Forum Moderators: open

Message Too Old, No Replies

Google's 2 indexes & you

Google is alternating between 2 indexes, every 3 days.

         

Namaste

10:31 am on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



I am sure you all are seeing this too, but every two-three days the index changes and if a page drops, it drops to the same position it was 3 days back.

For example, my result for Buy Widgets is alternating between position 6 and position 11. This has been happening for 2 weeks.

To me, this shows that Google is publishing 2 indexes and is alternating between the two.

What is the reason for this?

Theory 1: SEO Neutralisation. Since Jan, we have seen steps towards this. Obvious step being reciprocal links neutralized. Why would Google want to do this? "It's the money stupid", Less predictable results means more spend on Google Adwords by webmasters!

Theory 2: Continuous Update. Inorder to publish a new index and allow it to settle across all data centers 3 days is required. So, DeepFreshBot scours the depths of the web continously and a new index is published every three days. GoogleGuy allows the new index to settle and then pushes the red botton,making live the new index. (Best way to test this is to make title tag modifications to deep pages and see if they appear with the new index).

Theory 3: Both. How do you kill two birds with one red button? Simple, you undertake 2 above, but use two seperate algos, one for each index(theory 1). Thus, index A is on for 3 days, meanwhile index B(with different algo) is being prepared and published. Then Index B goes live and index A goes in for updation, and so on and on. To make it greater fun, every now and then GG inserts experimental algo C for 3 days.

Enjoy!

amazed

12:28 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



Theory 4: Google has finally accepted the responsibility to educate users in that the web is random and has nothing to do with relevancy :-))

mcavic

12:48 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Well, the web isn't random, and neither are SE's - they just look like it sometimes. But if you're bouncing between position 6 and 11, then at least you're getting a shot a position 6, which is good for you. And so is someone else, which is good for them.

Nick_W

12:55 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Thanks Narmaste, that's an excellent post.

I've not been folloiwing the serps much since Dominic (just logs which are averaging out) but I like the theory and will test the deep page thing just to see.

Thanks again.

Nick

tommyleef

1:04 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



In my major keywords it has been about a 5 to 8 day cycle. Number 6 then no where in the top 200 or so. Then, BOOM--back on top again.

The one where we are no where to be found is the same index top results found 4 months ago almost verbatim for the top 20 or so.

My site was still a rapid climber then and didn't already have first results page honors.

I think what is happening is there really was no 'update' simply because there has not been any degree of stability within this 'update' since it took place. I DO NOT think it is a sign of things to come and that this up-and-down, old-and-new will happen every month.

IF Esmeralda was actually an 'update' then every few dates we need new 'update' alerts. Esmeralda has to be the shortest lived 'update' in G History if people think it deserves to be called an update.

There is no way that top 10 then no where, then top 10 then no where can be argued as relevant results.

The only keywords that change that much on revelancy would be truly 'Breaking News' type items where yesterday's headline is trumped at the top by this morning's follow-up. However, even in breaking news a page that was top 10 worthy today shouldn't drop to off the index or so far down no one can find it in just a matter of hours.

So, Esmeralda is over, then we had F, G, H, and whatever the next hurricanes are to be named because all these happened since 6-15-03

mfishy

1:21 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



<<In my major keywords it has been about a 5 to 8 day cycle. Number 6 then no where in the top 200 or so. Then, BOOM--back on top again. >>

Woke up this morning to find another page vanished from #1 to oblivion.

The cycle has been every 2 days or so pages will come back to #1 from nowhere, but each time it happens I get a bit more worried. :)

<<So, DeepFreshBot scours the depths of the web continously and a new index is published every three days. >>

It is unlikely that the flux would be so dramatic, assuming they are using the same algo, in just a few days. A page cannot fall 400 places in 3 days due to updated links and content.

Marval

1:44 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



First off..I was about to post a completely different theory but figured that my looking at SERPs might be a little skewed.
Ive been one of the many that had the index problem and it has followed the 3 day pattern lately.
To really look into what was causing this, it takes looking at the sites that go to the top when you dissapear, as alot of the sites that make it to the top for the off 3 days are not "regulars" in the top serps.
Unfortunately there are some compounding problems that just dont make sense...a keyword that is highly competitive just lost some 20 million results on most datacenters...out of 202 mill to start with, so about 10 percent.
Couple with this, some unusual bot behavior...someone posted earlier they had noticed a new bot moving around that in my experience since yesterday, has been requesting pages like a deep bot, but is getting 200 responses on every page, right after the normal "freshbot" comes through and gets a 304 for no changes. The new bot group is 64.68.88.* and acts more like its trying to reindex the whole web.

