For going on three years now, I’ve an evangelist for the use of social graph signals over older link-based authority algorithms such as PageRank.
Last week it was confirmed that Google is now in fact using social signals such as Twitter Retweets and Facebook Likes as ‘votes’ in their new ranking algorithms. The details were highlighted in a recently uncovered USTPO patent filing: U.S. Patent # 7,765,209
It’s an exciting time to be in Search. The voting-block responsible for collaboratively filtering the web just shifted; away from webmasters and links – and towards users and their social activities.
A Case Study on Social Media Publishing & Analysis
Marshall Clark – Head of Search, Organic
As a follow-up to my earlier ‘Futurecasting Social Search’ presentation at the 2008 Search Insider Summit; in Summer 2009 I wrote an article for the Organic blog ‘ThreeMinds’ titled ‘Docs are Old School, Why We Need PageRank for People’.
The central idea was that Google’s PageRank algorithm was becoming obsolete by continuing to assign reputation to web documents/URLs at a time when the web is moving to syndication and distributed content. The article also outlined a novel solution to the problem by assigning reputation to content authors via a system integrating Social Graph data, OpenID, and existing search ranking strategies.
Following publication I launched a campaign promoting the article through a variety of social media channels – the following case study documents results and learnings from this campaign.
The promotion campaign for the ‘PageRank for People’ article was a success. Shortly after publication on ThreeMinds the article went viral on Twitter, spreading rapidly through the search and technology communities. The article was commented and retweeted by several tech luminaries including:
- Dave Winer – Creator of RSS
- Tim O’Reilly – Founder of O’Reilly Media & originator of term “Web 2.0”
- John Battelle – Author of the book “The Search”
- Leo Laporte – Founder of TWiT.tv Network & popular tech commentator
Analysis of traffic patterns showed a clear pattern of trickle-down content distribution/syndication originating from each of these Twitter Influencers.
ThreeMinds is one of the most highly visited blogs in the advertising agency world. Despite an already large volume of baseline traffic, the ‘PageRank for People’ article generated a huge boost in traffic to the ThreeMinds site, increasing total site traffic by more than 60% during one two-week period. The graph below shows total traffic to ThreeMinds for the last year with the two spikes on the right illustrating PRP traffic. Additional reports show the article was the #1 viewed article for the last year and one of the most popular posts ever published on ThreeMinds.
Traffic sources for the article were unusual as nearly all of the traffic for the post came direct from Social Media (85%+). ThreeMinds posts typically get a fairly even distribution between Direct, Referral, Search, & Social Media sources.
In preparation for promoting the article, I talked with several top Twitter and Social Media experts to learn how to best increase the chances of having my post picked up.
Some of the things I learned:
1. Make sure your article has a name that is both catchy and descriptive. Catchy to grab people’s attention, and descriptive to let them know what they’ll get when they click-through.
2. Build a sound bite into your title, a ‘meme’, something that people can borrow, repurpose, and pass along but will still connect back to your original content. Ex: “PageRank for People”.
3. For Twitter, keep your original tweet well below the 140-character limit. This encourages users to retweet your content, since ‘Retweeters’ (yes – seriously :/ ) can add their own comments and still be below the character limit.
4. Use Bit.ly to shorten your article URL. This allows additional room for content and Bit.ly also provides some great analytics tools for tracking traffic to your post.
Other things I found as the project progressed:
1. When people retweet your content, the first thing most do is snip out your Bit.ly link and replace it with their own shortened URL link. This link still points to your same destination URL, so you still get the traffic. I’m guessing they do this to track how much traffic they send. In many cases they replaced my original Bit.ly link with their own Bit.ly link. – Fortunately Bit.ly allows you to view aggregate data for all Bit.ly links pointing to your destination URL, whether you own them or not. Smart.
2. On Twitter, Influencers are everything. Several times during the project I thought I had reached a traffic plateau, only to have a new Influencer retweet the link and deliver a new, higher traffic spike. What’s more, these traffic spikes and the prominence of the Twitterers responsible for them, tended to increase over time. This seemed to point to a hierarchical web of connections, one where influential people follow other, more influential people. Seems like with a good connection mapping program and a database of related topics, this could become something very useful for PR firms.
3. Your content will live everywhere, including some places you wish it didn’t. Several bloggers ‘borrowed’ the content of the article as the basis for a blog post, many giving little to no attribution. While there’s not much that you can do to stop this, I found that friendly comment on the blog, building on their ideas and linking back to the original article was usually permitted. In one case this even led to a productive conversation on how my original idea could be improved.
1. Twitter has great potential as a PR/Marketing tool, provided you can get Influencers to pick up your story.
2. Your content will be mashed-up multiple ways and spread across the web. Make sure your core concepts are tightly connected and contain at least one ‘meme-like’ component so you keep credit for your concept. See Meme-tracking and the Dynamics of the News Cycle for an analytical perspective on why this is important.
