influence metrics.
Mahendra, your wish has been granted. Thanks for letting me know about Peerindex. I started replying to your comment here but it got so long and became relevant to the post that I added it as an update above 🙂
I’m a topic for you? That’s funny.
Heh, I love both the content you create and curate Robert, but I have no idea why you’re listed as a topic for me on Peerindex.
Hi Mark
Quick response since you raise many issues that are best addressed in a blog post–which I will do in the next few days.
Starting point – we take the view that authority / eminence is domain specific. And that someone who is ‘big’ in soccer is not necessarily ‘big’ in oncology. Single scores work very well where there is data to back them up–a great example is credit scoring (both of consumers and of corporate bonds). People know what AAA means to a bond (or did until the CDO debacle)–and that is the value of a single number. But how do you compare an oncology-specialist with a score of 45 against a tech and a geek with a score of 45? (That wasn’t a rhetorical question!)
On the calculation methodology – we describe it in plain English, and I’ll describe it again: conversations with people matter, conversations with high ranking people matter, retweets (of your content) matter, activity matters, noise hurts you, your behaviour relative to the people near you (and globally) matters.
We can’t break this down into a simple statements of “every retweet of you by someone with a score of X, increases your score by Y” because as you’ll understand the maths doesn’t work that way.
And in terms of giving you more benefit for higher status users engaging with you–that is the meat and drink of classic social network analysis (like eigenvector centrality or Pagerank) and I would be very surprised if that isn’t being implemented by other players in this market.
And yes–we run these calculations hundreds of topics.
Now our approach is we think fairly robust in the fat-middle. The fat-middle is where people like you, I or Scoble reside. Basically, not non-entities but not superstars. Not people who are already household names (like Gaga). People who you’d recognise if you were in that community, but wouldn’t recognise if you were outside it.
The critical thing is the robustness within a given community or domain. As CEO of PeerIndex, time after time I am delighted with just how well we seem to be able to describe these communities and the authoritative people within them. Of course, being in early beta stuff goes awry from time to time — we wrongly identify people as spammers (we did this with @fredwilson last week) and occasionally spammers fall through the cracks.
Where it can’t easily be stretched is measuring the extremes (think Gaga, Spears, Bieber) – but these are use cases which we don’t think have huge applicability to our customers or to consumers at large, however good ‘scoring’ these is for generating hoopla.
In terms of topics, we are doing a genuinely horrible job of communicating what we do with topics. When we launched in July our objective was to unveil publicly 3 to 5 topics per week. As we got user and customer feedback, we realised that the presentation of topics in this way was not achieving what people wanted so we stopped doing that–just after we created a topic on Scoble (something we created for fun).
For that reason, you’ll have Scoble appear as as topic–we’re choosing a selection of the 40 topics we’ve made public, rather than the several thousand terms we track. And as you point out it’s counter intuitive. Or as I like to say a ‘cluster fuck’ — we’ve got a major revision of the consumer platform due by the end of the year, and it should clear up the issues you’ve raised.
But the counter point to the horrible communication job of topics, is that we’re doing an increasingly good job of identifying what people care about and how much other people care about the stuff you care about. And we’ve found a really nice way of presenting it. Stay tuned!
An interesting blog post. At mBLAST, we agree with your assessment that any calculation of Influence should factor in blogs. In fact, mBLAST is taking it a step further to include blogs PLUS many additional content sources to ensure that topical relevancy is an integral component in the calculation of Influence. As we have outlined many times recently in our blog, without topical relevancy it is virtually impossible to have Influence. Please follow us at @mblast and http://www.mblast.com. We’re putting the final touches on the mBLAST Influencer application and will be sharing more details very soon.
Thanks for providing more insight into Peerindex Azeem.I look forward to watching your service improve with new releases. Any information on the inclusion of blog analytics to a user? It’s nice that you allow them to be added but what are plans for incorporating their metrics in the future?
Pingback: Quora
Pingback: Klout is a Good Start But We Need More Ingredients for an Influencer Recipe | krynsky.com
Mark, excellent remarks. Here is (in part) the answer to your question on what the top PR influencer list would look like with more thorough metrics: http://lists.traackr.com/PR2dot0One important precision: Traackr authority lists are generated based on topical areas defined by a set of keywords. In this specific case, the topic was “PR 2.0” rather than PR, and the keywords used had to people moving PR forward. Here is the post where we explain what that means: http://traackr.com/blog/2009/12/faq-for-the-top-25-pr20-list/
Pingback: Klout versus Reality | Geoff Livingston's Blog
I am pretty sure Klout counts the Twitter API retweets, they had to actually adjust their algorithm when Twitter launched it because retweets became much more common.
They are very transparent about what makes a score go up, consistency and reaction to your activity (replies, retweets, bit.ly clicks, etc)
They also have plans to include many of the other metrics you have mentioned (I know they are looking at blog comment systems) as well as LinkedIn, Foursquare and others.
Keep in mind the algorithm they are trying to build is a massive undertaking. They are barely through the first inning of what they are trying to build.
