26 Mar 2014

Google’s Algo Is Improving … But How Smart Was It In The First Place?

Some friends a few years ago were pushing a piece of linkbait on Digg, whose title was something like “They Threw The Book – The USA’s Harshest Sentencing Judges Ever.” You would imagine that their client was someone in the legal field, right? Actually, the important keyword in that title is “book” and the client wasn’t even a bookseller, but a travel agency looking to rank for booking keywords. (Actually, I changed the industry and target keyword, but the example accurately represents the exploit.)

Why did this work?

This worked because Google then works like today’s Google Translate – understanding each word in a query independent of its context.

So that same piece of linkbait could also have been for a dvd series about “throwing” the javelin farther in your next track and field meet. (To Google’s credit they did understand even then that ‘throw’ and ‘threw’ and ‘throwing’ were related.)

In a fascinating article about the knowledge graph, AJ Kohn highlights that Google’s moved past that.

Context matters, now.

Google understands “entities” now, making the above exploit outdated.

Another fascinating tactical point he makes is that it’s important to link out (hello newspapers) to related entities. (He shares a lot of other useful tactics, in fact, which I strongly encourage you to go read.)

They understand that a query for Zillow + location is often a search for “real estate suppliers + location.”

As a side note, the fascinating implication is that this devalues brands in a way that I don’t think I’ve seen anyone talk about. Competitors don’t need to bid on your brand in AdWords, now. Google will suggest them on its own. It’s possible that Google feels it’s gone a bit too far down the Vince update path and is looking to balance things a bit by helping small businesses.

An alternate explanation may be that this is just another explanation of Google devaluing a competing brand (for the sake of good user experience 😉 ). Zillow is competing with Google by fighting for real estate ad budgets, and are likely to spend 10s of millions on ads to that sector, this year. Google also got into the travel search space, harming Kayak and other aggregators that were selling ads to online travel agencies and airlines. It remains to be seen if Google brand attack, say, Mercedes by showing Aston-Martin ads.

</end side note>

The French have a great expression about relativity:

“In the kingdom of the blind, the one-eyed man is king.” (Sorry, content and links.)

Applying this expression to Google’s algorithm, it’s emerging from the land of the legally blind to having one eye. But as most of us know, that’s not saying much. If you scratch the surface, I’m sure there are clever minds that have moved on to exploiting the next loophole.

And before you chime in that any exploit is a short term thing, the exploit I’m talking about dates to 2007-2008ish. Schema came out in 2011. That’s 3-4 years of catching up, just to launch the product. We’re in 2014 and I daresay that outside a few industries, Schema has relatively low adoption and that entity understanding at engines is only in its infancy. So my friends’ exploit was based on a loophole that lasted at least 3 years (2008-2011), more realistically 2007-2013 (6 years) and perhaps longer since 2007 is only when I became aware of it… for all I know this loophole existed since Google started in 1999.

So yes, the Googlers on the search quality team are certainly smart, hard-working folks and they’re making big strides … but it takes time to make big strides. And there’s still lots of room to improve.

My real point here is that many in the SEO community often think that theory matches reality. The claim is that since Google gathered lots of intelligent guys together, Google has figured it all out. Just like some people claimed that Google had it all figured out in 2007…

No, the algorithm isn’t perfect, not even close. It’s gotten better… but it had big problems in the first place, so what does improvement mean, really!


Leave a Reply