The main repeatable things that I see when the indexes change is that the onew where the index pages drop, the top serps have loads of new content but that content hasnt been through any ranking, filters or backlink checks. Most have 1 or 2 backlinks, and would not normally be in the top serps for real competitive terms. Most also have whitebar PR (if you can count on that) and no directory listings.

The other noted changes are that the directory (I think its been noted here before) doesnt change for the change in SERPs. As has been pointed out before the directory seems to be very old and has not had a DMOZ RDF in some time(or has reverted to a much older directory)

Just my observations, but I am starting to think that the update was truly over, as I did notice a small percentage of pages get a PR change, but backlinks from authority sites have not been brought back in.

tommyleef

2:01 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



Great observations Marval

I wish I were on the 3 day cycle myself.

The fact that different keywords are on different number of days between the 'old' and 'new' rankings is even more disheartening when trying to understand things.

My site has kept it's 5PR and there are more high PR pages inbound, and Google shows this in backlinking, than the guys that hold the spot when 'old' results are there. More relevant content than the other guys too and those reasons are why my site climbed steady for 13 months now. Right when we finally broke the top 10 Dominic rained on my parade and then Esmeralda did the same with tropical storm rains.

What I am seeing is when the 'old' results reign high--it is more spammy.

One thing that may be factoring in for my site is that it took forever to be listed in DMOZ. I finally was included. That may be one part of why I go up and down so dramatically. The oblivion results are being taken from crawls prior to the DMOZ listing.

Never, ever thought DMOZ carried that much weight though.

claus

2:18 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



The only keywords that change that much on revelancy would be truly 'Breaking News' type items

Good point tommyleaf. It's interesting that recent changes to pages are not making it to the serps, even though you've had visits by the bots in the meantime.

Personally, i'm getting more and more convinced that it is actually an old scan that is being used for the serps, with some "feeding in" of new items from time to time, but without registering changes to the pages that were part of the old "snapshot"

The things i see supporting this line of thought is a) increasing difficulty in finding relevant "breaking news" type items (or just, say, events that happened a few days back)

And b) major recent changes not being shown in the serps (such as a totally diferent title for an index page - not even a deep one)

Plus c) entries that really are obvious to a spam filter sometimes gets through.

At the same time, i have been seeing increased bot activity for some time - i'm not all that convinced it has levelled out yet, so all the new stuff is getting indexed, it just doesn't make it to the serps.

Ohyes, and then there's the up-and-down-thingy...

The "Two indexes" idea is very interesting. To me it seems that G is collecting a lot of new stuff these days. Combine that with the notion that the serps are filled with old stuff (plus a few all-fresh listings from time to time) and you can't help thinking.

Perhaps G is "just" building a brand new and much larger snapshot but has not yet released it to DCs. If that's the case, the update is still running.

Even if it's not the case, judging by bot activity (increased) and serps (outdated+all-fresh+spam) alone, i'd say that this update is not over yet.

(just my personal thoughts, really)

/claus

my3cents

2:26 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



I too am noticing a 3-4 day cycle, only what I find is that when the site is listed up high, where it has been for quite some time, it lists www.domain.com

this holds for 2-3 days and then vanishes, but it is showing the same page much further down 60-80 and lists it as www.doman.com/index.html

Sometimes it lists the same page twice, using both paths, where each path has it's own ranking. It still makes no sense how Google can determine this is the same page for a couple of days, and then after that thinks that the two paths are two seperate pages.

I have also seen this same thing with www.domain.com and domain.com - again, like it's is sometimes acknowledge this is the same page (same file), and sometimes does not.

Another strange occurance is that certain index pages are showing up for serps that are not highly relevant.
example:

This company sells metal widgets in a types and sizes, but all of them are metal. The site used to always be in the top 10 for metal widgets, with sub pages ranking well for small metal widgets, large metal widgets, etc.

when the index changes in a few days, the main page is gone, or listed twice using different paths, but some of the internal pages show up high for "blue wood widgets" "wood widgets" "blue wood" - This is strange because we do not sell wooden widgets, on one page of the site, it mentions one time, that metal widgets are more durable than wooden widgets. Oh, and it also offers blue metal widgets (which ranking drop for every few days).

