Google Updates and SERP Changes - June 2010

[Hissingsid]

< continued from [webmasterworld.com...] >

I've been trying to work out what happened on May 17th and May Day. I seem to spend a lot of time analysing on-page factors and back link profiles for my own ranking pages and those of competitors but have fallen into the trap of focusing on a few big traffic volume terms.

Recently I've been looking at some second tier terms and this has been much more enlightening. This has also made sense of what is happening for the big volume terms.

In site linking seems to be out ranking external back links if done in a particular way.

The pattern I'm seeing is this, when all other factors are roughly the same, pages on sites where the site is organised in such a way that there are a large number of back links containing the keywords in anchor text within the site. ie Sites with many pages each page focusing onto one page with the same anchor text and very few back links from external sites are out ranking those that have a better external back link pattern.

So sites that have Javascript or CSS drop down navigation systems with the same links on every single page seem to be doing very well.

The other thing I'm noticing and this may confirm what I'm saying above. I'm seeing SERPS loaded with home pages.

Anyone else noticed similar patterns?

Cheers

Sid

[edited by: tedster at 11:32 am (utc) on Jun 1, 2010]

[mantucket]

walkman, is that 'old caffeine' testing IP - i.e. has caffeine but not MayDay alg? If so that would be really useful for testing, can anyone confirm?

[curioustoddler]

walkman,
Can you please check google.co.uk results? At one point i had asked you and you confirmed that google.ca was showing caffeine results at that time. But now google.co.uk seems to doing that job.

[Robert Charlton]

I think for us the problems reside with the synonyms and other matching /intuitive data going too far.

scottsonline - I'm thinking that synonym matching set too high might in fact compound itself as a problem. Am assuming that the algo is trying to get rid of a certain kind of content by ignoring normal relevance signals and looking for others. Suppose that the algo is first filtering sparsely worded content too deeply, eliminating so many pages/sites from consideration that it is left grasping for some kind of a match (even a bad match) to satisfy the query.

In such a situation, an algo might...
a) at one end of the spectrum force matches towards keyword-stuffed pages, as you were reporting earlier...
...or else...
b) go in the other direction and grab pages that don't really match, but are just synonymous, as has also been reported.

I'm wondering what people who reported lots of dupes are seeing. Are those still there?

[walkman]

Can you please check google.co.uk results? At one point i had asked you and you confirmed that google.ca was showing caffeine results at that time. But now google.co.uk seems to doing that job.

The name I search for is a European company, but it does come #1 in UK too (as it did in on the old caff IP)

[Robert Charlton]

@gford: That is TOO funny!

Re the music search in the Google announces Caffeine is completed [webmasterworld.com] thread...

I should mention that I had a pretty good idea that WebmasterWorld would end up outranking the pages that were returned, but I decided not to edit Slinger's post that asked about the search. (Notice, though, that I'm not repeating his search terms again). It's pretty instructive, actually, to try the search on all three engines, maybe even read about what I found when I tried doing that.

... I can pretty safely say that in English language serps as indexed by the three major search engines, there aren't enough results to fill the first page.

I compared Google, Bing, and Yahoo, searching a variety of ways, and arguably Google came up with the best results, with Bing arguably the worst. There are only two sites on the entire web... two pages, if you enter the search as suggested... that provided immediately useful results for the query.

Before WebmasterWorld intruded by posting the search, Google ranked those pages as #1 and #2. Now WebmasterWorld is in there too, in between those two... one of the reasons we don't like to post specifics. ;)

Anyway, it's very clear to me as I look at Google results that when matches are sparse, as they are in that search, Google is returning a variety of different match types in different positions, one of which has Blackduck run together, with "winds" an onpage match on that result. Google's results appear to be differentiated by match types when there's not much in the index to satisfy the query.

I find this kind of interesting, as earlier I'd mentioned the following in the History Repeating [webmasterworld.com] thread...

...I'm seeing no overall patterns either for relevance or quality... and again, no consistency throughout a search to suggest reasons for rankings. It's as though in different markets and different spots on a serp, different conditions trigger entirely different algos...

