Google Dance Syndrome Strikes Again, Part 1

There’s been a new outbreak of the Google Dance Syndrome. Last month, it caused some Web sites to lose top positions for some search terms. Unlike previous outbreaks, this time a “cure” exists that makes it easy to compare results from “old” to “new” Google. These comparisons have some marketers convinced recent changes at Google are designed to boost ad sales — a charge Google flatly denies.

For sites to lose rankings on Google (or any search engine, for that matter) is nothing new. Search engines constantly add and remove pages, as well as alter the algorithms they use to rank pages. There’s an extra degree of obsession with Google due to the large amounts of traffic it can deliver. Of the four most popular search engines: Google, Yahoo, AOL, and MSN Search, Google’s results are used on all but MSN.

Every so often, Google makes massive changes to how it ranks pages. When these happen, they’re usually accompanied by complaints from some search engine marketers and Webmasters that Google’s ranking system has deteriorated. Judging this is difficult. Often, those injured by the changes indeed point out examples of searches where Google fails in terms of relevancy. Examples of where Google’s new system succeeds also abound.

My previous column on the subject explores in depth the difficulties of measuring how well Google performs following such changes. Unique to this latest change is the fact there’s now a method for comparing “old” and “new” Google, something not previously possible.

The Filter Test

Google Watch‘s Daniel Brandt discovered including a made-up word as part of a search may cause Google to display radically different results. Since his original post on, hundreds, if not thousands, of site owners tried this. Based on their reports, the “filter test,” as it’s become known, seems to show how Google previously ranked things.

Here’s an example of the filter test in action and why it works.

  1. Search for “laptop rentals.” Thousands of matches tell you Google knows plenty of Web pages that contain both words.
  2. Search “laptop rentals ghghghghgh.” No matches, which tells you Google knows of no Web pages containing all three words. We already know from step 1 there are plenty of pages containing “laptop” and “rentals.” Really, it’s “ghghghghgh” that doesn’t exist on any of these pages.
  3. Search for “laptop rentals -ghghghghgh.” This should bring back exactly the same results as the search for “laptop rentals.” That’s because we’re asking for all pages that contain both “laptop” and “rentals” (which we know exist from step 1). But we command Google to exclude any pages that also contain “ghghghghgh” (the “-ghghghghgh” part of the query). As we know from step 2 there are no pages with “laptop” and “rentals” that also contain “ghghghghgh,” we should get the same results as step 1. Instead we get very different listings.

Why does this happen? One popular theory is Google uses a new filter to prescreen results for “money words,” or searches for which it hopes to sell AdWords paid listings. You can understand the popularity of this theory by looking at the before and after for the “laptop rentals” search. Before, you got mostly businesses that apparently specialize in laptop rentals. Afterward, they’re all gone, replaced primarily by university Web sites discussing laptop rental programs for students.

Ah ha? Well, more like hmmm. There are plenty of exceptions. Some people did run ads before the change and found they lost “free” rankings. There are also people who maintained their top free rankings after the change, despite the fact they never bought ads. The changes have had positive and negative impacts on all types of sites. Big sites don’t necessarily trump the mom-and-pops.

Post-Dance Analysis

In part two, I’ll look much more closely at the fallout from Google’s latest changes. I’ll explore my thoughts about why I believe we’re now seeing two different Google ranking systems being used, rather than a filter being applied to a single system. With luck, we’ll have some comments from Google itself.

In the meantime, if you want to explore more now, there’s a variety of existing resources.

Scroogle: From Google Watch, this site lets you easily see what pages are no longer in Google’s top 100 results for any search you enter. Conduct a search, and you’re first shown what’s missing and where it formerly ranked (based on the filter test), followed by the current top 100 results.

Brandt tells me the results are created by sending the query you enter along with three made-up words, then comparing these with a search that doesn’t use the fake words. So for a search for “used cars,” it might send something like “used cars -dhdhdhdhdh -hdhdhdhdhd -jfjfjfjfjfjf.”

Why more than one made-up word? As was later discovered in forum discussions, only one made-up word is enough to bring back “old” Google when entering a two-word query. But for a three-word query, two made-up words are necessary. By using three made-up words, Brandt hopes to cover situations in which someone enters two to four words to test.

Be advised: What Scroogle reports may not exactly match what you see doing the filter test yourself. For instance, Scroogle reports for “used cars,” “new” Google gives the Kelley Blue Book positions one and two, Edmunds positions three and four, and positions six and seven. However, a search on Google for either 100 results (the setting Scroogle uses) or the regular 10 results setting actually gives each of these sites only one listing in the top 10.

The name “Scroogle” seems to be a play on “Scrooge” and “Google.” Many site owners complain the ranking change hurts them during the Christmas shopping season, so they consider Google Scrooge-like.

WebmasterWorld: The popular forum has several threads on the topic of changes. “Update Florida – Nov 2003 Google Update” is open to the general public. The “Florida” reference is the name WebmasterWorld gave the latest Google index. Like hurricanes, updates are given new names whenever one emerges.

Be forewarned: The thread is long, divided into four parts, and involves upwards of 3,000 posts. For every definitive explanation offered, you’ll generally find an exception posted.

Paid supporters of WebmasterWorld can also access some private threads. An interesting one is “Update Florida – what happened?”

Other forums: If WebmasterWorld’s discussions aren’t enough, has a “Current Algo Update-Dance” thread. Over at JimWorld, you can read the “Google’s new filter” discussion.

Articles: Barry Lloyd has a good summary of forum discussions in Been Gazumped by Google? Aaron Wall has an earlier summary in Google Sells Christmas. Gord Hotchkiss offers a summary in Florida Fever: The Google Update Uproar. And if you want a visual guide to the forum theories out there, see Vaughn Aubuchon’s Google Florida Update – Dictionary Filter Flow Chart.

Related reading

Brand Top Level Domains