3. Keep your concepts short and memorable. This encourages distribution and commentary while reducing the effects of dilution, misinterpretation, and short attention spans.
If you have questions on anything presented in this case study please feel free to ping me on Twitter. I’m always happy to discuss.
Head of Search | Organic, Inc. | 555 Market St., 4th Floor | San Francisco, CA 94105
Here’s a copy of the deck I presented this week at SES Chicago on ‘PageRank for People and Distributed Reputation Systems’. Feel free to Comment or Tweet any questions. Links to the full version including my speaking points is at the bottom of the post.
Docs are Old-School
I’m going to let you in on the Search industry’s dirty secret.
Google is slipping.
Google’s big innovation was in realizing that a link to content is the same as a vote. By tracking all the links pointing to a page of content Google assesses how influential that page is – its reputation. Google calls this ‘PageRank’ and it’s old tech.
PageRank assigns a reputation score to the URL where content is published. This makes it a great fit for content that stays put in one location. However, evolving content distribution via blogs, RSS, guest columns, and syndication are a challenge for PageRank. Micropublishing, Tweets, Retweets, ratings, and comments – even bigger problems. The solution lies in associating reputation with the identity of the author – a PageRank for People.
Reputation is Personal
At issue is how Google attributes reputation.
If marketing guru Seth Godin publishes an article on NYT.com, marketing wonks want to read it. If he publishes it instead at PodunkMarketingBlog.net, they still want to read it, because hey – it’s Seth Godin. Google would rank the article at NYT highly, but Seth’s work would be next to invisible when published at Podunk.
We assign reputation to people; experts, advisors, consultants, coaches, gurus, friends, etc. Search engines to date have relied on some proxy for this real-world reputation.
Content Lives Everywhere
In the physical world, your reputation follows you. If you’re the world’s foremost expert on AJAX, your opinion on the topic will be respected wherever you go. Imagine if the same held true online. Publish an article on an obscure web dev blog, it ranks highly, because hey – you’re an expert. Pen a guest column on “AJAX and You” for Women’s Day magazine and it ranks great, because you’re the best in your field. Post a comment on the blog of an up-and-coming developer and that post gets a boost, because one of the luminaries in the industry judged it worth weighing in on. These are just a few of the possibilities, I’m sure there are plenty more.
Mapping reputation to people instead of URLs makes PageRank portable. It’s PageRank for people.
PageRank for People
What we’re talking about here is a fundamental shift in how we track and evaluate content online. One that mirrors how we assess information in the physical world: “Who’s this guy talking? What’s he done before? Do I like him? Do people I like, like him?”
Some of the criteria the PageRank for People algorithm would look at:
• Topics of expertise
• Number of links to your posts and the reputation of those doing the linking
• Number of comments and the commenters’ reputations
• Who’s in your community (those linking, commenting, reposting your content frequently)
• Ratings, weighted by the reputation of each rater
• Where you’ve been published
• How long you’ve been online
• Maybe even semantic analysis of your content (are you a troll?)
What About Privacy?
“Ok – so we’ve got a new search engine that knows who you are and on the topics you’re an expert on. What about privacy?” The key thing here is that the search engine doesn’t need to actually know who you are, only that you’re the same identity it has been tracking across the web. The New York Times recently published an article “A Service to Prove You Are Really You” describing a new Equifax I-Card identity verification service that could provide verifiable, anonymous ID on the web. Imagine the ability to use one or more aliases online, all of which track back to an I-card capable of verifying these aliases while keeping your real identity anonymous. OpenID with modifications might also provide the necessary anonymity + verification needed to make the system work.
The Future is Social
In the physical world we’ve got couple millennia worthy of debugging invested in how we evaluate information and people in real life. Moving these same systems online will deliver huge gains in the accuracy and trustworthiness of information and will help align Search with the increasingly social aspects of the Web. As associated technologies improve, I’m sure we’ll see some of the ideas discussed here implemented in one form or another.
Here’s a republish of a POV paper I wrote for Chrysler in June regarding the proposed Yahoo/Google ad partnership.
Yahoo – Google Search Advertising Partnership
In early June 2008, Yahoo received and rejected an unsolicited buy-out offer by Microsoft of over $47.5B. In the wake of this collapsed deal. Yahoo CEO Jerry Yang and others have proposed a number of alternate proposals in hopes of reviving an ailing Yahoo. Most surprising perhaps is the recent proposed advertising partnership between Yahoo and arch rival Google.
Under the terms of this proposed deal, Yahoo would grant Google a non-exclusive license to sell search advertising on Yahoo search results and web properties in the U.S. and Canada. The value of the deal is estimated at $800M per year and had a maximum term of ten years [(1) four year initial term, followed by (2) three year extensions at Yahoo’s discretion].
