A Search Engine’s View of Eric Ward’s “Key Problems With Current Social Link Graph Signals”

Posted on 24. Jun, 2011 by in Digital Marketing



Filtering the mass of Google Reader subscriptions by magic for my daily consumption of digital, I was presented with a post from Eric Ward on Search Engine Land entitled “Key Problems with Social Link Graph Signals“. For those who do not know Eric Ward personally, like me, he has been in the link building and content publicity business since 1994. Wow, 1994, in ‘the game’ longer than Google. Snap, just like bigmouthmedia.

Eric’s post presents a link builders stance on social link graph signals. Eric starts off by saying:

Social sharing was fine when it was not a search signal. And someday, maybe 20 years from now, it might make search more than just the social curiosity it is now. What bothers me about the social link graph is not that it’s so easily gamed/spammed. It’s that people can be part of each other’s social circles yet have very little in common.

As a fan of social search, I thought I would play the role of a search engine such as Google or Bing, by looking at Eric’s comments and arguing the possibility that social signals are actually useful in a lot of his scenarios.

Nice Pants Jim…No Really

Eric Ward gives the example that he and a friend Jim, who share a social connection, may not appreciate one-another’s shares (tweets, likes, etc) in music.

My friend Jim is a great guy, and I’ve known him a long time. I’ve been in his house, seen his bookshelf and CD/DVD collection, and I know how he dresses and what he drives. And while I enjoy seeing him and catching up on the kids, life, sports, etc., I have no desire to let Jim’s Tweets and Likes and Plusses impact my search results.
Why? Well for one thing, his love for Yanni will be at war with my love for Nine Inch Nails, and isn’t that an interesting battle for the engines to make sense of?
There’s a bit of forced comformity lurking underneath the social link graph, and that is, in my opinion, evil.
This leads me list a few ideas that would give me more confidence in allowing a social circle to affect search results.

Let’s add some context to Eric’s example. If Eric was to search for [good music], Google may use the fact his friend Jim has chosen to 1+ the band Yanni to return them in Eric’s results. Eric wouldn’t be best pleased with seeing Yanni and suggests it’s a case of ‘forced conformity”.

However, as a search engine, I would apply an algorithmic system like Facebook’s “EdgeRank” to counter-act ‘forced conformity’. Instead of blindly assuming that because one friend likes something another will, I would apply Eric Ward’s affinity and interaction with Jim’s social share of Yanni, and with Jim in general, to influence whether his +1 is returned.

For example, if Jim’s +1 is returned but Eric doesn’t click on it, retweet or +1 (interact) why would Google return it again? If Eric did click on it, bounced fast or not share it (enhancing their connection), I as a search engine may let this impact my confidence in the result, suggesting I use the information to make a change i.e. remove social result. I can learn from mistakes and improve. Personalised search is about learning that Eric Ward does not like the musical tastes/content as his connection Jim.

Would it not also be helpful to see that Jim likes Yanni before Eric clicks on the result? That would suggest to Eric that he might save his time and skip that result.

Going a little further, a search engine could establish what results Eric shares, compare them to others out with Eric’s social circle and return those results instead of Jim’s. Of course, if Eric doesn’t share, he’s not playing the game, and that could make a search engines job harder.

Keeping Up (Twitter)

Eric argues that it’s impossible to keep up with more than 100 people their tweetstream, We’ve altered the heading to ‘Keeping Up’ to make it sexier.

It is impossible to follow more than 100 people and actually keep up with their tweetstream, unless you are unemployed. You know it’s true. And if you aren’t unemployed, 90% of those tweets sail right by you because, well, you are working. I know from my own work day that I don’t have time for the noise, even from those 78 people I follow.
The fact that I have 4,000 followers via @ericward amazes me, but it also leads me to believe that rule #1 for Twitter signals must be if you have anything of value to say or share, then you should have earned way more followers than people you follow. So as a starter, any Twitter user with who follows more people than they have following them is not a useful signal.