Hi Mark
We’ll start to add blog analytics at some point, but our sense is that companies building personal business cards like about.me or the blog analytics specialists like PostRank are well placed to do this.
We’d love to talk to them about partnering–why build when you can partner? But right now the dance card is pretty full handling soc net profiles–and it achieves what we need.
Really Robert, you are a topic for *a lot of people*, in fact on a like-for-like intensity levels, we have 15,000 people who talk about/ share and actively engage with your stuff. Which is the same order of magnitude as we see for Tennis or Economics.
So within the twecho chamber, I am not sure if you are a person or shouldn’t be reclassified as a sport or a whole domain of enquiry.
Mark-it’s because of the thing I mentioned above. We privately have several hundred topics for you, but we only show 40 right now, and one of the nearest matches is Scoble. And by your own admission you love Robert’s activity online (and know him) so it’s not unreasonable there is a match.
I think Klout provides some insight to how their scores are calculated but I wouldn’t go so far as saying they’re “very transparent”. In my mind that would require them to show us what’s behind the curtain (the algorithm).
I understand that what these services are trying to do takes a massive undertaking. But that’s the challenge they’ve decided to tackle and to do it properly they definitely need more data and even then it will be difficult to automate because of what I believe are a larger number of edge cases than normal here. I think lowering that factor is key.
I’ll give you that. very transparent is wrong. But if you ask them they are very open about how it works, what exactly impacts your score, how to raise it, etc. Even what percentages Twitter and Facebook represent respectively and how that interacts.
The exact algorithm will never be public obviously. But they let enough of the magic be seen to show how they work, what they emphasize, and if you care, how to improve an individual’s score.
They have a ways to go. And more data sources is a big part of that. However, from the edge cases I have seen, they have a system that is working pretty damn well. People are and will be reluctant to accept their results. That is unavoidable. Objectively, if you stripped names away and looked at activity, you would be hard pressed to argue with their results.
Keep in mind what they are producing is a very narrow data point. They are saying how successful someone can be in getting reactions from others on Twitter and Facebook (for now). That is useful to a lot of people for sure, but by no means should it be the end all be all of measuring influence or choosing who to interact with online from the point of view of a business or agency.
Thanks for the discussion Mark.
Solid points here Mark. Check out Spot Influence (http://www.spotinfluence.com/company;jsessionid=510E6239A02D25F09085818CC6D892F6) currently in Alpha.
Based out of Boulder, CO and nearly ready to launch.
Pingback: What Klout Is And Is Not | mitchneff.com
Pingback: Klout: How Brands Are Sizing You Up? | She Posts
Pingback: Got Klout? Get Free Beer! | DCeventjunkie
Pingback: 4 Ways Klout Can Evolve
Pingback: Changing To-Do Lists To To-Be Lists
Pingback: Social Media Marketing HQ | Learn Social Media From the Industry's Brightest Minds » My Social Media 2011 TO BE List – What’s Yours?
Pingback: TummelVision 63: Mark Krynsky on X Prizes, competition, cooperation, and “influence” | Tummelvision
Generally I do not post on blogs, but I would like to say that this post really forced me to do so, Excellent post!
http://uggbootssale.hotinternetsecrets.com/2011/04/30/latest-coach-fashion-handbags/
http://adoodleaday.co.uk/uggbaileybuttonboots/2011/04/30/latest-coach-fashion-handbags/
http://ugghandbags.aixin.us/2011/04/30/latest-coach-fashion-handbags/
http://newuggboots.petzplus.net/wpmu/2011/04/30/latest-coach-fashion-handbags/
http://sotolog.net/uggtallboots/2011/04/30/log22.html
http://math-lib.org/blogs/classicshortuggboots/2011/04/30/latest-coach-fashion-handbags/
http://math-lib.org/blogs/tallclassicuggs/2011/04/30/latest-coach-fashion-handbags/
http://cheapuggsforsale.smooth-zone.com/2011/04/30/latest-coach-fashion-handbags/
http://uggbootskids.petzplus.net/wpmu/2011/04/30/latest-coach-fashion-handbags/
http://uggboot.acrestrategies.com/2011/04/30/latest-coach-fashion-handbags/
chaussures louboutin marque rouge les semelles sont à la mode incontestable des timbres de l’excellence.christian louboutin pas cher. Depuis le lancement,christian louboutin homme, son label élégant,Chaussures Louboutin Soldes, en 1991, le designer fran?ais a été fidèle à son but “chaussures réparées,Christian Louboutin Escarpins, comme des bijoux?. De la spectaculaire fines chaussures à talons hauts,Christian Louboutin Wedges, bottes et chaussures en tissu,Christian Louboutin Sandales, Christian Louboutin marqueterie est le principal de chaque femme pour le talon du ciel et accessoires convoitable.Christian Louboutin Bottes.
Pingback: can you start the hcg diet the last two days of your period or will that make the hcg not get into your system? | Diet Plans That Work
coal price of
0.175 yuan to receive tax / kcal. Hubei Enshi district heat for the
semi-bituminous coal in north face winter
coats October 5500 kcal for 2 times higher, the overall increase of
50 yuan / ton, the price is now up to pit block 750 yuan / ton.