More reasons for SEOs to learn about IR
The recent buzz around the SEO space is the (poorly termed) Google brand algo which has been more than a little strange. As usual the blog regurgitators are talking in circles, donning tin foil hats and otherwise missing the mark as far as what is happening (some wild theories out there).
Now, I am (obviously) not disputing the recent changes that took place as even Matt, in his usual way, stated there are adaptations that could be producing such an effect (reminds me of the sandbox). That doesnt bother me. And to be honest, Aarons assessment is fine as well, though a little over-the-top, tin foil hat flavouring for me. He terms it as an algo update; which works fine.
My bitch, (as always) is how some SEOs are now talking about the brand algo and theories on what is causing the effect (one person even asked if bounce rates were the cause *#$&@ ) which does little to make us look like we know much about search engines beyond how to spell the words. I have already seen more than a few new Google brand algo type posts popping up around the blogosphere. Beyond those making up theories as they go, others are using the evil empire corporate agenda angle...also flawed. Please people
stop that would ya?
Its not rocket science its computer science
To think that Google (or any search engine) would be crafting a specific algo to pump up brand names is just too fantastical and smacks of PR not scientific SEO. What is more likely is a combination of other existing elements that produced said (perceived) end effect.
As I mentioned on Aarons post;
For me though, any type of 'brand specific algo' isn't really viable. My first instincts are that, if indeed it's happening, that it could be related to harmonic/trust rank type of approaches. New algos mean more processing and greater expense; the search quality team needs to get by the boardroom. To that end it also makes sense that existing systems (such as mentioned above) could be tweaked to attain the desired end result.
This is an important distinction; there is a difference with new tweaks that produce an effect and actually creating a brand centric algos. It is all encompassing in search engine intent. The intent isnt as obvious in many cases.
Can you say brand bias?
Search engines are all about trying to present the best results possible and as such QA testing may have shown that users were concerned why certain brands werent visibly present in certain query spaces and thus they felt the need to tweak things as to fix this problem. Write an entirely NEW algo for this though? Not generally the most effective/economical way to go about things.
Let me tell you a little story (the parts I can discuss at least);
Back in 2006 I came across a few VERY LARGE web entities (owned by the same corporation) that were doing some fairly dodgy keyword injection schemes. Considering the stature of these entities I was kind of shocked to see this. Being the good little web citizen that I am, went off to talk to my Googler friend to bring this to light.
A few days later, he said they spoke to them about it and that theyre essentially caught between a rock and a hard place. People expect to see certain companies/websites showing up in the results for certain query spaces. If they ban such netzians, beyond potential legal ramifications, there is a huge quality control issue here.
Yes my friends, I have seen brand bias at work. But that doesnt mean this is all cloak and dagger and Google is somehow taking actions based on some corporate agenda. As troubling as it may be to the tin foil hat crowd, sometimes, just sometimes, they are actually acting in a way they believe best serves the end user
shocking right? Sigh
Custom algos in tough times?
Thats the next problem; introducing non-native elements means more complexity. Each time you introduce distinct algos into the mix, it makes the entire system more complex and unwieldy. It also increases processing and storage needs, which means $$$. Why are we so inclined to believe weve found a new shiny object when the obvious answers could be right before us?
On the topic of the 300+ tweaks per year (TPY?) once computer scientist said to me, while discussing all of this;
] you simply don't tweak things in a complex system without some sort of testing, analysis and results you can't analyse something in flux so you can't tweak something and then tweak something else before analysing the first tweak. The second is Google is a large organisation with a QA dept a host of engineers, testers etc can you imagine the paperwork that lot must be generating? What Is suspect they meant is they are constantly testing tweaks away from the main search some might be live via dedicated data centre but most will be offline. Once they have passed it through the various QA processes they decide if to implement it across the board immediately or as part of an update.
This becomes exponentially more problematic if we were to incorporate an entirely NEW algorithm into the mix, not merely changing weights/parameters on existing ones.
He didnt just say behavioural signals? oh no he dint how that worked its way into this discussion is beyond me
seriously. Aaron even puts out (albeit briefly) a hint that implicit user feedback may be at play here. My goodiness, methinks weve beaten that horse well and truly to death here on the trail simply not viable (yet) beyond personalization and query analysis. It wouldnt be much use for brand bias type improvements.