Search engines will agree with Eric and his Twitter motto that there is no way to keep up with all posts, and yes 90% of users tweets may “sail right by you”. But, search engines would argue that if a tweet were important, and I mean – of value, how likely would it be that of Eric’s 70 followers, he hears or sees that message again – thanks to retweets from his community?

That’s right, critical mass promoting content. One tweet of huge value, followed up by your community interacting with the ‘topic’ increases the likelihood of you seeing that tweet, or topic again. Remember, if you see the topic mentioned, no share, it may ignite your quest to search for it. That search would then result in you seeing the tweet via social search.

It’s hard to miss an entire conversation on Twitter that enthrals a large number of your connections.

As a search engine, I would consider the number of impressions tweets received (we want Twitter stats, chant with me now!) as part of my algorithm. Referring back to Facebook and how it tries to resolve missed posts, with EdgeRank, with the help of its “Most Recent” post list. Twitter may sell Google tweet impressions to enhance their social algorithm.

Paid Likes and Warts (Facebook)

Eric gives a scenario where a brand bought 70,000 ‘Likes’. Again, we’ve altered the heading to ‘Paid Likes and Warts’ to make it a little catchier.

I could write volumes on the flaws with Facebooks social signals, but here are two simple ones. A few weeks ago, I had a call with a prospective client who had a “revolutionary diet product”. I’d never heard of this product in my life, their Facebook page was just a couple months old, and yet it already had 78,000 “Likes”. When I asked how this could be, I was told they had been purchased. They were paid likes. I already knew this, but his willingness to admit it was almost refreshing in it’s dishonesty.

Basically, Eric insinuates that ‘Likes’ can be bought thus making ‘Likes’ a less credible signal.

If the users who were paid to ‘Like’ the brands page don’t interact or have any affinity with the profile, a search engine would make the profiles shares and posts hold less value.

You could buy all the users under the sun, but without interaction and affinity, why would I, the search engine, value their updates? I would be looking for more than just ‘Like’ numbers.

Problem 2: I can kind of understand why a product page like Advil has 12,000 Likes, but what if your product happens to be something that might be very helpful but is not something one wants to disclose they use? And I don’t just mean something like Preparation H (731 Likes, bless them all), what about a product for wart removal, or heaven forbid, genital wart removal? Or how about a Facebook page for a treatment facility for substance abuse? How eager are people to let the world know they are drug abusing hemorrhoid sufferers? Go ahead, Like that.

This scenario is somewhat unexpected from Eric. As a link builder, I’m sure he’s been up against the situation where nobody apparently wants to link to a website due to their nature. I have. However, as marketers we find routes around those problems. Creativity springs to mind.

If you think about it, ‘Why would I like warts?’ also applies to ‘Why would I link to warts?’. Getting links to a wart website could be tough. But there are opportunities; you could get health experts, doctor’s blogs, to link to your product within their website or blog as an avenue to build links.

The exact same scenario applies to social; get health experts with social presence to share and vouch for your product – have associated profiles of business that sell your product ‘Like’ your profile. Social media campaigns are often best suited at tackling an ‘itchy’ situation.

Using the Masses (Google +1)

Here Eric gives an example of a large company and considers their ability to use masses of employees to +1 their website, unfairly influencing results.

I like being able to click the +1 button in the search results, but I also wonder just how this impacts corporate search marketing behavior. Does a company with 250,000 employees have an unfair advantage because they can send an internal note to their entire company asking them the click the +1 button? Is a competing company with only 10,000 employees at a disadvantage?
Remember, none of these +1′s are legitimate anyway (because they are mandated, not earned) and if we are looking only at social circles, wouldn’t IBM employees be more likely to have social circles that included other IBM employees? What’s the point of +1′ing your own company to your own social circle?
Now that the Google +1 button is lose in the wild, this will change things, hopefully for the better, but there remains another far greater problem with Tweets, Likes, and +1′s

Search engines are likely to believe that large companies do not hold an unfair advantage in social search. Eric’s gives the scenario that all 250,000 employees could +1 a website to boost its results.

