17 replies to this topic

[copywriter]

Hi,

Just a note about something strange that happened. After the infamous "Florida" update, I - like many other people - found myself gone for a couple of keyphrases. Over the next week or two I was gone for just about all my keyphrases.

For some reason, it dawned on me that I still had a 301 redirect point from my *old* domain to my new domain. I decided to remove that just to see if it would have any effect. This was yesterday.

Today, I'm back in the top 5 and top 10 for ALL my keyphrases again!

Now, I know 301's are supposed to be search engine friendly, but this makes me wonder if there's been a change in Google's algorithm regarding redirects.

[Grumpus]

Hmmm. I don't think that would be a part of the algorythm. Sounds more like it's a nasty belch in their DNS database or something along those lines... Weird weird weird. Every scenario I'm running in my mind right now as to how this could happen just doesn't work out.

This does seem to go along with my previously unpublished hypothesis that this wasn't so much a change in algo, but rather a major backtrack to an older way of doing things. The problem is that I can perfectly well see that a 301 on another page could cause the problem - it simply doesn't make any sense that removing it would fix it.... Weird. Weird. Weird. Have I said that?

Fun! I love puzzles. Granted, a plausible scenerio is going to pop into my head just as I'm trying to go to sleep tonight and I'll be up all night trying to work it out and see if that's what really happened....

G.

[Jill]

Which site is back, Karon? The one old one or the new one?

Perhaps you didn't do the 301 correctly? I have a 301 from RW to HR, and I'm reaping all the "benefits" that I should. Googlie sees them as one and the same site.

Jill

[Grumpus]

I've got a scenario that "works" now, but it has a lot of ifs in there... It could work, though...

As I alluded to above, I suspect that the major upset in the Florida update wasn't so much a bunch of new things being added, but rather taking part of the data within Google and actually stepping backwards in time with it. The discrepencies come because we're dealing with brand new "on page" factors, but several month old "off-page relational factors". This is stuff like PR, inbound link text, outbound linking power, common local-rank factors, etc. (Some of the things on that list are technical - I won't bother to explain them because understanding them won't affect the ability to understand this theory).

In essence, I'm of the mind that Google keeps a separate database of information about each page and how it relates to other pages on the web and how other pages relate to it. By using this database, it's easy to take a new page that someone puts on the web and applying some assumptions from the relational table in respects to the page(s) that linked to it and thus got the new page "found" by google. Those assumptions could be a fairly accurate guess of the eventual "real" relationships that the page may develop.

It is this technology (described above in an overly simplified manner) which allowed Google to go to the constant update model and not have to do the monthly "dance" for the relationship data to be merged in with the physical on-page crawl data. A good enough "guess" could be made to make it okay to rank the page for a few weeks before actually applying the "real" data to it.

My theory (that I've had since Florida kicked in), is that there was some horrific flaw in the whole thing. Most likely, something in that relational data kept getting skewed over time - and with each update that went by, the skew became increasingly apparent - geometically compounding itself. A flaw like this would be easy enough to fix so that it doesn't happen anymore, but you are left with an entire database of skewed and errant data.

The only option is to go back to some older, unaffected data and start again - and thus Google begat Florida and saw that it was good. Every page was working on "guessed" data, but those guesses were coming from non-flawed (just old) relationship data.

So now, with this 301 case, the old relational data was for the old page. The Google programmers, when they pushed the button to start the wayback machine, anticipated that some sites would be gone or moved to a new server. So, if the location on the old locational data was throwing a 404 but the URL still had new crawl data in there, then it would just assign that old data to the new location. But they forgot that 301's were used when moving servers, too. Sure, it's the right thing to do, but no one does that when they move, so it didn't pop into their heads. When Florida hit, the relational database didn't send the "Hey, the page isn't here, let's compare" error that they'd anticipated, it just threw the standard "it's moved and it's moved here" error - which didn't trip the flag to tell the two databases to get back into synch with each other. It just sort of sat there thinking everything was okay.

So, when you removed the 301, a 404 came up and it gave the error it needed to sort itself out - and so it did.

----

This theory up until the paragraph that begins with "My theory (that I've had since Florida kicked in)" is rooted pretty solidly in how Google works. There are surely two (or more) databases and one definitely contains relational data that is separate from actual on-page factors. Exactly how it all works is speculation.

