What was "Florida" anyway?

[I, Brian]

Something I really really meant to ask people at the SEO Roadshow - but instead spent a lot of time simply meeting people - was the issue of Florida.

Was there ever a consensus on what it was?

Did anyone here even care to speculate?

For my own 2c, it seemed like Google threw in not one single change, but a few all at once.

As for the main engine behind the change - I've seen some interesting arguments by Dan Thies about it having been TSPR, and Phil Craven about LocalRank. Personally, I always thought something based on Hilltop's "experts and authorities" seemed a particular suspect - or maybe it simply seemed a fascinating idea to implement.

Either way, does anyone here dare speculate on ideas of what the Florida Update was or was not? Could be an interesting discussion at SEW, unless considered worn out at WMW.

[mcanerin]

I agree it was more likely a combination of filters and changes than just one.

That would also explain some of the (obviously unanticipated by Google) behavior.

Take a bunch of concepts that may work fine in isolation, then mix them up in a complicated, self referencing, nearly "organic" system like the web, and interesting things are likely to develop.

I remember reading a science fiction book a long time ago, where one of the premises was that the internet had become so huge, interlinked and pervasive that it began to actually have "weather patterns" of slow access, strange results and so forth...

I don't know if that will ever happen, but referring to major changes in algos using hurricane terminology always struck me as an interesting coincidence

Ian

[seomike]

I was under the impression that G updated it's technology on semantics especially how it deals with word stems.

I think that's why people were dropping. Not because of filters but because as G was regrouping sites with similar root words more sites were added to results that might not have had an exact phrase of the keyword serached for.

Example: Do a search fo "Trailer hitch ball sacks"

You'll notice that G will bold the stems of the words in the results like

hitches
hitching
balls
trailers
sack

That's pretty nutz in my book

[Jill Whalen]

I think it had something to do with trying to keep more commercial sites out of the regular listings in favor of informational stuff.

They're still doing a pretty good job of that for many, many phrases. And I truly think that they are hoping to eventually go completely that way. All info on the left, all commercial on the right (and paid for).

[Marcia]

Brian, some of the Roadshow type guys (and gals) pretty much don't pay a whole lot of mind to updates that they generally talk about. A lot of people just wait 'til they're all over to even look.

combination of filters

Ian, that's most plausible which is why most everyone really couldn't figure it all out. Though added on to that, tougher to beat than filters, is that just maybe there were new "requirements" to rank that weren't being met.

What was shared back then that could be the most important bit of wisdom ever dispelled by the Big G:

"The things that used to work just aren't working any more."

A little historical recap, as best I can remember - it's been almost a year:

They'd introduced and announced stemming on their site, and there was a lot of talk about LSI - but little specific about how to implement it. One helpful bit was an on-the-fly interview with someone from Google with a hint that Google was looking for "breadth of content." But IMHO that goes further than just within a site itself, linking factors are way too prominent for that. A "corpus of documents" can theoretically include a site plus the sites linking to them and the sites linking to the sites linking to those - and sites linked out to.

Google (in my mind) is about "relationships" so it isn't hard to assume that.

Then there were a few factors evident that had been mentioned in Hilltop, so that theory was jumped on - but it's far from likely that Hilltop was implemented, though there can always be individual elements from an algo utilized.

There were clues, some of which were evident when looking at some specific spaces on a local level and seeing what got hit and analyzing. Namely, real estate and web design. But few conclusions, though when some "fixes" were suggested and implemented sites came back. Those were VERY productive. And then the filters started to be loosened up over time.

Some sites were restored, but some never came out of it. And oddly enough some (like me) believe that there's a striking similarity in some respects between the Florida symptoms and some of what's being evidenced with the "sandbox" phenomenon, although this is about ten times worse - except for those who know how to avoid it. And there are some who do know, which imho proves that its algorithmic.

My *intuitional* suspicion - and only that, *not* factual - is that, while not implemented as such, there may be some elements contained in the Local Rank patent issued this year that are coming into play. And with Florida there did seem to be a connection with linking and for lack of a better way to put it, semantically based connectivity. Very evident in local searches once the "fixes" were put into place.