It is highly unlikely that this is a new brand algo, not beyond the realm of possibilities, but just not likely. Id prefer to call it a brand bias at this point, there is certainly a difference.
An industry afloat on the river DUH
And why do I care? Why do I bitch? It is because, much like bounce rates, the masses of SEOs that follow along blindly will now run with this like the beasts of legend such as ToolBar PageRank, behavioural metrics and to a lesser extent the magical sandbox. All given value and grace without a lot of foresight; never mind research from SEOs into cause and effect.
All love and respect to Aaron, but when I read things from peeps on Sphinn like;
I'm sure Aaron Wall's kind of insight into the dynamics of the search market will always teach us more about what Google does today (and why) than the titles of a couple of scientific papers.
I just have to speak up.
Excuse me?? So Im right about SEOs?
fuck research and computer science.. just read some SEO blogs and yer golden?? Holy shit
no wonder the industry isnt respected. My favourite part is how he says Aaron is giving us the WHY
seriously, did I miss something? Sure, maybe I gloss over some when Aaron goes after Google (with the tin foil hat crowd) but I am sure hes not glued to his screen with all of my posts either. I certainly did not find my why or how in that post, nor elsewhere since.
Do I know what is going on? No. But I am equally as certain much of the assertions in Aarons discovery arent the answer either. I am equally convinced some of the ideas floated out in the SEO spaces are, for the most part, frightening in their silliness (yes, I even heard one person suggest bounce rates sigh). This is what angers me
how people can be band-wagoneers that would rather take a blog post as gospel and, ignorance be damned, shout from the mountain top that what they have read is now the truth.
I turned 43 yrs old today
I am far too old to believe ANYTHING from a mere excursion over the surface. This is no different.
Building a brand bias
so what could be at play here? After some cursory digging and reading anecdotal evidence, the following come to mind;
- TrustRank/HarmonicRank right away these are the ones that leap to mind. This was rumoured to have taken hold with the so-called Jagger updates, but as always who knows. We can be reasonably sure some relative method exists because, well, Google loves links. This type of valuation of authority based on the quality of the link graph associated with a given page could certainly be tweaked to give more weight to the stronger authority sites/pages.
- Named Entities this particular area also came to mind initially with the brand bias. Essentially these algorithms associate specific people, places or things with a given webpage/site. These methods are used in the main index as well as blog search and might be able to give weight to brands that are synonymous with a given query space.
- Personalized PageRank (topic sensitive) once more, going back to 2004, much of todays anchor text scoring likely came from PPR. Given this, we could also surmise tweaking this to give greater weight to branded links/concepts could give a domain greater global PageRank and have the desired effect were seeing.
- Semantic/phrase based this is also an interesting area as it seeks to develop a better understanding of gleaning concepts and themes from a page/site. If such methods are used, its not the direct targeting of KWs and anchor texts but also related concepts and themes. In theory the probabilistic matrix could make brands that are synonymous more weight than the standard on page and straight link text approaches.
- Query analysis although click-bias can be troublesome, if theyve determined brand affinity is important, additional weight could be given to certain URLs (brands) in popular query spaces. As with many behavioural approaches, click-spam could make this unattractive. (Of note, is Matts comment that, The short answer is that we did change some of our algorithms for some queries)
- Personalized search another long shot would be serving more branded results to users whose profiles show a tendency to brand affinity. From what I have read, this also seems unlikely; but worth testing.
- Domain age this is another authority related signal and while it may be able to play a role, it seems unlikely in all but a supporting role to other elements mentioned here.
Are you starting to get the idea here? It seems to me that what were seeing is rooted in existing methods, more-so than a new term for the lexicon (Googles Brand Algo).
What we need as search optimizers are actionable elements. This means we research from a solid base founded in information retrieval; not conjecture. Lets pool our data and work out the HOW so that we may better understand the latest developments in a context which is useful not just fantastical and newsworthy.
Aaron has set the ball in motion, it is up to the rest of us to continue the momentum and dig into what is actually happening. And please, its not a new brand algo ok? Seriously, just dont