179 replies to this topic

[Michael Martinez]

Several people have responded to my request for URLs, and I will get to them all, but please keep in mind that I have to work today, as well.

Anyway, the first few sites I looked at seem clean. I'm still collecting data, and I will make a more detailed post later today.

Please keep the URLs coming. I promise to keep your sites and targeted search expressions confidential, and I am NOT soliciting business.

I would prefer to keep my notes in this thread as I can easily identify the subject line. And I wanted people to be aware that the community is coming together on this one.

Matt Cutts has indicated there may be two more parts to this update over the next two weeks. Everyone needs to hang tight. For all I know, even my own (presently well-performing) listings could go south in the next two weeks.

I am as curious as everyone to know what Google changed, but we may have to wait 2-3 more weeks just to see how it finally shakes out. All your listings COULD come back at that time without your doing anything.

[Jill]

All your listings COULD come back at that time without your doing anything.

Usually that is the case when they throw the baby out. I certainly wouldn't change anything if you had a site that you know is in good condition that previously ranked well.

Sometimes it's not YOU, it's GOOGLE.

[oneofthe3lions]

Sometimes it's not YOU, it's GOOGLE.

Sometimes? You mean 'Usually'?!

[Michael Martinez]

I have looked at over 10 Web sites now. This was a pretty good response, so thanks to everyone.

I've got to go eat lunch.

As far as I can tell, with one or two possible very minor infringements that I would not expect to see penalized, all the sites look clean on first glance.

Everyone please keep in mind that I don't have time to do thorough analyses. But nothing obvious leaped out at me.

I'll say this much: I think Jill's instinct is right. But let me mull this over through lunch and I'll come back.

[Michael Martinez]

Okay, I've now stuffed my face--er, eaten lunch. I feel better.

Well, let me start off by providing a brief aggregate profile of the sites I've looked at.

AGGREGATE PROFILE
In my opinion, all of these sites offer unique, legitimate content. Some of them compete in very heavily invested markets, so their competition is fierce. The site operators deserve kudos for their previously good placements. None of these appeared to be average affiliate farms or anything that the general public would regard to be "low quality" sites.

All of the sites use external CSS files, but most of them don't do anything anywhere remotely sneaky. I've made one specific, but hesitant suggestion regarding one thing I noticed, and that site operator is looking into it.

Most of the sites use Javascript, including many external Javascript files. The only comment I have regarding the Javascript is that I think more people should make use of the NOSCRIPT element. Some of the sites use it, most did not seem to.

I found no obviously hidden links and no obviously hidden text. Some of the sites do have what I would regard to be "stuffed" KEYWORDS meta tags, but I don't believe that really has any impact on rankings.

The sites address a broad variety of industries. Some are travel related, most are not. Some of these are devoted to very small niche markets (in my opinion). Competitiveness just cannot be measured for those sites in the same way as, for say, the travel industry sites.

All of the sites have what I feel is good linkage. I found anywhere from 200+ to over 140,000 references to their URLs in Google (I did not check other search engines because I specifically wanted to see what Google "knows" about those URLs). Only two sites had thousands of inbound links.

Nearly all of the site operators report having had more than 1 top ten ranking for targeted search expressions. The sites with the most top ten rankings seem to have retained more visibility and traffic than the sites with the fewest top ten rankings.

A few of the sites have hidden INPUT elements on their pages. In the past, this has not been an issue because hidden INPUTs are used to transfer data from HTML pages to CGI scripts. I don't believe this SHOULD be an issue, but I'll dicuss this in my HYPOTHESES section.

Finally, most of the sites come up first or at least in the top five results for the full text of their titles (without quotes -- I used only FIND ALL searches, the most common searches used). A couple of sites have real issues with their title tags. You should be able to rank in the top ten for most title strings, except in a very few industries.

SOME ADDITIONAL, NON-SITE RELATED FACTS

As I mentioned above, Matt Cutts has suggested on his blog (which I have not had a chance to look at today) that at least two more phases of this collective update process are yet to be rolled out. Look for one next week and one for the week after. He may alter what he posted yesterday, or provide additional information.

One of those updates involves what GoogleGuy refers to as "a new binary" -- that is, a new program (that is usually what programmers mean when they refer to "binary" or "binaries"). What does the new program do? Matt didn't say. I haven't looked for GoogleGuy's comment, but I doubt he provided details (he rarely if ever does). So, the binary could be an updated crawler (we should see a change in version number in our referral stats if that is so), an updated query processing tool (the software that actually generates the search results), an updated database administrative tool (back end stuff), an updated indexing tool (back end stuff), etc.

