Somehow, my flight from Seattle landed just before the newest Icelandic ash cloud began shutting down airports across UK airspace. As a result, I was able to present at SMX London this morning. The presentation is included below.
This slide deck focuses on things that are probably in the search engines' ranking algorithm today (e.g. the reasonable surfer model), might be in today and probably will be more in the future (e.g. tweet data) and may or may not get in (e.g. Facebook's open graph).
SEO Ranking Factors 2010 SMX London
Everyone at SMX noted that this year's event is much busier and more active than last year. Goodbye recession; hello London as a new center of the search world.
Looking forward to your comments as always!
p.s. I'm planning to write a much more comprehensive post about the "reasonable surfer" patent, but in the meantime, be sure to read Bill Slawski's analysis.
Thanks Rand, have to be honest and say I was largely disappointed at the quality of presentations and information on offer at SMX London.
I didn't think much "advanced" information was provided a lot of the time I was just thinking "well yeah....we know this, that's obvious".
I came away with 1 A5 page of paper half filled, which says it all really.
I don't know whether this is because actually I am doing a good job personally of keeping on top of this stuff, or the presentations were simply of low quality this year.
I know I'm late to this thread and few will read this, but I couldn't agree more. SMX London Advanced was frankly a waste of money with very limited takeaways in terms of learning valuable advanced tips.
In a way I guess Ishould feel comforted, maybe I do know a lot more than what I thought if the quality of the presentations is anything to go by. One or two good presenters outweighed by numerous medicore/poor ones. I'm glad it was my company and not me personally paying for my attendance.
Many of the presenters are clearly in the game of self-promotion rather than educating their peers via sharing their documented research and findings.
Good post Rand and good presentation yesterday.
To everyone who was disappointed by missing the SMX event yesterday - I wouldn't be too disappointed, I went with a colleague and we were "underwhelmed". As a Paid-For event I was expecting a lot of new techniques/up to date information - we didn't get it.
Rand's presentation was good, but I might as well have saved the money and checked the slides here. Sam Crocker from Distilled also gave a good presentation, as did Rob Kerry from Ayima also came across very well and had some interesting thoughts on external 301 re-directs not passing link juice any longer.
Apart from them, the level of SEO knowledge/info was very, very basic. Good if you're interested in networking, poor if you're looking for SEO tips.
Cheers for the slides Rand.
All of the events I have recently been to in London have been very themed, last one was all paid search, the one before was all soical media, the one before that was all content management systems.
What was the theme of SMX? was there one?
Great presentation and a real pleasure to actually meet you.
Hi Rand! Could you explain slide 20 (H1s Did they ever matter..), slide 21 (alt attributes:..), slide 22 (KW density: correlation...) and LDA (latent dirichlet Allocation).
I'll try to explain briefly, but it may be worth a deep dive post at some point.
H1s - basically, back in 2004/5, lots of SEOs who used nonsense word testing said that H1s were barely useful and the engines didn't really care about them. Many other folks in the industry, myself included, ignored these data points, didn't do our own testing, and continued recommending them to clients. When we finally did do testing, and correlation studies and put them into our ranking models, we found that H1s had almost negligble value. It seems to be good to put your keywords at the top of your content into a "headline" type area, but using the actual <H1> as the wrapper provides minimal value. Thus, I question whether I was wrong all along and those early testers were correct. Maybe H1s have never mattered much at all and it was just an SEO myth of association. Note that recently, Covario's research (cited by Stephan Spencer in his posts), confirmed this same result in their research.
Alt attributes - these seem to be slightly more useful for rankings than H1s, within the margin of error. Plus, they're quite good for image search SEO. Hence, it might be more worthwhile to recommend keyword use in the alt tag of a relevant image than in an H1 tag.
KW Density - technically, it correlates with rankings, but when you put it in the ranking model for causation, it appears useless. The takeaway is - it's good to use target keywords in your content, but KW density (the formula of "total number of words / target keyword instances") is a terrible way to measure.
