Why Facebook’s EdgeRank can make you invisible. And how to fight back.
Person or Brand – the Social Network’s algorithm now determines how relevant you are.
Have you ever wondered who shows up in your Facebook TOP NEWS stream? And who does not? You probably felt that there’s a pattern behind it. You probably also felt you don’t have a clue what that pattern is.
Facebook gave us a quick look at the math behind this, when they explained their EdgeRank algorithm at F8 developer conference in April. But in the last few weeks, more and more light has been shed on how it actually works. (I added a list of references at end of this article).
Here is the official part, the algorithm shown by Facebook at F8.
The EdgeRank is basically an individualized relevance score for each social object placed in the network: be it status update or video. So what is an Edge? Figuratively spoken, it’s like a finger pointing at your content. EdgeRank sums up the importance of these Edges for each individual facebook user. It does this by looking at three factors:
- User Affinity (U): How closely has the possible receiver interacted with you in the past. If someone talks to you frequently, then his chance of seeing something you post increases. Other factors might also play a role. Facebook mentioned at F8 that it’s aware that Brand Pages are different. We rarely chat or send messages back and forth with a page. Be aware that one-way interactions from your side don’t count. You clicking on someone’s profile won’t make your stuff show up in their feed.
- Weight (W): What kind of interaction is taking place? Some interactions/Edges are more equal than others. A comment seems to get more points than a Like. But we still know very little about this. When asked about the value of a click on an outbound link, etc, Facebook developers at F8 made it clear that some EdgeRank secrets will remain untold.
- Decay (D): How fresh is the post. This one is easy. The younger, the better. It also seems that posting new content improves the value for older objects.
These values are then multiplied to create an Edge value. Each object’s EdgeRank is the sum of the values of all Edges that point to it.
By limiting the Top News Feed to Objects with a high EdgeRank, Facebook created an automatic News Feed Optimization that vaporizes the social information tsunami and turns it into a manageable flow. While pleasant for me as a reader, and in line with Clay Shirkys prediction that the future is all about information filtering, this has tough social consequences: It can be the equivalent of a party where you start telling a joke, and everybody turns to leave.
With EdgeRank, Facebook is redefining the way we interact with our peers the way Google influenced the way newspapers write. (Ask a journalist :-) At it’s core it re-interprets “important for us” to be the same as “important for others”. It isn’t exactly a stretch to assume that we will adapt to fill out that reward strucuture. Wheter we have a Brand Page to promote, or whether we just want to make sure that our friends see our vacation pictures of Thailand.
A new thing to optimize: Social Relevance
Call it Social Relevance Optimization (SRO), call it dancing to Facebook’s tune: either way, chances are you’re already doing a good bit of it. From the party invitations to the funny videos that you post, you probably make sure not to annoy your friends with duds. And if someone leaves a comment on your wall, your tend to answer it. It’s all pretty much common sense and being polite.
In the old show-all-newsfeed this was actually the best we could do. The one-size-fits-all information filter was radical decay: When we looked at a feed we would see everything in our network, but only the 1-2 hours of it. Twitter still works like that. We probably missed most of the content, but at least everyone had an equally miserable chance to be seen.
Now that Facebook’s “Top News” is doing most of the filtering for us, brands and people who don’t know how to please the system will quietly fall through the cracks that EdgeRank pries open for them.
Difficult for virtual friendships. Desasterous for brand pages. Your brand might be on Facebook. You might have enticed hundreds of people to “Like” it. And yet: In the darkness that is a low EdgeRank, no one will hear you post. Social Media consultants BrandGlue estimate that since the introduction of EdgeRank, less than 1% of Likers/Fans see a new BrandPage post.
Welcome to SRO 101
Here is how you drag your Facebook brand page out of that non-relevance hole. The Daily Beast did an interesting one-month experiment with a new profile, and described what it took to make “Phil Simonetti” show up in other people’s feeds. They list a bunch of cool insights, but it really boils down to three things:
1. Promote your way out of the new page curse fast
A young brand page with few friends has a hard time for two reasons. First, the Interaction history with your Likers contains almost nothing (1 Like), so the Affinity value of your new object/post to anyone is rather low. Secondly, as long as you have less than 10000 Likers, you’re going to get very little interaction. It takes a lot of readers to get a stream of comments. Vitrue offers a free Whitepaer that claims the interaction on Facebook is around 0.5%. (Still better than Twitter where only 1 out of 300 will be retweeted.)
Your way out? Don’t trust the “Social Media are free.” hooopla when it comes to building your network. Use everything you’ve got to get past the treshold. Other websites, print, radio, TV, promotions.