Thinking about it, the argument could also be applied to links. A CEO sends out an email saying everyone needs a blog so they can write blog posts with rich anchor text pointing back to the company websites. Business doesn’t work like that. I bet the idea has sprung to many search-savvy CEOs minds, and in fact been tried and tested. Plus, I bet search engines will have I.P. detection in place to prevent a ‘machine’ automatically +1 websites or creating multiple social spam profiles.

Google can certainly be aware that 10,000 Google Profiles are all accessed during working hours via the same corporate IP range. That’s easier to spot and therefore safer to trust than some link based quality signals.

However, a search engine may see a large business or network of people as a strong signal, not unfair advantage. If a business’s reach i.e. their employees profiles and social credibility, then if they all +1 a website, their social circle will see and benefit from the share. However, going back to ‘forced conformity’, if it doesn’t appeal to their users, the +1 value will degrade in value.

Old Gold Gets No Social Love

Eric argues that old content won’t receive as many shares as new content due to the lack of people in the social web willing to pass it on.

The problem is with older yet still awesome content. No matter how fantastic and evergreen it may be, it’s less likely to be Tweeted and Liked and Plussed, because back at the time it was created there were not as many people in the social web world to do so.
Here’s an example: Danny Sullivan’s incredible Introducing: The Periodic Table Of SEO Ranking Factors was Tweeted, Liked, and Plussed over 6,000 times in just 2 days. Compare that to the equally important column Danny wrote on Jun 2, 2008 titled Microsoft Wins Deal For Live Search To Be Default On HP Computers.
That was some really big news at the time. I mean Big News. Yet that column does not have a single Tweet, Like, or Plus. So, rule#1 for Google +1 is there needs to be a way to reduce any bias against “older” content that didn’t have the same chance to be “Liked”, since the “Like” functionality didn’t exist yet.

If “Microsoft Wins Deal…” is still linked to, for example, the term discovery, it has equal opportunity to appear and get shared. If it was a competitive search term, and others, e.g. >10, fared better with shares, it wouldn’t rank as well. That’s assuming that shares totally replace links in rankings.

Like links, shares change overtime (peak and troughs), and timeliness will appeal to search engines when ranking websites. Extremely applicable in the news world, Query Deserves Freshness (QDF) springs to mind – links are nowhere to be seen in the ranking process, they’re too slow to be discovered. Shares on the other hand can be requested at the end of an API. It’s no secret that shares have been fuelling Google News for a while now.

There are millions of pages of great content in existence with no backlinks, but it doesn’t mean they won’t rank. The same applies to shares too. Search engines aren’t throwing away their PageRank model they have so long used to ‘work’ today’s web. One could describe social as an additional ‘layer’ to rankings, fuelling the socialites and human nature of using and consuming content. If you like it, don’t leave and return later, tell the search engine, world, your social circle etc.

Search engines would lead us to believe it would be Danny Sullivan’s job as the webmaster to promote his older article to gain exposure once again. He may choose not to, and let discovery do its job for promotion.

Gammed/Spammed

Eric didn’t feature gamming/spamming social signals as a topic, but he did touch on the topic, suggesting it’s not what bothers him. However, while I have my writing gloves on, I thought it was worth covering.

In comparison to links, I see social signals of as equal opportunity for gamming/spamming.

  • Links – buy them, create link networks etc.
  • Social – buy profiles, create profile networks etc.

As a search engine, how would I learn from gamming/spamming links, all the money I have invested in Search Quality, and apply to social signals? Quality and Trust. From links, search engines will have learned their lesson on how people can spam signals. Search engines hint that not many links pass much value today. Search engines have been forced to apply stringent criteria to websites before passing link value.

That applies equally to social signals too. One must gain, share and act in a social manner that is credible. Once having done so, they’ll gain social appreciation in the form of quality and trust. Therefore, assuming that creating mass profile networks etc. makes social search easier to spam is a false assumption.