The music search suggests, at least for obscure searches, that this is what's going on. Each position was in essence a different "algo", with results returned in the order of match types that Google, because of the history of matches it keeps, thought might make me happy. Note that I'm not saying that we should be happy about having our queries changed. I am saying that if certain types of pages are getting filtered out at the start, this might describe what we're seeing with the rest.

Back to the music search... for Google to run two words together and return Blackduck as one word (after first returning the two matching pages in the #1 and #2 spots), might make statistical sense... more so, perhaps, than what Bing returned as #1, which was: "D2: The Mighty Ducks" as #1, with no "winds" at all. No winds at all just blows me over. ;)

Anyway, that's why I think we're seeing some of what we're seeing.

[Robert Charlton]

PS to the above...

I can't say whether Google is "stratifying" less obscure searches, but I'm fairly sure it's what they're doing for queries with very few results that they deem useful.

[vandread]

Thanks, they were singular and plural forms of my trophy KW that were generated by an SEO Plugin. I really don't know if that was the problem, but felt compelled because it was over-optimization, and a blatantly obvious SEO tactic. All my internal linking will now be done manually with some careful thought.

I do however suspect that it was an indexing thing because G is giving far to much weight to my homepage. It is returning the homepage for the SERP ranking for many KWs, rather than the actual individual page. The pages are still indexed, I think G is still sorting itself out.

Interesting. It seems as if I'm in the same boat as you. I also noticed that my HP is ranking for new entries that I write in my blog and not the article itself.

I'm also using an internal keyword tool but not to link to my HP but to inner pages if keywords match. I might be inclined to turn it off for a few days to see what happens.

I'm just not sure of the possible negative consequences this could have.

[aspdesigner]

"I've been frustrated in my searches with google and find myself using tighter and tigher parameters, or using a metasearch engine or bing."

@Freedom: Sometimes, when we are doing SEO, we tend to have too narrow of a focus to see the big picture.

It was my own experiences accessing Google from a USER's viewpoint, as detailed in my first post in this thread [webmasterworld.com], that alerted me to the fact that something was seriously wrong with Google.

"I'm not trying to jump on the me too train, but when doing searches, the pages that show up set me back a bit as it never seems to be what I am looking for."

Take a look at the titles of the pages that are coming-up. Are many of the titles showing-up in the Top-10 only containing SOME of the search phrase you entered, even for competitive searches? Or containing an alternative word instead of what you actually searched for, that has a similar but not identical meaning to the word you actually entered?

[curioustoddler]

Thanks walkman,
Let me share my own experience. I am closely watching two sites in the same trade. One is 6 years old site and other is new site( launched in August 2009). Initially i was check caffeine results at that 209... IP. then google.ca started to show new results. And for more than six weeks google.co.uk seems to show new results. why do i feel that.
For new site we have to work on backlinks and content for hundreds of pages. We pick up 20-30 pages each month and work on them. So for few pages we worked in Oct, then for few in Nov and so on. Now i need to check when these pages start to appear in allinachor and when in actual results.
Earlier new pages used to show in that caffeine URL only. But for last few few weeks, they start appearing in UK, then gradually move to ca, and in US google.com i can only see what we had done by Feb second week only. But in UK i can find pages for which we worked in March last week and April first week. There is no visibility for pages for which we worked after first week of April.
Now another interesting thing. One old site is gradually losing backlinks and dropping in allinanchor and actual results in UK and somewhat Ca also but not at all in US. In UK it has dropped a lot, almost disappeared in last two months for many pages. Drop was almost half in ca and not at all in US.

[aspdesigner]

"I also noticed that my HP is ranking for new entries that I write in my blog and not the article itself."

@vandread: As detailed in my post on June 9th here [webmasterworld.com], one of the symptoms we are seeing both now, and also with Florida, is pages that are simply LINKING to relevant pages (like affiliate pages), coming-up higher than the relevant pages themselves!

I would be hesitant to edit your linkage right now, for given Google's current state, rather than getting it to switch to your relevant interior pages, it is quite likely you will simply loose your rankings for those specific searches.

[vandread]

@vandread: As detailed in my post on June 9th here [webmasterworld.com], one of the symptoms we are seeing both now, and also with Florida, is pages that are simply LINKING to relevant pages (like affiliate pages), coming-up higher than the relevant pages themselves!