After that point in here, it's all wild guesses that tickled my fancy and worked out with a certain sense of plausibility (given that the set-up I laid out is at all accurate).

In other words, take it for what it's worth. Chances are that one or two things in there are exactly right and the rest of it is a complete load of hooey. It's fun to play, though.

G.

[copywriter]

Hi Jill,

Well, I had several people check it and they said it was right. When I first did the 301 (what? 4 months ago?) it worked great! MW was showing in great position and KTA was no longer in the SERP at all.

But then... for reasons that Grumpus explained (thanks for your thoughts, G) or other ones... all of a sudden MW vanished. Little by little each page was dropped from the SERP.

After I removed the 301, I was back in full force within 24-hours.

Now... the other posibility is that my host for KTA changed servers and upgraded services during this time. *Perhaps* something on their end disagreed with Google???

At any rate, I'm back. I've just put up two "We've moved, click here for the new site" pages on the old KTA site (on the most popular pages people will click to). The rest will have to sort itself out later.

[Grumpus]

the other posibility is that my host for KTA changed servers and upgraded services during this time.

Google is notorious for not updating their DNS cache frequently enough. It hasn't happened much recently, but a year or so ago, it was common to hear that people changed hosts/servers and it'd take Google three or four weeks to find them, even though nary a URL had changed.

This is a possibility I had thought of and it's plausible - but only if you accept that removing the 301 page and the sudden re-emergence were a co-incidence. Removing that simply can't help something find a new location - unless it already knew the new location in some way.

In reality, my theory up there is a mostly a load of horsepucky. Certain aspects of it are rooted in truth (and I do believe that Florida was more of an emergency backpedal due to an oversight than a progressive surge to improve ranking quality). Fact is, no one will ever know exactly what happened there, I don't think. It is fun to make up conspiracy theories, though.

G.

[qwerty]

I had something odd happen recently, not with a 301, but a changed IP address.

We've been making major changes on a client site that only had a few pages listed in Google. I noticed about a week ago that a search on the company name brought up some site I'd never heard of, so I went to the URL and found an exact copy of the home page, complete with links to interior pages on my client's site.

At first I thought this was a very stupid example of theft, but it turned out that my client's host had changed their IP address, and that address had been used by a site that no longer exists. We contacted them and they took it down, but some damage had been done -- Google had picked up the changes on the other site before they'd caught them on ours, so now we were apparently guilty of duplicate content.

The odd thing is that (even now that the page is no longer up), if you request Google's cache of the non-existent page, you get the copy of my client's page, with the header at the top indicating that this is Google's cache of my client's page -- with their domain name rather than that of the nonexistent site.

[copywriter]

Weird!

Well, G, while a good mystery is something I enjoy, I'm not willing to put the 301 back up to see if I disappear again We'll just have to keep wondering.

[Grumpus]

Qwerty - sounds like yours is just a matter of waiting for the crawl to start throwing 404's on the newly removed site and you'll start to come back to life. It appears (from all of my experiences with duplicate content) that this filter is applied at search-time. In other words, certain off-page factors vary from search term to search term, so when a search is made and two pages that come up identically, the one that gets filtered will vary depending upon which one of them ranks better for the search term being used. Thus, you may not even have to wait for the PR/Offpage factors database to update (which still does and always has happened monthly) for some pages to start showing back up. The floogey variable here is the IP change because Google keeps that data (it seems) in yet another completely different database.

I'm not willing to put the 301 back up to see if I disappear again

<- The closest available emoticon to a "chicken".

G.

[qwerty]

Yup, and wait we shall. But I still find it odd that Google would identify the cached copy of one page as the cached copy of another. If they know it's the other, why do they display it at all?

Let me show you: go to Google and request cache:www.lionspawsanibelplus.com

[robwatts]

Interesting take on the 301 Grumpus, there is quite probably a healthy degree of truth in your theory.

Second guessing what those Googlies do is a fun part of the game. Its always good to think in terms of "If I were a search engine looking to protect my algo I would do these things..."