Google could happily and easily place more trust in Google Profiles associated with a unique telephone number than accounts with no telephone number. Equally, Google could consider the presence of 10 apparently unique accounts each connected to the same telephone number as a negative quality signal.

Putting Away the Hat

This leads me on to the second big decision search engines had to take with social search, and part of Eric’s argument “It’s that people can be part of each other’s social circles yet have very little in common”. I often need to remind myself that outside ‘the world of search’, it’s users that make the searches, browse and click results that make the world go round. Those users don’t have websites. The common Joe cannot easily contribute to the link graph, but can easily affect the landscape with social search. There are more Facebook, Twitter and social profiles in general combined than there are personal websites. What does that tell you about signals? It’s about empowering the ‘Average Joe’ to make decisions.

On occasion, yes there will be shares that don’t have value to you. Don’t we have that same problem now with links? Do I trust ‘Random George’ and his review?

I believe search engines have solutions in place to assess the cause and effect of social signals. I would like to think search engines are data mining the quality of their results constantly, looking at impressions for shares vs. clicks, and not forgetting, interaction via the social network. I think search engines are confident that they will find social shares coming up trumps, and better integrate them in ranking results.

Who Are You Backing?

It’s a question search marketers are starting to ask themselves, what signals do I want to focus on? I read an interesting article on innovation at HBR, quote:

Suppose you were alive in 1884 and were approached by an aspiring entrepreneur who had developed the most efficient and durable horse carriage ever created. Would you invest in this company?
Hopefully not, since one year later, another inventor by the name of Karl Benz would patent what is now considered the first automobile. Within 20 years, the horse and carriage industry would be under assault from the automobile. And within another 15, the nascent automobile industry found itself rocked by yet another innovation: Henry Ford’s system of mass production, which obliterated hundreds of competitors who could not produce cars as quickly or cheaply.

Ignoring the investment aspect and focusing on links vs. social, you could compare links to horse carriages and social signals to the automobile. Would it be a good decision to invest all resources in links when social signals have the potential to replace links in an ‘assault’? Hypothetically, that could happen. Who knows what search engines are planning next?

Finally, I would like to leave you with the thought that it is Internet protocol to link and human nature to want to share, whether it be because something was funny, informative or offered you an enjoyable experience, or you want to gain attention, be a pioneer, help, reach out or be neighbourly – it’s what makes us human.

I would also like to thank Eric Ward for taking his time out to contribute to Search Engine Land and help spread the knowledge and opinion that makes us digital markers who we are – Henry Fords. Hopefully we’ll write a collaborative post sometime.

Tags: , ,

  • Eric

    Michael- a perceptive dissection of my SEL post, and I commend you for it. In some ways, I was just having a little fun, poking at wart removal creme and Yanni in the context of personalized search. I completely agree that the algos are able to discern which signals should impact results. Using an old school metaphor, I see the social algo working like an old graphic equalizer, with sliders set differently for each searcher. A little Twitter for me, a lot for you, none for my mom. Lots of facebook for some, less for others, etc. Imagine each slider represents a potential social signal. There could be 10 of them today, 25 in a couple years. The key is IMO, there cannot be a universal social algo. This is where the challenge begins. The guy I made fun of in my column because I don’t share his musical taste is a smart financial wiz. I’d definately want to see his picks in my results for that topic. So this adds still another layer of complexity. It’s all good, though, and I was only having a little fun. You’re not related to Yanni are you? :)

  • Pingback: How To Help Clients Understand Social Link Outcomes

  • TMeck

    I’m His engineer

  • http://twitter.com/Tambourinos Michael Thomson

    Hey @9b60a8d06de13279bc8cb4b33cb57f21:disqus just back from holiday hence late reply. Yeah, I did find your examples funny. I think your metaphor is great for visualisation. I just read your follow up post with the graphic. It sheds a lot more of your opinion on the subject.

    p.s. fingers crossed we both never have to work on a wart based website ;) the irony after writing that post would be brilliant!