LDA - I'll leave that for another time; essentially, it's a way we hope can produce some good recommendations about additional terms/words to consider when targeting a particular keyword. If you're into vector space modeling of terms/phrases to do topic detection, LDA is part of that sphere (which also includes LSI, LSA, pLSI, etc.)
Rand, can you define "but when you put it in the ranking model for causation"? How do you perform extended tests with the ranking model - Just being curious...!
We know that correlation != causation, and thus we want to find ways to determine if a signal is merely "correlated" with high rankings, or if it actually directly impacts those rankings.
To do this, we use machine learning against search results, building a model that represents, to the best of our ability, the rankings in a fashion that has good accuracy with the engines' orderings. In this case, we've used only Google.com, and our accuracy levels hover in the 70% range. However, by applying even these crude models, we can see a big difference between those things that merely "correlated" vs. those that, in the absence of other signals, appear to influence (and thus are likely "causation").
Thanks for this in-depth explanation!
Nice presentation Rand, thanks for sharing!
As someone wo always wants to get the most out of theories etc, I am really curious about your vision/ opinion on LDA! I've been taken some reads into LSI and LSA so filling in the sphere in the future would be great!
Thanks for the info Rand. Do you have any test results about the H2 tag. It's just a theory but I'm suspicious that Google will decrease the value of H1 tag because every webmaster is abusing it nowadays, so they will increase the value of H2 tag instead. Is this possible?
Thanks Rand. I had the same doubt with that Keyword density correlation. That explanation was fine to understand.
Great Stuff Rand! But I think you missed to write about SITE SPEED as one of the factors for ranking.
I had the same question - why did you omit site speed from this preso, Rand? Do you think it is not as important in the ranking factors as the SEs (Google especially) have led us to believe?
I believe Google said that site speed was now a ranking factor in "less than 1% of queries" and that in those cases, it was one of 200+ factors. Site speed has always been very important, bt this announcement, IMO, did not make it more important, just made more people aware of it.
Site speed is important because faster loading pages means lower abandonment which yields more sharing, more linking, more social activity, more clicks, more traffic, more purchases, etc. Having your rankings possibly improve slightly on "less than 1% of queries" is not a compelling reason to focus on it - all that other stuff definitely is!
Rand,
Do you have the audio or your comments on those slides?
Best,
Marcelo
I don't have any audio I can upload, but if you have questions about any of the slides, I can likely add some extra value here in the comments!
Thanks for sharing.
Do you wonder if Google adjusts their algo just to spite SEO recommendations? I bet they would do something like that just because they can.
What do you think of the Golden Triangle now that Google has changed more on their sidebar? Seems like these days you can't settle for one, two, or even three main keywords for your site anymore.
Nice to know that I shouldn't spend time worrying about header tags anymore. Thanks Rand, this stuff is really useful and time saving. Keep it coming.
@arek-suroboyo it would be great if Ask picked up market share. After the big ol' merger and Google's 500+ changes in under 365 days, absense of Yahoo, and Bing's crowded SERPs, maybe people will prefer a straight up search engine and less universal search.. ya never know. If ANYONE knows what people will do before they do it, it's Google.
Eminent stuff, Rand. I'll return to this more than once. Just plugged this post in SEOchat. Questions could go on and on, but I am awaiting with bated breath your eventual in-depth posts on the relation of social media to Google search results, which are yet to come. As you know, many are skeptical that there's evidence for any impact (well, with the exception of real time search, obviously). I am perfectly open minded, but I too would like to know what the correlations are and how they were found out about it. I think this might be the big new thing in search, if it's true. Convert us, make us believe! :)
Thank you as always.
Hi Rand,Amazing presentation. I wish there was audio with it and that would have really helped a lot.Regarding Facebook and twitter, does this mean that from now on, we should have more presence on them (have more Facebook pages/groups and twitter lists/fans) to increase our rankings? Thanks again,Abhinav
Rand this is by far one of the posts yet (not sure how I missed it). I sure wish there was audio to this as it would clarify a lot of the questions I have. lol maybe an incentive to us pro members to have audio the slides?
good factor. i use this for my competition in seo thailand.