If you can’t throw around money like that, do it the hard way: Bring some time, and be very nice to the first fans. Especially to the ones with few fans! Build deep recurring engement with them, because you might actually show up in their feed. Your changes of coming through on the feed of a popular Facebook member with 500 friends or more are close to nothing. (However, if you can convince someone with a 1000 friends to maniacally click on your content, by all means do it. :-)
2. Create Objects that invite an Edge
An Edge only happens when someone interacts with your content. So this is kinda obvious. As soon as you have enough Likers to be reasnoably certain that someone will answer, invite feedback. Ask questions. Throw in a provocative update to get the discursive juices going. Create comment surveys.
Keep in mind, though, that the relationship between EdgeRank’s AFFINITY and WEIGHT is complex. Facebook attributes different Weights (W) to different Edges. The Daily Beast experiment hints at the following sequence of feedworthyness: Comments beat “Likes”. Likes beat posted links. Links beat the opening/view of fotos and videos. This makes a lot of sense, as mirrors your effort and thus your commitment to a piece of content.
Logically, the likelihood of any of these things to happen is almost inverse. And the amount of responses influences your future AFFINITY ranking. According to a whitepaper by Vitrue, if you are trying to get a response, photos beat videos, videos beat links, and links beat Status Updates.
You’re going to have to decide what is more important to you: If your page has very little interaction yet, you should post photos and videos. If you already receive lots of views, and Likes, upgrade your strategy by asking for comments.
3. Get the right people to act first
Finally we get to the splendidness of Social Media, where an Idea will travel beyond its initial reception or seeding and go its own merry way reaching ever wider circles. An adventurous voyage, fueled by chance and serendipity. Facebook’s EdgeRank is turning it into a chaperoned walk in the park.
Why? Because the love and attention that you receive from a “Liker” will not be featured everywhere along the Social Graph, but only in his/her area. And As we already said, it is rather unlikely for posts to feature in the feeds of a wildly interactive power Facebooker who receives loads of attention from numerous friends.
A much more likely scenario is that your content will travel along a quiet backroad. If the first guy or girl is from a not so releveant subgroup among your Fans/Likers, that subgroup might be the only place your object every travels to, before it’s EdgeRank Decay value (D) buries it for good.
If you’re no Facebook bigshot yet, your task is to create reactions quickly, and in the right interest group. One of the ways to make sure your object gets Edges from the right people is an intial burst of facebook advertising. You can even target the ads to the most relevant subset of your BrandPage’s Likers, which should make it very efficient. This will invite the right people to act, and create early simultaneous actions/Edges. Which enables that object to show up in many more feeds, creating more interaction, which will help keep decay at bay.
Filter failure avoided. What about Filter dominance?
Facebook is neither evil nor innocent. While their ranking obviously helps me filter the information of 500 friends, the algorithm may also clandestinely support behavior that Facebook wants to happen: Facebook needs to promote groups? Tweak a line of code, and interactions/Edges with groups will score better in EdgeRank and move up in the feed.
But the even bigger challenge for Facebook lies in the inherent promise and problem of a post-broadcast communication system. We’ve all become information providers. But what if the information I produce isn’t deemed worthy by an impartial and utterly unemotional algorithm. What if EdgeRank thinks me and you are just noise.
Did 500 million people join Facebook to find out that only 50 million should be heard?
I’m a narcissist. ;-)
How about you?
The Article that made me start looking into this: http://www.doseofdigital.com/2010/07/facebook-page-exist/
Navigate to the Techniques sessions, and click on ‘Focus on Feed’. Questions about EDGERANK begins after about 20 minutes:
Video from F8 conference: Go to “Techniques”, click “Focus on Feed”, EdgeRank discussion starts after about 21 minutes. http://apps.facebook.com/feightlive/
Vitrue Whitepaper on interaction rates: http://vitrue.com/blog/2010/09/21/anatomy-of-a-facebook-post-vitrue%E2%80%99s-data-behind-effective-social-media-marketing/
Informed speculation on what could be measured in EdgeRank: http://www.quora.com/How-does-Facebook-calculate-weight-for-edges-in-the-EdgeRank-formula
Link to HP Labs study on “Influence and Passivity in Social Media” which of course opens the same can of worms for Twitter ;-): http://www.jeffbullas.com/2010/10/08/average-twitter-user-only-retweets-1-in-every-318-links/
And finally: You can now vote for the “Black Art of Optimizing Facebook Wall posts” for SXSW 2011!: http://panelpicker.sxsw.com/ideas/view/7760