I remember well how, seemingly, one flick of a Google switch impacted the entire SEO community.
Google’s Florida Update of 2003 affected Google search results for the industry and end users.
I was at a huge gathering of SEOs who all had the same thing on their minds. I saw many almost grief-stricken faces of website owners who “bet the farm” on free traffic from Google for their online businesses, only to wake up one morning and discover their once top-ranking webpages had dropped – to somewhere close to the center of the earth.
It began in December 2003 at the now-defunct Search Engine Strategies (SES) conference in Chicago. A couple of weeks earlier, the SEO community had witnessed the most major shake-up in Google’s ranking algorithm, and they were still reeling from it.
Thousands of attendees, which included many first-timers, were hoping that they would get some sort of explanation or at least some answers from the SEO expert speakers. Or, more to the point, from the few Google representatives also speaking at the event.
The impact of this particular update was so massive that it’s still lore in SEO circles. Exactly 20 years ago to this day, the so-called “Florida” update “upended” the ranked results for so many queries.
This article will explain what caused this huge shake-up in Google Search rankings.
Why was it called the Google Florida Update?
Where did that Florida moniker come from?
“I was busy planning my next conference which was due to happen in Orlando, Florida,” said Pubcon founder Brett Tabke.
Yes, he gave the Google update a name, and subsequently, many of the following updates, in the same manner as hurricanes get names.
Describing the update as a “slaughterhouse,” Tabke still believes that Google targeted a certain kind of website.
And though many believed at the time it was specifically affiliate sites, he says that many of his own sites that were not affiliate linked also “got creamed.”
A brief history of search ranking algorithms
To get to the “why” of the story here, we need to have a basic understanding of the ranking mechanisms used in search. I wrote a more in-depth piece identifying and comparing ranking algorithms.
But for the reminisce that this is, I’ll keep it pretty high level. When Google launched in 1998 there was a huge amount of PR given to the “magical” new algorithm named after its inventor (and Google co-founder) Larry Page.
Heralded as the secret sauce that separated Google out from the rest, PageRank is a hyperlink-based algorithm.
However, in the same year, another algorithm based on link analysis called Hypertext Induced Topic Search (thankfully shortened simply to HITS) was invented by a young scientist at IBM called Jon Kleinberg.
Now a professor of computer science at Cornell University, his algorithm was strikingly similar in its approach to that of PageRank. The difference between them was also striking.
- PageRank is keyword-independent, meaning your webpages get a score regardless of the subject matter.
- HITS is keyword-dependent, meaning your webpage score is based on keywords on the page and in the user query.
But perhaps the most striking thing about both is – neither of them worked in real time.
Meanwhile, another young scientist by the name of Krishna Bharat was also working on a hyperlink-based algorithm called Hilltop.
Being aware of both PageRank and HITS, his approach was more similar to HITS, meaning it was keyword-dependent and considered the user query. Kleinberg’s HITS algorithm is based on a score between “hubs” and “authorities.”
Bharat’s Hilltop algorithm is based on “expert documents.” You’ll notice the early inferences to E-E-A-T.
So, we have three algorithms, all hyperlink-based and developed in 1999. What’s the importance of knowing this?
Well, 1999 was also the year that Bharat joined Google as a research scientist. And let me just drop the final clue here to where we’re going with this: in 2003 Google acquired Bharat’s Hilltop algorithm.
And certainly, implementing a new ranking mechanism into the mix is bound to shake things up – and it surely did!
Get the daily newsletter search marketers rely on.
Users generally only look at the top 10 results, so there’s not much room for repetition
Quite a lot has been written about the Hilltop algorithm in the SEO community. Some of which, it must be said, are written by authors who, seemingly, may not have read the original paper.
Described as a search engine based on expert documents, quite a lot of what it does is right there in the title. But the term repeated most in the paper is “non-affiliated.”
It’s because of that there is so much speculation that Google was simply doing something to target affiliate marketers. While that’s partly true, it wasn’t fundamentally why this methodology was introduced.