Thanks for that neat presentation Rand. Most of them are self-explanatory and understandable. Still I feel of missing the audio session. Hope I would attend one of the events live next time.
thanks :)
Thanks for sharing the deck Rand. I definitely enjoyed the presentation, only wish you'd had more than 12 minutes to whip through this stuff! I found the Twitter influencing search beyond QDF to be the most interesting insight and can't wait to see some more testing done around this :)
Hi Rand, maybe this is a silly question, but I was wondering what you mean by slide 11 "Are they treated like links?". How do you mean that the tweets are treated like links, and links to what? Thanks for a great presentation :-)
I'm not surprised at all about the "links in content are more valuable." This plays right into the contextual aspect of the web. I believe intrasite links in the main content of any page are one of the most underused aspects of any SEO technique.
I'm curious to run some tests on the confirmed slide 5, I'm also intersted to see the data behind the idea.
Hi Rand, could you please talk more about slides 16 and 17?
Thank you in advance.
Interesting. I've done only one experiment recently, regarding alts. But I have found that for a few keywords, have really bumped my rankings from 5th or 6th position to 1st or 2nd. Although this has only happened on medium tail keywords and the experiments certainly were not 100% valid. I didn't account for many exogenous factors. But with your data, I'm thinking of actually putting an alt tag experiment together.
On slide 5 do you mean that at the higher position links appear, the more value they have even if they are pointing different URL from each other?
Rand, do you think that eventually Twitter will become a "realtime search engine"?
Thanks for posting the slides for us Rand, I was absolutely gutted that I wasn't able to make it to SMX - maybe next year, eh?
Thanks for sharing the slides on SEO Ranking Factors 2010 SMX London.
I think after blogging only Twitter is the microblogging platform which shall have an impact on the search scenario. As Twitter appeals more to the intellect and aims at Less Is More by making the person say more using less words.
The real scenario of search shall will be unveiled only when all data centers of Google shall be caffiene enabled . That should be sometime in mid 2010 latest by August 2010 ( My assumption) as the last Major Google update which was in 2005 was truly reflected by August 2005.
SEO becomes more and more interesting and challenging with every Google update.
This has been amplified too with the new "Google Updates" thing.
Good stuff Rand.
Hope your liking your trip to London.
I missed the conference this year ! have been bogged down, did web seminar record it and will they show it online?
Regarding the actual content, its intense but so much learning also seeing how things are moving towards the social sphere is amazing, but will Google, Bing and Yahoo be able to keep the right balance in order to actually deliver relevant results rather than social information ?
Will be interesting to see.
Security Error Loading Document Info:'Error #2048'!
All their base, are belong to us unfortunately :-(.
Clear cache & refresh. Everything is working fine for me!! if all else fails, reinstall Adobe rubbish Flash.
Congratulations on beating the ash cloud. It's becoming quite the extreme sport in Europe I understand.
Thanks loads for sharing the slide deck with those of us that, to quote trax "we're gutted" by being unable to attend.
Going through the slides without audio reminded me of watching movies without sound. You can guess at half of it, but there are some things that are next to impossible to decipher.
Beggars can't be choosers, but in a perfect world, I'd love to have heard the audio too.
Seconded!
I think SMX is a really strange Exposition. I kind of expect them to stream certain events but I can't find any live streams of SMX whatsoever. Microblogging and live blogging can only give you so much information.
Besides, I'm gutted that I missed this:
"hahaha @robkerry says "you need to be Billy Big Bollocks with G News editorial rejections" I'll explain more in my blog post ;) #smx"
Thanks God Rand gave the link to the post of Bill Slawski about "reasonable surfer" (I've to admit it's the 1st time I hear this definition). By the way, even without audio the slides were quite understandable.
