Why Google Is Broken
This week we are showing you the 10 most popular posts of 2014 on State of Digital. The top 10 is based on a combination of reading numbers, shares and comments.
This is number 8, by Gianluca Fiorelli, about how ‘bad’ Google can be.
Originally posted: April 23, 2014, 10:45
No, this post has not being written by Barry Adams, even if I think he will agree with my conclusions. [Yes he does – Ed.]
Although I don’t not always see a second hidden motive in every action that Google does, as I have said many times in the past I try to keep a critical and healthy “conspiranoic” eye with respect to what Google does.
Personally, I consider it is a great search engine. It is far from perfect (nothing is), but it is the best option we have available. Even if it caused pain to us, Google’s attempts to address the spam issue in its SERPs have improved the overall search experience… but that effort was not the same everywhere.
When we talk about Google, we should always distinguish between Google.com and the regional Google versions.
Most of the things we daily talk about and discuss concern only the first, but in many cases do not have a real correspondence, for example, in Google.es or Google.it. And that means that the regional Googles do not offer that quality Google pretends to offer to all its users.
Obviously, these differences between Google.com and its regional versions are annoying, because they prevent us to take advantage of attractive marketing opportunities, and we must wait to be able to finally use them.
But there are other “differences” that make the regional Googles broken, because they affect directly the quality of search experience.
The forgotten EMD update
Let me remind you of the more recent history of Google, and I apologize already if you’ll find it “hasty”.
After Caffeine , the SERPs were inundated with results of dubious quality. Google, therefore, tried to correct this with a series of updates.
The first was Panda (and just before it the ” Scraper update”). Thanks to it the SERPs were cleaned of documents with thin content, which did not offer real value to users and whose sole purpose was to rank for the largest number of keywords. (Do you remember Demand Media and EHow?)
The quality improved, but not that much. Those documents, in fact, were mostly replaced by those that based their strength on manipulative link building tactics.
For this reason Google introduced Penguin.
And, in case that Penguin alone was not able to definitively eradicate the effects of link spam, Google has dramatically increased the number of manual penalties and actions against everything that Google tends to be defined as a link network (hence guest blogging platforms too).
With the SERPs “cleaned” of thin and poor value search results (Panda ) and the result of link spam results (Penguin), the quality has improved … but not that much.
As in a nightmare without an end, their place wasn’t taken by immaculate and highly valuable web documents. No, for the most part we began to see how the good old exact match domains sites conquered the SERPs.
As if it was Sisyphus, Google, therefore , launched the EMD update.
But it launched only on Google.com and for English queries, forgetting the regional Googles; and more than one year an half has passed.
I know what could be the answer of Matt Cutts: We have cleaned the regional Googles SERPs thanks to Panda, Penguin, link networks ban, devaluing article marketing sites et al, we don’t need to roll out EMD as they affect just a tiny percentage of the queries.
The problem is that Matt Cutts is in Mountain View and doesn’t deal with European Googles every day as I (and you, readers) do.
I think that all of you have examples like that one (maybe you could share them in the comments).
Let me tell one thing, though. EMD by themselves are not bad, not at all. But we all know how EMDs tend to rank quite easily even without a strong link profile. More over, many EMD sites were able to create a brand around their own exact match domain name: for instance Booking.com technically could be considered an EMD, or the Spanish Coches.net, but both created a recognized brand with their name, so that we cannot consider them as “real” EMDs.
Therefore, even if licit EMDs exist, it is quite easy also for not-valuable sites (I don’t want to call them “spam sites”, as they aren’t necessarily so) to outrank more valuable ones, making some Google SERPs almost useless.
The Hummingbird glitch?
There is also another problem that since some time now is affecting Google, especially those that share the language used such as it is, for example, Spanish.
As a Spanish SEO friend of mine, Miguel Lopez noticed, if we do a search for “lamparas hospitalarias” (hospital lamps) in Google.es, the results are filled with sites from Venezuela, Panama or Mexico.
Note how some of these not-Spanish ranking sites are not generic domains, but ccTLDs, therefore sites that should be targeting by default their respective Googles, not Google.es.
This “geo-targeting” confusion is not just affecting less common queries like that one, but also also more generic terms like “carro”
The word “carro” has different meanings depending if it is used in Spain or Latin America. In Spain it means “bandwagon”, in Latin American Spanish it is synonym of “car”.
Let’s now see what kind of SERP Google presents if we search “Venta de carros” in Google.es (venta = buy).
The results are even more ridiculous if we just search “carro”:
Where are the bandwagons, Google?
Ok, I may understand that people right now are not used to buy bandwagons, but offering to buy cars in Venezuela or some other Latin American country to the Spanish audience in Spain makes no sense (and I don’t think that those sites are really interested in selling cars to someone in Madrid or Barcelona).
Sincerely, though, this is not even my main concern. That these kind of glitches are possible is telling us that there’s a problem with Hummingbird.
Why Hummingbird? First of all because these geo-targeting mistakes by Google, randomly present also in the past, have become more and more common after Hummingbird went live.
From the very few things we know about how Hummingbird works, we know that semantics play a big role. But also that search and named entities understanding by Google, as word-coupling and, possibly, knowledge base.
That means that Google – probably – is relying on just one big Spanish dictionary, which includes all the nuances/synonyms of the Castilian (Spanish, Mexican, Colombian, et al), but also – in the case of “carro” – that Google is relying on how users commonly use a term. Therefore, if 500+ million users mean “car” when saying “carro”, then “venta de carros” will probably mean “buy cars” and not “buy bandwagons”, which is what only less than 50+ users in Spain are meaning (and not that frequently).
The terrible thing is that, somehow, this way of writing the SERPs is almost not considering what Google itself suggests in terms of geo-localization (these results are so also on mobile, where geo-targeting should be even stronger than on desktop). Therefore we see .com.ve or .com.mx or generic domains in Spanish but clearly targeting Latin American countries ranking in .es and even outranking .es websites.
This reminds me of when Caffeine was rolled out, being the reasons the poor quality of how sites in general deal with geo-targeting and the apparent impossibility by Google to really understand the patterns a language may have, and the somehow short circuit existing between the “rules” commanding Hummingbird and the ones commanding classic International SEO. Especially geo-targeted links don’t seem to help Google as much as before in determining for what region a site is specifically designed for.
These issues, quite possibly caused by Hummingbird, are evident in all Spanish Googles, but if my theory is correct it could be affecting also the French and German ones, and even the regional English based Googles, as you may confirm (or not) in the comments.
The regional Googles are broken.
And it is particularly a deception seeing how “latin-language-based” Googles and regional ones in general tend to be considered a B-Series market by Google.
I may understand that Google.com is economically more important, I may understand that owning almost the 99% of the search market share may discourage from paying that same constant attention than in markets where you “only” own the 67.5% of it (data by ComScore), but the frustration both as search marketer and user is getting bigger and bigger:
- as a search marketer I cannot simply tell “Google is broken” to my clients;
- as a user – at least in Italy and Spain – Google is far from being the flawless Star Trek computer its sales pitch pretends it is.
Google, I love you buying Nest, putting balloons in the stratosphere and thinking about robotics… but what about investing even a small percentage of your 15.42 billion USD earnings in this first quarter in native engineers, and opening local webspam offices for finally solving the quality issues of your regional search engines?