The binary is apparently being implemented one data center at a time, so this rollout is going to behave very much like a classic Google Dance (in my opinion).

Finally, let me point out something about Google's data architecture. This is a high-level description, but I think it's important.

Most people know that Google operates from multiple data centers located around the globe. Each data center replicates most if not all of the data found in the other data centers. Each data center sends out its own crawlers. Each data center hosts hundreds if not thousands of computers that work together to provide the services Google offers.

The Google database is so large that no single computer hard drive can hold it all, so Google has devised a complex data structure hierarchy where the data is distributed across multiple servers in what they call "shards". A shard represents or contains only a fraction of the database. And shards are replicated across multiple servers because every data center loses one or more computers each day.

A perhaps not-so-well-known fact is that some Web sites are recorded in more than one shard. I mean that, if we treat all duplicate shards as a single shard, there are still SOME Web sites that will be found in multiple shards. I'll refer to the collections of shard duplicates as "unique shards". i.e., if a data center has 5 copies of a shard, I regard them to be 1 unique shard. (NOTE: Google may not treat the shards this way.)

When a query is processed, the shards are individually queried for results. A Web site that is recorded in more than one unique shard stands a higher chance of being returned for a query than a Web site that is recorded in only one unique shard.


HYPOTHESES

There is really no way to confirm these hypotheses. I present them in order of my strength of belief (that is, the one I feel is most likely close to the truth is first, the most wild, radical one is last).

1) Google may have taken some of the shards offline, or has reorganized them, or has in some way altered affected some of the shards' abilities to participate in the normal querying process. It may be that each shard has a special managing program and that some of them have been altered, given lower priorities, etc. It may be that the shard data is simply not available.

2) Google may have implemented an algorithmic filter that looks at external style sheets and attempts to downgrade (but not fully penalize) sites which may be doing sneaky things. When I say "may", I speak from a programmer's point of view. It is extremely difficult to write code that can evaluate other code and deduce what the intended end result should be. CSS code is more like a list of definitions, but those definitions have to work with something.

3) Google may have altered its ranking algorithm to favor sites that meet certain criteria based on age-related factors (this has been a much anticipated and much discussed proposition, but it's too soon to know).

4) Google may have seriously broken its ranking software


ADDITIONAL THOUGHTS

Since I don't work for Google, I have to engage in what is essentially wild speculation. Having read much of the wild speculation posted in various forums, chiefly by the same people who have wildly speculated in the wrong directions many times through the years, I am inclined to dismiss all the usual hypotheses as being unrealistic and out-of-touch with Google's principles.

My wild speculation is based on Google's behavior, current and past. They appear to do things to shards when they roll out updates. Why? How? I don't know. But everything gets tilted at a weird angle for a while and then it rights itself. This is indicative of a shard-related situation, particularly for this update, because I have looked at a lot of query results and noticed something interesting.

Google is listing a large number of what I personally would regard to be "high quality" sites in top ten positions. I mean, they showing far more high quality sites than they used to, and I don't believe these listings will continue as they are for long.

Most of these sites are indeed relevant for the queries. Most of them are indeed large corporate sites, or mega sites, or otherwise high visibility, high profile sites. They are not simply operated by people conducting online business from their homes. They are not necessarily well-linked sites, but they are sites that are branded, recognized, and established by (in many cases) major, well-known corporations.

Google looks at what is important to the larger audience. Some people have described these new entries into the top ten listings as "spam sites", but the designation is absolutely ridiculous (I looked at one specific example where someone said the listings were spam, and I found nothing but relevant content from high-visibility corporate Web sites like Priceline, Travelocity, Hotels.com, etc.).

My feeling is that Google has implemented some sort of post-crawl filter that is or has been applied to only some of the shards. My feeling is that the shards which have not yet had their data filtered have been given a lower priority for placement in final search results.

If that is the case, then as all the shards are updated, people's listings should return to where they were before the weekend.

Unless someone from Google speaks up and says, "Yes, this is what we're doing" or "No, Michael is blowing smoke out of his ears", we'll never know what is going on. But I think the drops in listings will only last a short time. Ultimately, the only sites that should be suffering true penalties will be sites that trip the new post-crawl filter.

That's really all I can say right now. Thanks again to everyone who let me look at their sites. I'm glad to see that so much of the good advice here at HighRankings has been put to use. I'm sorry to see people lose their rankings, but this time I don't think the changes in results have nearly as much to do with factors we can affect as with Google's internal operations.


ON EDIT: I know some of you have clients who have suffered. I strongly advise you NOT to present this speculation to your clients as an explanation. Not unless you're really sure they can distinguish between what is essentially (my) BS and real fact (of which we have very little).