Ranking search results by reranking the results based on local inter-connectivity

Emphasize - NOT FACTS. Only anecdotal clues and right-brain speculation, which may or may not be correct at all.

interesting arguments by Dan Thies about it having been TSPR, and Phil Craven about LocalRank.

Saw those, and I got into it a bit with PhilC about the LR - which I'm inclined to favor to a degree, and he's kind of disagreeing with me. He's smarter than I am, so I have to kind of work at it to keep up with him.

[Mikkel deMib Svendsen]

What I personally find amusing about all the speculations after updates like Florida are the complexity of explanaitions layed out. What I often find is that you can explain the same happenings much simpler.

It kind of reminds me of the way people used to look at our universe back in the middle ages - with the earth as the center. Even back then they where able to predict the position of stars and all that but the calculations and math behind it where just so complicated. After we decided the sun was the center the caluculations just became so much easier. The results was still the same but it was a lot easier to get there.

What makes evaluating search algorithm updates even more diffucult than the stars and the universe is that it's not all logic. What if only 80% of the Florida update was in fact an "on purpose" change of the algo and 20% was simply a human error. What if 80% was human error? How do you compensate for that in evaluating a complex "organism" such as a search algo?

On top of that comes language dependency and the increasing dynamic nature of search algos in general.

As an example we hardly saw ANY impact on Florida here in Scandinavia. That imply to me that the update was, at least in part, linguistically based (Google knows far more about English than they do about Danish, Swedish or Norwegian).

[sem4u]

Aaron Wall & Barry Lloyd wrote some interesting articles about 'Florida':

http://www.search-marketing.info/newsletter/buyads.htm

http://www.searchengineguide.com/llo.../1125_bl1.html

[I, Brian]

If I remember right, there was a statement by GoogleGuy at WMW during the first news of the update, stating that it was occuring in two parts, with the second part following a few days after the first. That's when it all hit the fan.

I think what Barry argued - whether he meant to or not - was a very strong case for something like Hilltop having been applied as one of those parts. Reliance on a selected set of "authorities and experts" (not least .edu and .gov sites as automatically qualifying) to provide results for a narrow range of search terms selected, seemed the most obvious reason for the shelving example. The Austin Update seemed to be an expansion of the set of search terms being targeted by a Hilltop-esque system. Doesn't make it true, though - as Chris Beasley at SitePoint effectively stated, Hilltop was such an old algo that it wouldn't make sense for it to be applied in it's original form, though.

The other part of the update that GoogleGuy referred to would IMO have been behind the issue of stemming, as a result of something from the Applied Semantics toolkit being applied.

However, that's just personal speculation - there's always room for more possibilities.

Originally Posted by Mikkel deMib Svendsen

As an example we hardly saw ANY impact on Florida here in Scandinavia. That imply to me that the update was, at least in part, linguistically based

Indeed, but that doesn't mean to say that you shouldn't be worred. After all, the Yorkshire Update is coming.

[Mikkel deMib Svendsen]

Indeed, but that doesn't mean to say that you shouldn't be worred. After all, the Yorkshire Update is coming.

I am never worred - I am excited

[rustybrick]

Originally Posted by Mikkel deMib Svendsen

It kind of reminds me of the way people used to look at our universe back in the middle ages - with the earth as the center. Even back then they where able to predict the position of stars and all that but the calculations and math behind it where just so complicated. After we decided the sun was the center the caluculations just became so much easier. The results was still the same but it was a lot easier to get there.

LOL, very true. Also they explained whales and all types of fish as sea monsters.

I never thought of this update frenzy in such a way. Thanks Mikkel!

[seobook]

Originally Posted by sem4u

Aaron Wall & Barry Lloyd wrote some interesting articles about 'Florida':

looking back at it, mine was in all honesty probably a bit wrong...I am more with Chris Ridings / Mikkel now and think as much as anything it was some poor implementation of stemming and other linguistic problems...especially since it was so drastic in English and not as bad in some of the other languages.

they appeared to do something with topical authority sites at the same time though. I remember at SES NYC someone stated (probably gonna missay the name) Pati Smith I think it was stated that some sites actually improved their rankings by removing links to industry hubs because the search algorithm may have been viewing their sites as an extension of the hub while they were linking through to it.