I would be hesitant to edit your linkage right now, for given Google's current state, rather than getting it to switch to your relevant interior pages, it is quite likely you will simply loose your rankings for those specific searches.

Making changes now could be counter-productive, I agree. It is just hard to hold on for 4-6 weeks hoping that things correct themselves in that time. But I guess I still have the option to turn of the plugin if the traffic has not come back by then.

[aspdesigner]

"Am assuming that the algo is trying to get rid of a certain kind of content by ignoring normal relevance signals and looking for others."

@Robert: Sorry, but that doesn't match what we're actually seeing here. If I ask for "X" and it gives me "Y", but then I ask for "Y" and it switches to "X", that is clearly not ignoring that relevancy signal, it is using it to EVADE.

For example, if every time you glanced towards someone, they ran and hid behind the nearest tree, claiming they must therefore be ignoring you is NOT a reasonable conclusion!

"Suppose that the algo is first filtering sparsely worded content too deeply..."

Sorry, but that doesn't match what we're seeing either. Sparsely worded content seems to actually have an edge, both now, and also in Florida.

Here is an example from back then -

"A Clara personal page is first for keywords

Holiday + destinationkeyword

This page has only 10 lines and tells hows they enjoyed
their holiday a few years ago.

Yes well done Google its relative bloody useless for those looking for a holiday :)

I was worried when i first saw this update now I can't stop laughing because their is so much crap ranking well."

Or an example right now - that incomplete site we were building, that had a PR0 with no incoming links, and was not only coming-up Top-10, but out-ranking an established relevant site. Outside of headers & footers, the unfinished page that was actually ranking well ONLY CONTAINED TWO SENTENCES OF TEXT!

There is a simple explanation for all the spam & junk, one other members here concluded the last time this happened. Simply put, by targeting the well-built, SEO'd or relevant sites, what you are left with is the less desirable results, or as one member put it -

"a pile of irrelevant trash that was before now buried below 5 pages of decent SERPS."

Your second post, I will check out, and will get back to you in a bit.

"but I'm fairly sure it's what they're doing for queries with very few results that they deem useful."

But 136,000 matches, and it can't find 10 good ones? I've seen the old Google handle searches returning FAR fewer results than that, and STILL manage to find some reasonably good ones for the Top 10!

We are also seeing the same sort of results, even from competitive commercial searches, with lots of actually relevant sites to choose from!

It's also hard to imagine how evading the best matches, and/or giving us something other than what we actually asked for, would be considered...

"results that they deem useful"

Unless by that, you mean results that are more likely to cause frustrated users to click AdWords instead! ;)

[backdraft7]

We were experiencing a comeback, now again it seems like a switched was flipped off for the past 16 hours.
Ohno: do you concur?

[drall]

Same here backdraft7.

[backdraft7]

I compare my sales from last year Friday June 12th with this years Friday June 11th. Last Years traffic level for the day: 1415 unique visitor, sales for the day 13.
Now on June 11th 2010, traffic level: 1471 uniques, sales 6.
Last year I could tell time by the sales, nearly one per hour.
This year, (and just after May 17th I might add), sales are down at least 50% - all I can presume by this data is that the traffic is now much less targeted. Like I said above, we'll have a streak where traffic seems to return to normal, then a complete shutdown. I doubt the economy has much to do with this.

[indyank]

same here @backdraft7 @drall this is bad...i hate the weekends...

[gford]

We also saw the "switch" about 27 hours ago now from the time of this post and many kws are still percolating to the first page.

I been crossing my fingers this sticks but I am thinking it won't as 3 of the last 4 friday/saturdays I have seen very similar phenomenons.

The difference with this one is it wide-spread over many more kws and pages whereas before it was isolated kw and urls that I could barely even find and only noticed due to a spike on realtime page views.

Also...when did webmasters turn off the DATE last seen option when you download all links pointing to you, bummer... :(

[indyank]

@gford exactly...this fri/sat it is more widespread....