Back in 2003, Google (and other search engines) were very much a list of “10 blue links,” and there was often a lot of repetition. Frequently, two or three links could be from the same site and absolutely, for commercial searches, there could be multiple links from online affiliate marketers all pointing to the same product sales pages.
The purpose of the Hilltop algorithm was to have the top results from unique resources. So, as an example, even the presence of the same three octets of the IP address (suggesting the same host) is treated as an “affiliation.”
During a conversation with a Google engineer back then, I remember him using this analogy when talking about the number of affiliate marketing links that would often feature in the top results.
He said Google does not target affiliate sites, but rather than “directing users to the store” (effectively what affiliate pages do) we’d much prefer to “pop them right through the front door” of the store.
He went on to say that if there are several affiliate pages in the top results, basically it’s just one result as they all point to the same place.
Information retrieval (IR) science is based on satisfying an information need by providing information “about” a subject.
So, the more unique the top-ranking results are, the more likely they are to satisfy the information need.
As far as so many attendees of the previously mentioned conference were concerned, they just felt that Google was putting them out of business.
Here’s a true story from one of my sessions. Regardless of the session’s subject matter, the Q&A at the end would pivot directly to the Florida update.
One guy became very animated, shouting at me that I should rally everyone in the industry to unite. “And do what?” I asked him.
“Go to Mountain View and effing burn Google down,” he replied.
He then explained loudly to the packed room that he had been the number one result at Google in his niche for two years, and his business was booming. Then, after the update, there was no sign of him in the SERP.
I suggested that he, perhaps, should not have “bet the farm” on this one source of revenue to sustain his business. I also suggested that, yes, we should go to Mountain View and visit Google. And when we see either Page or Sergey Brin come through the reception, he should walk right up to whichever one it was, kneel down and kiss his ass “because he’s been sending you free customers for two years!”
Why Google launched the Florida Update
It wasn’t uncommon for me to find myself on stage with various representatives from Google back then. Marissa Mayer and I were frequently on the same panel when she was director of consumer products at Google, and we were at this conference.
So she took a lot of incoming from the audience. That said, earlier in the year in London, I had been on a panel with Google’s senior research scientist, Craig Nevill-Manning. Fortunately (for me), he was attending the Chicago conference, and I had secured an interview with him.
As you can imagine, with the senior research scientist, I felt I’d be the first in the industry to get the inside scoop on the volcanic Florida update.
Nevill-Manning invented something called Froogle, which went through a few transformations and is now known as Google Shopping. But the original Froogle was basically a kind of eBay with no fees.
I put it to him that the industry was rife with conspiracy theories about Google making these changes to the algorithm and hitting affiliates hard to keep them out of the SERP and force them to open AdWords accounts. To which he replied:
- “There’s no conspiracy. This was simply an effort to make our results better on Google. We determined, for certain kinds of queries, the search results had… Well, many irrelevant results. So, we’ve tried to weed those out by changing our algorithm. And whenever we change our algorithm, it’s going to hurt some people and help some others. This may be a larger change than many changes that have happened in the past.
- “But the bottom line is the engineers that have worked on that, and I know them very well, so I know their motivation, and they’re totally focused on making the top 10 results as relevant as possible for users. So, if you were inside Google, you’d know that these conspiracy theories just hold no water.”
He had to add that now Froogle was rolling out of Beta, those “free” shopping results would appear at the top of the page, above the regular results. I guess that was “what Google takes away with one hand, it gives back with the other.”
But that certainly didn’t prevent the “black hat” element of the industry from going to ground. As Tabke remembered, those that had whole networks disappear did the same themselves.
The paranoia level had them even putting fake names on their conference badges at the actual Florida conference so that Google reps wouldn’t know who they really were.
Could future Google updates have the same impact as Florida?
With AI, machine learning, neural information retrieval, and neural ranking mechanisms now replacing some of the old techniques, my guess is yes.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.