Since there are hundreds of websites that do in fact sell wooden widgets, I don't understand how having the phrase appear once on a page could put it at the top above other more relevant sites. I also see this same site showing up for a few irrelevant terms, and the page it lists is either an internal page or the www.domain.com/index.html

Should I fill out a spam report on myself for showing up on terms that are not relevant? j/k

One thing we've all noticed is a lack of consitancy, and of course the fact that GG has skipped addressing this topic all together and left us all wondering for the past two months.

jady

2:47 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



"There is no way that top 10 then no where, then top 10 then no where can be argued as relevant results."

Took the words out of my mouth! Plus looking at it from the other side of the spectrum (someone searching G for something - not checking ranks) my Mother called me the other day and said, "We found a site for ABC123 the other day and I went back on Google to try to find it and it was gone! I didnt save it in my favorites - how do I get it back?"

As someone that can always answer questions from friends and family, I simply told her - "I DONT KNOW"

I too find results non-relevant when a good site is in the top few, then you do another search maybe to find that site again to make your purchase or what not - and its literally GONE!

amazed

3:05 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



hi jady,

I love that you are quoting your mother, its a variation...

problem is, its a fact, I hardly dare to mention it as nobody will believe me...

my father has noticed too... :-))

but seriously Google, users don't mind number three or five or maybe even second page, but they want to re-find.

You are robbing them of the last illusions of stability!

JasonHamilton

3:08 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



If you watch your keywords on google several times a day, you'll see that the changes are not on a 3 day cycle, but happen several times a day. I'm guessing since the speed at which the index seems to change (changes immediately just by hitting refresh on the browser), there are either:

1) more than one index per data center
2) the query request is not going to the same data center each time.

I don't claim to know how frequent the changes are -- mostly because I'm not flooding google with continuous searches, but I've yet to see the search results the same during any one day for the full day. they always seem to be changing.

I do have to mention that while the positions change frequently, they do appear to be swapping between positions that I've seen in the past.

Here is some rankings I've had on 3 seperate keywords for the last few days:

June 29: 3/5/2
June 30: 3/5/2
June 30: 5/8/4 (afternoon)
July 1: 6/8/4
July 1: 8/8/4 (afternoon)
July 2: 6/6/4
July 2: 6/6/4 (afternoon)
July 3: 6/6/4
July 3: 6/6/4 (afternoon)
July 4: 8/6/4
July 4: 8/6/4 (afternoon)

mfishy

3:29 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



JasonHamilton

The ranking changes you describe are not that problematic for the user or the webmaster. It is the changes many of us are noting - #1-#3 to COMPLETELY GONE that are a bit odd.

JasonHamilton

3:35 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



True enough -- being completely dropped would be a problem.

Also, I don't know if this has been mentioned previously or not, but I've also seen some of my sites where the site will still rank, but certain sections will have all the indexed pages dropped. I've noticed one section gone for almost a week. They're back today though.

esllou

3:42 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



it is a 3-day cycle and we are on about the 5th one now. I search for a super important search term about 10 times a day. I was number one, then for three days, number 20, then number one for three days, etc, etc.

I just keep praying it will stick when I am number one! LOL. I remember my mum telling me the wind will change and my face will stay like that....it's a similar thing at the mo with Google and I am hoping the wind doesn't change now as we are 36 hours into a "bad" cycle. Big hopes from tomorrow evening!

ps...this is ridiculous! Am I the only one hoping good things from MSNBOT?

tommyleef

3:49 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



Bottom line is I feel everyone here pretty much loves Google. We just want them to produce what made them great and what serves users--Quality Relevant Results.

We all have been preached to about the overwhelming and daunting task it is to do what Google does. I fully agree it is just that--a HUGE JOB that is IMPORTANT.

Having said that, This task is the one Google took on and is the one Google gets unreal amounts of money to do for Yahoo!, AOL and many other companies.

Not at all saying Google should stop...just saying if this is your job--then do it better and stop telling the world how hard it is...if you are looking for a raise in pay then you already got that with the deals made in the past year.