Interestingly however, the deal would not be a simple substitution of Google ‘Adwords’ ads for Yahoo’s existing ‘Panama’ ads. Under the terms of the deal, Yahoo would have the option of serving either a Yahoo ad or substituting a Google ad, depending on which advertisement provided them the highest revenue. Yahoo president claims that Google ads would be limited mostly to long-tail results where Google has a deeper advertiser base. Industry insiders however, remain skeptical of the eventual scope of these ad substitutions.
Yahoo executives also claim that the deal would beneficial for all parties involved, since Google historically monetizes their search ads more effectively than Yahoo. A recent JP Morgan analysis seems to support part of these statements, claiming that Google monetizes search queries at a 79% higher rate than Yahoo. This higher rate would be a function of not just Google’s typically higher Cost-per-Click fees for terms, but also their superior ad matching capabilities for contextual ads and content.
What would be the results of this partnership?
From a user perspective, the proposed partnership would have little obvious effect. Advertisements on Yahoo search result pages would likely become more relevant and targeted as Google’s superior performance-scoring and relevancy-matching systems drive out some off-topic and branding-focused advertisements. Most noticeable improvements would likely come from contextual search ads placed on Yahoo web properties. Contextual advertising systems match web page content with appropriate search ads using sophisticated ranking algorithms similar to those used to score natural search results. Google’s contextual placement system is superior to Yahoo’s flawed system and users would likely experience greatly improved ad targeting on Yahoo web properties under the proposed partnership.
Would this be good for advertisers?
This is a complex question. All initial impressions suggest that the partnership will increase search advertising prices and consequently hurt advertisers, however the devil is in the details. Rationally, if this deal is a win-win for both Yahoo and Google, additional revenue must be entering the market at some point – the most obvious source being the pockets of existing Yahoo advertisers. Where things do get interesting however, is that paying higher fees for Yahoo search ads may not necessarily be a negative where existing advertisers are concerned.
One of the benefits of ROI modeling for search advertising is that it reveals hidden relationships between upfront costs and bottom-line revenue. While our seat-of-pants reaction to increased CPC search fees is an expectation of reduced campaign performance, this is not necessarily the case if increases come in conjunction with significantly improved targeting and conversion rate behaviors.
Google has historically commanded a CPC price premium over all other paid search platforms. This stems from a demonstrated ability to deliver superior KPIs even with higher prices factored in. It remains to be seen whether Google can bring these same performance gains to the proposed partnership. Advanced analysis of paid search performance will become even more crucial to existing advertisers as they evaluate the impact of changing CPCs and performance on their existing campaigns.
Is this good for the marketplace?
This is an increasingly murky aspect of the partnership. Google currently controls 50-60% of the search marketplace. However, when we look specifically at the market for paid search advertisments, they are much closer to 70-80% of the market. Many in the industry claim this represents an effective monopoly of search advertising – a situation the Yahoo-Google partnership will only exacerbate.
Microsoft has already issued statements to consumer protection groups claiming that the deal would “limit choice for advertisers” and “destroy a competitive environment”. Microsoft stated that the Yahoo-Google deal would place 90% of the search ad market under Google’s direct control.
Also concerning is the fact that the deal appears to have been developed in conjuction with direct DOJ oversight, leading some in the legal community to suspect that it may have been structured specifically to avoid regulation. David Turetsky of Dewey and LeBoeuf, LLP stated “They’ve had a dialogue with the DOJ…which is unusual. DOJ is not in the business of giving counseling to companies as to what flies.”
Additional concerns center on the impact of the deal on industry-wide search ad pricing. The structure of the ad delivery system may act to fix minimum prices across the search ad marketplace, since Yahoo would never sell an ad for less than what Google would pay them.
At the heart of any antitrust investigation would be the definition of what is Google’s core business. Currently Google derives most of its revenue from search advertisements. However, they are also actively working to expand into television, radio, and outdoor advertising, with only limited success. Central will be regulator’s determination of what defines Google’s core marketplace. Is it limited mostly to search advertising, or is their market online advertising or even the entire advertising industry. Google will try to portray their market as broadly as possible – however their revenue breakdowns will likely suggest otherwise.
How is this likely to affect Chrysler Search?
Google SVP and Adwords originator Omid Kordestani has stated that the deal “will not allow Google to raise prices”; this however is a matter of considered semantics. In Adwords, Google has built a brilliant system to harness market forces for delivering maximum profits while avoiding appearance of undue manipulation. With Adwords Google never has to “raise prices”, they simply adjust the rules of the system to encourage market forces to raise prices for them.
It appears likely that any Yahoo-Google deal will result in increased search advertising fees for Chrysler ads showing on Yahoo. Whether these fees are accompanied by performance improvements sufficient to offset additional costs remains to be seen.
We recommend continuing to watch specifics of this deal for new insights into pricing and performance.