Forum Moderators: Robert Charlton & goodroi
its not the trademarked ghost dataset that went missing, and it wasn't a rebuild like the halloween update.
No, but the overall technique has a familiar feel to it. More than one dataset may be involved this time - and perhaps many more. Interesting that three weeks ago we were hearing reports of googlebot spidering like crazy, and in recent days, reports of googlebot not even showing up for some sites.
[edited by: tedster at 5:09 am (utc) on July 15, 2009]
I keep thinking back to that rumor I reported in part one of the July Updates thread. A lot of the ranking changes I am seeing line up with the idea of "less weight for less relevant links" -- or maybe "more weight for more relevant links".
Gonna run some more tests on this.
Any particular industries you're noticing this occuring more so than others?
As my initial findings didn't support this. :(
One market where I "think" I'm seeing this "semantic theme" backlink phenomenon is the travel SERPs - those with [very specific location] in the query phrase. Sites with backlinks predominantly from other [very specific location] pages are now doing better than those that have more of a scattershot collection of backlinks. This is a change from just a few weeks ago, when backlink quantity seemed to rule all on its own.
One thing I'm becoming quite convinced of (and this has been a growing impression) is it's time to stop trying to control anchor text in backlinks. Instead, just attract backlinks from good, on-theme, pages and let the anchor text be naturally randomized as the webmaster chooses - external anchor text should naturally include a high occurrence of [domain name] or [business name].
[edited by: tedster at 2:46 am (utc) on July 22, 2009]
One market...
Ok, gonna agree with you here. Like you said, this started before the rest of the update and as you insinuate, this is a change among all [location + keywords] from what I can see.
Don't know if this translates across the keyword spectrum, however.
Off to test some more ideas.
After this shakeup I notice the first page of many of the area + real estate type phrases I search are full of the very large mashup type sites vs local real estate offices. And I noticed that the local offices with really old sites that ranked steadily on page one before are all over in the serps now - very few retained their position prior to this change.
Very interesting...
Presumably, this would be easier than the original geo-targetting, where smoothing to regional areas would be required (or the web would break). The tension between how-local Vs how-global would be difficult, whereas here it would be easy.
Could that be another ingredient in the secret sauce? Google rethinking or rebalancing its geo-targeting?
I haven't looked at any geo-targetted terms, so I'm just putting this out there: would it not make more sense to base this [location]-related boost on the the OUTBOUND link profile ("this is where I am") rather than the INBOUND profile ("this is relavent to me"). I mean, your local stuff in Widgetville might not link to the travel company, but the travel company (or local information site) would link to local businesses.
I have another thought that is just beyond sensible articulation. Basically, "location" is a very distint word that will crop up ALOT for popular places, but only on pages ABOUT that place. Thus it will represent DISPROPORTIONATLY STRONG semantic clustering, relative to natural language keywords. Thus, it may not be a special case, but more an uber-case-study for the semantic re-ordering. Apologies I cannot be clearer, but I hope you get the gist.
One market where I "think" I'm seeing this "semantic theme" backlink phenomenon is the travel SERPs - those with [very specific location] in the query phrase. Sites with backlinks predominantly from other [very specific location] pages are now doing better than those that have more of a scattershot collection of backlinks. This is a change from just a few weeks ago, when backlink quantity seemed to rule all on its own.One thing I'm becoming quite convinced of (and this has been a growing impression) is it's time to stop trying to control anchor text in backlinks. Instead, just attract backlinks from good, on-theme, pages and let the anchor text be naturally randomized as the webmaster chooses - external anchor text should naturally include a high occurrence of [domain name] or [business name].
I could not agree with you more Tedster. The more we concentrate on [country] + [keyword] the less (especially new pages) we seem to perform for these keywords. We will be ensuring relevancy of link location but not so much with the anchor text.
There are a number of possible changes in the algorithm that could produce the results that you are seeing including your links from semantically on topic pages hypothesis.
I'm in the UK and have a site that was badly hit post Florida. I am convinced that this was because Google introduced semantics into its algorithm at that time and the UK dictionary needs to be different to the US dictionary but in fact was not.
An alternative hypothesis to (or development of) your own is that Google is changing the way that it semantically evaluates terms as opposed to words. So developing your example [placename][hotel]. When someone is searching for a hotel in placename generally they are not interested in hotels anywhere else only in that placename. This means that [placename] needs to have more weight in the analysis wherever it is analysed. Including on page factors, backlink topicality, anchor text etc.
Some folks have suggested that this update is a "big brands" update. I wonder if this effect is caused by a basic error in the weighting of words within a term. So for example if someone searches for [generic][accessory] if more weight were placed on the word generic then big brands semantically close to that generic would be given a boost.
In our own case we provide services for widgets. When someone searches for [widget][service] they don't want to buy a new widget they want the service pertaining to the widget. However a big brand associated with the term widget as entered the top of SERPS.
I think that we are onto something but need to do more testing.