If you gave you boss up-and-down performance and often went on about how hard what you do is to do...they would soon start to think you can't handle the job.

Google can handle it and is rewarded with great pay.

So, the Hollywood interviews and press releases touting how wonderful a worker you are over. The employee of the year award for '02 has been received and is now dusty on the wall, the training and experiment time has passed. The inner office meetings have ended. Smoke break time is finished.

Now it is past time for our highest paid employee to punch in and perform like a good dependable worker.

Most people giving examples of the past 2 months from Google are not just run of the mill Google bashers. The past months have been a mess. CNBC needs to cut the cameras off and let Google get back to work.

claus

4:14 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Mozart
>>Google has multiple personalities

One month ago i made a change to some website. The URL is the same, changes are not major (same topic, different weight on subtopics), they are reflected in content as well as index page title.

The red widgets/blue widgets thing is exactly what i have observed, eg. for a two word combined keyword like, say, "widget gadget".

-----------
Typing in "widget gadget" gets my old index page on #2 (with www) and #3 (without www). There's no sign of the new pages, these are old (> 1 month) listings.

Typing in "gadget widget" gets my new index page on #1 (without www and there's not a sign of the url with www in it or the old pages)
-----------

The two keywords are exactly the same, the only thing i change for the search is the order.

This is why i keep saying that there are old pages in the database, that hasn't been updated yet ~ that the update is still not over.

/claus

<edit>clarified a bit</edit>



added: just tried adding "and" between terms (in English as well as in the site language) - a 404 page on another site came in #1 in one combination, the other combinations yielded entirely different results.

Normally - in the "good old days" - i would find the page among top results regardless of order of terms and the "and" keyword.

rfgdxm1

4:54 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



I have a domain where the home page, for one certain keyword every few days bounces from about position #5 to between #50-70. I haven't changed the home page of that one for a long time. Dunno if it is 2 different indexes, or perhaps 2 different algos applied to the same index. Google is schizophrenic in that whether or not it likes this domain changes every few days.

James_Dale

4:56 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



Google is building a hyper-complex automated categorisation engine. Clear your cache and keep your eyes on that status bar.

Fearless

5:09 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



I'm responsible for five not for profit sites, four established and one new. Modest size, PR 3 to PR5, occasional updates and new content.

I've been monitoring our results across all the data centers. Once the index "settles" I don't see the "up and down" that Namaste describes. Maybe our sites are too "small"? Maybe our search terms aren't competitive. Dunno.

If Google has been doing this deliberately, our sites haven't been affected.

AmishJohn

5:27 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



"Theory 1: SEO Neutralisation. Since Jan, we have seen steps towards this. Obvious step being reciprocal links neutralized"

Has Google Guy or anyone else at Google said that reciprocal linking was not a good thing and or that we shouldn't do it?

My own view is that having a thousand reciprocal links demonstrates a profound desire to have a high PR, but doesn't say alot about relevancy.

Of course dropping the ranking value of true reciprocal links will encourage SEOs to develop clever schemes to have what appear to be non-reciprocal links.

annej

5:44 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Some things I've noticed. In the last few days things seem to be settling in. All the data centers aren't the same yet though they seem to be getting closer. Results seem to have gone back close to the pre Dominic era on the key words I've been following. Yet the snippets and cache is new. I haven't seen a 3 day cycle, it seems more chaotic than that. How I miss the good old days when Google was semi predictable.

netnerd

6:11 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



A site i had in the top ten for a very competitive search term has disapeeared today.

I cant find it anywhere - yet it still has PR6 and shows all its backlinks.

It is i confess similar to another site in the top ten i have, - anyone know what could have happened? Will it come back?

It is still in ex and one of the other datacentres, but dropped from #3 to #5 on them.

g1smd

8:09 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member Top Contributors Of The Month



For those people with results that vary over a period of a few days are you searching at www.google.com or directly at the various datacentres?

The results from www always vary because each new search will bring results from a different datacentre compared to your previous search.

Searching directly at a datacentre should bring consistent results for that datacentre.

The datacentres are www-fi, -in, -va, -ab, -dc, -cw, -ex, -zu, and www-sj but here is an additonal point:

Each datacentre at www-xx has an alias at www-xx2 so there is also www-fi2, -in2, -va2, -ab2, -dc2, -cw2, -ex2, -zu2, and www-sj2 to check as well.