[Andylew]

things seem to be on the move, I think they have a huge backlog of crawled pages to process. The latest change we have seen is a site: search filtered to past week or 24 houra has always returned next to no pages. However this morning that changed, still nothing in past 24 hours but past week now shows over ten thousand. These have been added overnight. The interesting part is these pages we crawled weeks ago, I know this as we have altered our page title structure 3 weeks ago and they have all been listed with thw old style titles - google is playing cathup. Hopefuly now caf is live it wont take them long to catchup on the backlog

[backdraft7]

I think they have a huge backlog of crawled pages to process.

What happened to the blazing speed of indexing the equivalent of a stack of pages 3 miles high every second? (source: [googleblog.blogspot.com...] )

[Andylew]

it could just be they have a months worth of backlog to get through first. has anyone else seen spider activity drop through the floor?

[backdraft7]

All I've seen is sales drop though the floor. I gotta bet that there will be an economic backlash from this algo and Caffeine update that might last months. I'm going on one month of highly reduced sales, which means I will spend much less than normal...and it all trickles down. This Saturday is looking like another wash.

Looks like Google CFO knows something we all know too...he just sold off 1137 shares of GOOG (only a half a million $). Maybe he just fueled up his boat. (source: [gurufocus.com...] )

[Andylew]

Anybody any ideas what Sprievodca stavom lokality means?

It just showed up as a sitelink on webmaster tools searcing on .co.uk!

Come on google pull your socks up!

[tedster]

Google Translate says it means "Site Guide status" in Slovakian. Well, it seems to show one thing - sitelink menu labels are algorithmically created.

[tangor]

Spiders? May12 - Jun12

Bing: 22K
Yahoo: 23K
Google: 1.4K (yes, that is a decimal point)

[ohno]

@backdraft7, yep, the switch WAS flicked then deffo flicked back again, how much longer G? I have two thoughts in my mind.....

(1) Testing is STILL going on, G is not happy.......
(2) This IS the new G, get used to it, G IS happy....

Either way I'M not too happy.

Thankfully the world cup is on!

[ohno]

BTW, Gbot still very active,WMT shows a jump of URL's for one site BUT a DROP for another site.....

"I doubt the economy has much to do with this."

The economy doesn't shut down the same time as a G "update" ;)

[backdraft7]

The economy doesn't shut down the same time as a G "update" ;)

it a way actually, Google shuts down the economy...

One big problem is the Wikipedia results.
First off, if I need a result from wikipedia, I should just have to add "wiki" to my query (a poet...)
We don't need the "dictionary" jammed down our throat as a generic search result.

[Robert Charlton]

If I ask for "X" and it gives me "Y", but then I ask for "Y" and it switches to "X", that is clearly not ignoring that relevancy signal, it is using it to EVADE.

aspdesigner - Agreed that "ignore" is the wrong word. For me, "reject" might be a better choice than "evade"... as "reject" would suggest the step in the process that I'm talking about, whereas "evade" might describe the whole process. I considered using the word "filter" when I wrote that up.

In any event, I feel that certain sets of conditions appear to be triggering a filter that drops the page from consideration for a given search... "de-valuing" as you've called it... and replacing them with best matches from pages left unfiltered, which might include synonym matches.

Acknowledged, again... when the results are bad, they're sometimes very bad, as they don't seem to degrade in the ways we've previously expected. I'm seeing lots of extremely good serps now, though, where all the traditional optimizing factors appear to be in place and working, not "evaded".

On the rest, Google may inevitably make some adjustments, and we also will make some adjustments. I'm thinking that the information we can gather now, about what trips the filter apart from traditional optimization, will be helpful to strengthen our own sites.

In cases where there are assumed known good pages that aren't being shown and the results are substandard... I'm not convinced what we're seeing is as simple as the algo just rejecting a title match, though it may end up looking like that's what's happening. The algo is rejecting something else, or a combination of something elses. Those factors, IMO, are what are important to identify. Apart from all of those white hat seo title matches that we're seeing, which suggest that Google isn't routinely rejecting title matches, we've also had pages with exact title matches outranked by pages without such matches for a long time.

There are of course also likely to be off-page factors, like patterns of external inbound anchor text, which are probably interacting with the title element and onpage text differently than they were before.

< continued here: [webmasterworld.com...] >

[edited by: tedster at 12:04 pm (utc) on Jun 15, 2010]