Cheers
Sid
Most queries I've checked are lucky to show 2 local sites out of the top 10.
It just looks like more sites with Keyword in the URL have been hit hard with yet more filters
i.e Keyword in title tag, meta description, H1, onpage, anchor links to site and in url = penalty
whilst = keyword in title tag, meta descrption, H1, onpage and in backlinks BUT not in url = Rank High
Keywords in URL means more links to that site with those keywords in it whilst a brand site doesnt have that same issue because a percentage of its backlinks wont be keyword
If i was seeing just big brand name stuff ranking, i would agree but i currently see loads of junk ranking.
This is a mix of weakening backlink values, authority/ human review site links no longer counting for anything and semantic overkill on sites that are 100% on topic being chucked out with the junk.
Serps are currently a mess
I have one site optimized for "widget", which is gone for "widgets" (plural), but still ranks fine for "widget".
Here's the analysis:
Incoming links for "widget": 72
Incoming links for "widgets": 31
Appearance on mainpage (title, meta, text):
"widget": 2x
"widgets": 2x
The searchvolume is:
400.000 "widget"
40.000 "widgets"
So everything would point to being kicked-back for "widget", which I could understand somehow. But I got no idea why "widget" is still ranking and "widgets" is gone when it's exactly the same theme.
And this is definitely the core of placing more emphasis on semantics and the meaning derived from any keyword phrases when a query is made by a user.
I see this phenomenon as well in travel, and in some sectors of health now, including the 'practitioner' type searches.
the UK dictionary needs to be different to the US dictionary but in fact was not
I don't think Google relies on a standard "dictionary" approach to semantics. Rather they dynamically create their own semantic relationships through spidering and measuring actual usage.
Their approach to semantics was first patented as phrase based indexing [webmasterworld.com]. While the precise methods in those patents may well have evolved (it's been three years) I'm pretty sure this is the foundation of their approach. At least it gives us a better mental model that thinking of dictionaries alone.
US stroller = UK push chair
US elevator= UK lift
US customized = UK bespoke
US TV = UK telly
US better half = UK her indoors
Here's more from the patents:
The phrase identification operation of the indexing system identifies "good" and "bad" phrases in the document collection that are useful to indexing and searching documents.In one aspect, good phrases are phrases that tend to occur in more than certain percentage of documents in the document collection, and/or are indicated as having a distinguished appearance in such documents, such as delimited by markup tags or other morphological, format, or grammatical markers.
Another aspect of good phrases is that they are predictive of other good phrases, and are not merely sequences of words that appear in the lexicon. For example, the phrase "President of the United States" is a phrase that predicts other phrases such as "George Bush" and "Bill Clinton."
However, other phrases are not predictive, such as "fell down the stairs" or "top of the morning," "out of the blue," since idioms and colloquisms like these tend to appear with many other different and unrelated phrases.
Thus, the phrase identification phase determines which phrases are good phrases and which are bad (i.e., lacking in predictive power).
When I mentioned [very specific location] earlier, I was pointing to exactly this kind of "good phrase" in the linking page.
[edited by: tedster at 10:08 pm (utc) on July 22, 2009]
Do you think that there is some kind of manual intervention on certain phrases going on? ie more than usual, like a concerted effort to manually evaluate terms semantically.
The reason I say this is because about 6 weeks ago all of our long standing #1 terms dropped between 1 and 5 places. Now slowly they are moving back up to #1. Last weekend I came back from vacation (holiday) to find that the two most important terms were back at #1 with mini site links. Today for the next most important term we have gone from #5 back to #1.
Cheers
Sid
Take a look at the pages and then take a look at the cached pages in Google and you will see and you will see the hack.
BTW, Zen Cart has patches already wrote and instructions how to remove the code, if you are a zen cart owner go to their website for the patch and instructions
Moderator's note: We're allowing this particular search term to indicate a widespread problem affecting many ecommerce sites. We generally don't post specific queries.
[edited by: Robert_Charlton at 7:41 pm (utc) on July 23, 2009]
I worked on a client site not so long ago where the client chose to ask the webmaster to manually remove injected code on the index page rather than pay the $2000 to upgrade the cart and fix the error for good
When I mentioned [very specific location] earlier, I was pointing to exactly this kind of "good phrase" in the linking page.
I've seen for some time a kind of " semantic matching " going on with terms surrounding the link text pointing to sites.
We found rankings for phrases not found in link text pointing to some sites we watch. To be honest I haven't dug very much deeper . But there's enough discussion to break this out at some point into a seperate thread.
[btw ] sorry to hear about the ZenCart problems - hope this gets reversed and fixed asap
On most of my pages Google have put a really old DATE next to the results. I think this many be effecting the site as Google no longer thinks the site has fresh content even though we add at least 3 in depth news stories everyday.
i.e Keyword in title tag, meta description, H1, onpage, anchor links to site and in url = penaltywhilst = keyword in title tag, meta descrption, H1, onpage and in backlinks BUT not in url = Rank High