Does anyone see that for each datacentre that -xx and -xx2 are the same, similar, or completely different?

I can't check it, as for me the results stabilised many many weeks ago with a #1 listing in all of the datacentres all of the time.

[edited by: g1smd at 8:14 pm (utc) on July 5, 2003]

Namaste

8:12 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



All would be great had I not checked the red widgets SERPs! There my page was as well! I repeat, same page, different SERPS, different display! Both SERPs had the full www.domain.com/dir/filename.html with the red widgets result having no fresh tag and the blue widgets result having the fresh tag of the last spidering! But one with the old title, the other with the new title!

this is actually a very old problem and I have seen this for as long as I have been watching Google. Google often displays old title tags.

rfgdxm1, yes it could be 2 algos applied to the same index but the only purpose this would serve is seo neutralization. Publishing 2 indexes had the added benefit of freshness and is better for users.

I have started looking verr closely at the results above me during the cycle that my position drops.

here is no way that top 10 then no where, then top 10 then no where can be argued as relevant results.

there is actually a way in which this can happen. The site has undertaken what I call Micro SEO as compared to Macro SEO.

Mirco SEO is when you have achieved a position by scoring on 2-5 factors in the Google algo, such as Keyword Density, Insite links, Reciprocal linking, etc. Thus if Google is undertaking heavy SEO neutralisation in one cycle, then you drop drastically. I recall in another thread the point about "SEO gernerated links" was mentioned; the theory was that Google was penalising for perfect links, for eg., Buy Widgets, on all sites linking to you.

Macro SEO is when you have focused on building a site that withstands algo changes. Focus on all factors and try and get to the top using robust methods rather than gimicky stuff. This is what Brett keeps talking about "build a quality site, don't focus too much on variations in Google algo"

steveb

9:01 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member



Building a good quality site doesn't keep you from having links missed or algorithmic calculations done poorly. For three days I'm #6 for allinanchor; then for three days I'm #42 for allinanchor. Huh? Building a quality site, or even any aspect of SEO has nothing whatever to do with a LOT of the bizarro fluctuations going. Yes, garbage SEO allows sites to burble up to the top twenty (but not very much to the top five) in this current half of the Jeckyl/Hyde dance, but that is totally different than the ranking of inappropriate pages (contact pages instead of content pages for example) and inconsistent calculation of objective numbers.

midnightcoder

9:48 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



>>Theory 2: Continuous Update. Inorder to publish a new index and allow it to settle across all data centers 3 days is required. So, DeepFreshBot scours the depths of the web continously and a new index is published every three days.<<

I'd go with this theory at the moment.
As I mentioned elsewhere, I have a site that had 10k new pages added to it recently.. instead of adding the lot in one update google is progressively adding approx 300 new pages every 3 days or so. The new pages stay in the index, followed by another batch a few days later. The new pages can be seen in -fi before they hit the main index.
Rolling update

g1smd

9:55 pm on Jul 5, 2003 (gmt 0)

WebmasterWorld Senior Member 10+ Year Member Top Contributors Of The Month



steveb: you seem an obvious candidate to answer the questions that I asked just two posts above your last post. [webmasterworld.com...]

My observation is that Google dropped all of my old pages on an old site (now containing noindex,follow as the site is no longer updated) on 2003-06-15 on -fi and added all of my new pages from the new website, on a new domain, the very same day (about 5 weeks after that domain went live). That new result spread to all of the other datacentes, one each per day, ending with -in on 2003-06-22. The old site was badly listed; but the new site is #1.

I have also noticed that the number of results in the listings has gradually crept up during this time. In the hundreds, the number of results go up by 3 or 4 more results per week. This points to continual new data being poured into the mix.

Googlebot was a regular visitor every 2 days until nearly the end of June, but then went missing for over a week; eventually returning yesterday or the day before.

[edited by: g1smd at 10:07 pm (utc) on July 5, 2003]

James_Dale

9:58 pm on Jul 5, 2003 (gmt 0)

10+ Year Member



g2smd, for the phrases I am checking, both -xx and -xx2 versions of all DC's are identical.

However, there are fairly regular changes in specific datacentres. When these occur, both -xx and -xx2 change simultaneously.

This 104 message thread spans 4 pages: 104