I'm thankful that Rand posted the link and mentioned the post at SMX as well. Thank you, Rand.
As far as I can tell, the reasonable surfer concept was introduced in the Google patent that I wrote about, though many of the ideas that it discusses have been introduced in other ways.
For instance, Google, Microsoft, and Yahoo have all published either patent filings or white papers or both that discuss how different sections of a page may be segmented based upon a combination of reading the DOM model of the page and analyzing visual gaps (white spaces, horizontal rules, etc) that appear upon a page, to give those different sections differing amounts of weight.
Microsoft's paper on Block Level Link Analysis explores how a page can be broken down into semantic blocks, and a link analysis of the Web can be done upon a block level rather than a page level, and the paper describes both the idea of a block level PageRank and a block level HITS algorithm.
Even in one of the orginal papers that discussed PageRank, "The PageRank Citation Ranking: Bringing Order to the Web," some potential problems were raised, such as copyright warnings and disclaimers receiving overly high rankings based upon links pointing to them.
The newly published Google patent mentions the same type of problem with "terms of service" pages, and addresses that and other issues with the idea of providing a number of examples of features involving links, the source page of the link, and the link target page, as well as user behavior data as factors to be considered together to determine which link or links are most likely to be followed by a reasonable surfer.
It might be tempting to draw a conclusion such as a link appearing higher on a page, within the main content area of a page would carry more link weight then one lower down on the page, but that simplifies the concept of a reasonable surfer too much.
The patent tells us that the search engine would consider a wide range of factors together as a whole to determine which links would carry the most weight, including things such as how related the anchor text within a link might be to the content on the page, how commercial the text within that link is, and much more.
If Google is using this approach of a reasonable surfer, and it's quite possible that they are to some degree, then it has a number of interesting implications in many areas. For instance, people considering buying links would have to more seriously question the value that they might be getting from a link. A link from a page showing a toolbar PageRank 2 might pass along more weight than a link from a page showing a toolbar PageRank of 7 based upon the reasonable surfer features.
Nothing to add but thank you Bill
Thank you for sharing the presentation Rand,
As far as facebook building their own search engine, I think you mentioned it earlier on a WBF where you may be searching for a particular item (ex. Bird watching guide) and none of your friends would have any "likes" since they probably have similar taste and likes and that model would not be beneficial.
They would have to take that into consideration and that maybe where Bing fits in, who knows what will happen and like all proficient SEO'rs we learn, understand and adapt.
Shawn
thx for the link to the "Google's Reasonable Surfer" model. Never heard of this model before or maybe it goes by a different name?
Really neat presentation, can't wait till I actually get to go to one of these events!
While I'm waiting about your post about the Reasonable Surfer issue, can you confirm me what I've understood from the slides that the concept has somehow having to do with the behavioural use of the web?
About the influence of Social Graph... actually I've not a clear opinion about it, even though I recognize myself in what said in this interesting article I read yesterday on the NYTimes
I'm curious to see the evolution of the Bing collaboration with Facebook. If Bing is going to rely very much on the Social Graph in its SERPs maybe we are going to have two different kind of Search Engine to work with... but, honestly, I believe that from a user point of view would be moreeasy to make a search directly inside of Facebook.
Be great for users and SEOs if Bing grabbed more market share - Google is too dominant and at times compromises all those things that users liked in the first place.
I agree with you... Seeing an emergent SE would be something interesting for everyone...!
SEO is all about AUTHORITY eh! Ask Mr. Milgram about it... Or those frenchies who recently made a remake of his experiment! :)
Guess I just found my next blog post...
The backlinks should contain variations. If you always contain the same text in yours backlinks this is not natural.
PaidLinks should be included among their own menus or content of the website. It will be more difficult to buy links
Can you give more insight into the orders of the links mattering? How much does this matter? Enough that we should care and make a strong effort, or is such an infentesimal percentage of the algorithim that we have better things to do, like make banana bread?