Its certainly in their interests to create as much mystery as possible. I liken it to a game where you have a pack of cards spread out face down on the table, if you invited a friend to guess where the ace of clubs was, then they'd eventually guess correctly, if you never moved it from its spot then they'd guess correctly everytime and the game would get boring. By moving it around every so often it gets harder to find, and you keep the game alive.

I'm not willing to put the 301 back up to see if I disappear again  We'll just have to keep wondering.

Karon, once upon a time a useful test would have been to put the 301 back in and see where things went. If you disappeared again then it would be reasonable to assume that it was the action that caused the reaction. The thing with the Google of today is that very few people know what Google is doing, lots of suppositions and theories of course, but no real concrete evidence that can be applied query wide. Hilltop, TSPR, localrank, a blend of all three, adrank

The Google of today may well have a variety of algos and db's for a variety of serps, which subtely alter based on the time of day, day of week, geographic locale, length of query, type of query etc etc I like to think of G as a multiplicitous entity that enjoys playing a cat-and-mouse-catch-me-if you-can kind of game.

Mice like cheese, some mice only like certain types of cheese. There are 1000's of different blends and varieties. If the cat wants to be sure of catching the mouse then the cat needs to make lots of different cheese and let the mouse choose for himself. Sure itl'l take a while to place the cheese in various positions and some might never get eaten, and the mouse might have already had his fill elsewhere, so the cat better make sure he does everything to ensure that the mouse gets to his piece first by making it as attractive as he can. How to do this? well, the mice read all about cheese in various places, some cheeses are over-advertised or hyped in IP static cheeseville, making the mouse think twice, some cheeses are similarily hyped, but from such a diverse source of cheeseland locations and properties that the mouse deduces that the particualr cheese may really be something special, so checks it out. If it likes what it tastes itll come back for more, so best make that cheese as tasty as possible. If the cat can't give the mouse what he wants the cat will never catch him. The mouse also knows that the cat is smart and will try to catch him, so the mouse varies his tastes occasionly, he is also lucky that the cat just wants to play, although the mouse is delicate and thinks he will get killed if he lets the cat play too long..

OK, back to work.

[Grumpus]

That could very well be a symptom of the "backpedal" portion of my theory above. As I say, I don't know where the DNS portion of Google's database is stored, but it's most likely not in the same section as the "on-page" factors - which would include the cache.

The section that contains the "how this page relates to everything else in the world" data is now old (I suspect that it's most likely 3-5 months old). The cached information database is still brand spanking new, though. That one knows that the dog's choice site is in a new place.

Think of it like the GPS system in your car (or if you're me, the GPS system in the car of the guy who lives across the street). Last week they moved an onramp to the highway two blocks south. You've seen it before, but you can't remember how to get to the new onramp. When you close your eyes (i.e. check your cache) you've got a picture in your mind of what the new ramp looks like, but you have no idea how to make use of it because you don't know how to get there.

So, you fire up the GPS (the relational database). Unfortunately, the GPS uses an old database and keeps insisting that you've got to go to the original location of the onramp. Now, you've got the problem. Your mind/cache knows that the GPS/Relational Database is wrong, but it's impossible for you to teach it the new location because you don't know where it is, you just know that it is.

And so, you've got to drive around for a bit until you find the ramp. Now that you know where it is, you can program it into your GPS and now your cache and the GPS match up and all is good.

G.

[Jill]

After I removed the 301,  I was back in full force within 24-hours.

Karon, still not clear on which site is back. Is the KTA site showing up in Google, or the MW? If it's the MW one then I'd just stay with whatever you're doing. If it's the KTA one, obviously, you'd prefer it the other way around (I imagine!).

J

[copywriter]

Sorry... it's MW.

[Jill]

The Google of today may well have a variety of algos and db's for a variety of serps, which subtely alter based on the time of day, day of week, geographic locale, length of query, type of query etc etc I like to think of G as a multiplicitous entity that enjoys playing a cat-and-mouse-catch-me-if you-can kind of game.

Rob, most definitely!

Which explains why some people swear that the latest algorithm penalizes the same keywords in hyperlinks, while others swear it's a better strategy than ever! Same with Headers, Title tags, keywords in the body copy.

For everyone who says that a certain method works better than ever, another person says it's now being penalized.

Google must be laughing their little asses off at that. Too bad they haven't quite got it to work so that the results are any good!

Jill