[lisety]

I hope you're right when you say that those of us losing ranking for no apparent reason will get it back soon. This Jagger algo is definitely not making my day
Lise

[Jill]

Interesting speculation. I don't understand the whole "shard" thing (did you make that up?), but I find the part about the branded sites showing up first to be a possible key to all this.

Perhaps they're testing out their TrustRank algo? That would certainly explain big brands showing up first.

[Michael Martinez]

Interesting speculation. I don't understand the whole "shard" thing (did you make that up?),...

No. Google engineers have publicly spoken about the shard concept. I had never heard of it before.

...but I find the part about the branded sites showing up first to be a possible key to all this.

Perhaps they're testing out their TrustRank algo?  That would certainly explain big brands showing up first.

Well, after I posted that, I thought about my On the Googleness of Being paper, which I posted on Spider-Food in February.

There I wrote:

Hence, fresh content can now be shot straight to the top of targeted listings provided it's posted on a Trusted Content Site which has little if any competition from other Trusted Content Sites.

I followed that with four definitions that were picked up by some other sites:

Trusted Content Site - A Web site of long-established reputation in the Google index.

Reputation - The track record of a Web site according to Google's measure.

Child Inheritance - The instant or near-instant ranking success awarded to a child page because of the parent site's credibility as a Trusted Content Site.

Listing Inheritance - The transference of a search ranking from a static page to a dynamic page where the static page includes user-visible language indicating a site has changed location (URL).

Google proceeded, shortly thereafter, to trash a lot of the so-called Trusted Content Sites in their listings for several months. A lot of SpamAd sites began taking over the SERPS. In June, the SERPs began reverting to more stable, traditional results. For example, the rankings I had lost in March came back, some better than ever.

In late July Google began filtering out the SpamAd sites, and they did a very good, quick job of it.

I'm beginning to think they may have reorganized their shards. Suppose they now have shards which are composed primarily of what we call Trusted Content Sites? The TrustRank factor, while still highly speculative, could determine which or how many shards your site gets into (or is otherwise associated with).

I can see how it would take several months to reorganize the shards, if they is what they did earlier this year. They would not have been able to do them all at once. They could have been shifting a tremendous amount of data around. But that would explain why so many listings vanished, and it would explain why so many sites began reporting cache dates going back as much as two years.

It's all hypothetical from my end, but it sure seems to fall into place.

[Michael Martinez]

I hope you're right when you say that those of us losing ranking for no apparent reason will get it back soon.  This Jagger algo is definitely not making my day 
Lise

I strongly encourage people NOT to use those silly update names. They don't provide any chronological reference. This update is best described as the October 2005 update because it is rolling out over a 2-3 week period.

[lisety]

Since Google asks you to use the name Jagger in contacting them about the updates, I think it makes sense for the rest of us to use that name as well.
Lise

[Michael Martinez]

Since Google asks you to use the name Jagger in contacting them about the updates, I think it makes sense for the rest of us to use that name as well. 
Lise

All you will do is confuse new people as they start to read about this update in the past tense. Most people already seem to be unclear on exactly when the "Florida" update occurred, and yet it is still mentioned often in many contexts.

Use the Jagger1 reference to correspond with Google, but keep in mind you are doing no one a favor by using that reference in general discussions where many people who have no idea of the workings of these update namings came from, how they are handled, or how to find out which updates occurred when.

How many people today remember when and where hurricanes Donna, Betsy, and Vicky hit?

[SanDiegoMedia]

I'm beginning to think that the chosen name should have been "Jager". It seems a lot of sites are f-ed up, and only time will get them back to normal.

[Michael Martinez]

Hm. Well, to offer a compromise on the naming issue (which should not be allowed to derail the appropriate discussion), I noticed someone at SEW started a thread title "October 2005 Google Update 'Jagger'". I can live with that if people will use it consistently. If the update names have a context that help people place them correctly in reference, I have no objection.

[Jill]

Actually, High Rankings has their own naming system. We don't like those silly WMW names either, Michael.

In the past we had Gladys and Jason. Geez, I can't remember some of the others.

This one I am going to dub "Scrooge" because it is an obvious attempt to spoil the holidays for many businesses.

[juliesjewels]

This one I am going to dub "Scrooge" because it is an obvious attempt to spoil the holidays for many businesses.

"Scrooge" best describes what is happening right now but the term "unfair" fits too because that is how I and probably many other small business owners feel about this Google crap.

Thank you Michael for taking the time to look over various sites in an effort to offer an explanation. Many of us read this forum daily and value the input of everyone willing to contribute.