Search with Design

This question has resurfaced in the SEO-world, and in my world recently, and so I’d like to weigh-in. Here’s the latest situation presented to me:

My friend sent an email to 100,000 subscribers, of which 1,000 of them were GMail addresses, advertising “college widgets”. The link back to his site contained the anchor text “college widgets”. He went from ranking #10 in a Google search for [college widgets] to #1 in just a few days, then later he dropped back down to #4. I looked at my GMail account and the email page has a PageRank of the 7!

The first thing to understand is that the way the PageRank in your toolbar works, is that when it doesn’t find any data for your current page, it extrapolates an approximate value from the previous page you visited as well as the domain. Because the Google domain is a PR10, your toolbar extrapolates a PR7 for your mailbox.

So where did the resultant change in rankings come from?

Well, the page in question has a title tag whose first two words are “College Widgets”. The page also has 59,000 links pointing at it (according to Yahoo!) most of which are from the same domain, with the anchor text “college widgets”.

In a sample of 100,000 emails, I have to think that at least 1-3% would have the Google Toolbar installed. Add in the GMail subscriber base, and Google has 3-4 thousand pieces of user data to use about the page in question. I would estimate that amount of user data spread over just a couple days would be more than enough for Google to trigger a deeper crawl of the page, and potentially index or count more of the backlinks.

As RC mentioned, you can’t triangulate with only 2 points. Google uses user-data, it’s time to face facts and build your websites for users too.

The Article Distribution Service iSnare.com has been billed as one of the best tools around to increase a website’s presence. And I’ve been a big proponent of it since I first came across the service.

The idea is simple enough: submit an article to this service, it is reviewed by humans for quality and then gets auto-distributed to 1000s of article-aggregation websites, many on general topics, and a few on whatever topic you choose for your article.

After using it a few times, I began to notice that pages I promoted with the service would tend to rise in Google’s SERPs for my targeted terms, and then slowly fall back down. They would usually settle at higher positions than where they started, but I wondered why the Rome effect was so strong (that was a subtle reference to a rise/fall timeline).

So, I decided to study the Google results on fresh articles, and their mentions in search engines. I used the old trick of searching a unique phrase. On August 4th I used a unique phrase from each article on Google’s engine: 0 results. I then submitted both articles to iSnare for distribution. On August 8th I got an email that both articles had been approved and syndicated; a second Google search revealed 0 results for both.

0 results again on Aug. 9th. Then on Aug. 10th I saw the first signs of life: 7 results for Article 1 and 8 results for Article 2. By Aug. 15th, Article 1 had 437 results, and Article 2 had 458 results. There are two points of note here:

Point 1: I submitted both articles under the same category. They were approximately the same length (around 450 words). I submitted them on the same within minutes of each other, and yet Article 1 lagged behind Article 2 for some reason.

Point 2: At this point (Aug. 15th) there were no supplemental results for either article. All 400+ results were fully viewable in the main index.

On the 16th of August the dupe filter must have kicked in on Article 1, because supplementals appeared and total results dropped to 361. Article 2 continued to thrive with 556 results on the 16th, with still no supplementals showing.

Eventually the dupe filter must’ve kicked in on Article 2 as well, and by August 30th, both result counts were below 50 (39 and 34 for 1 & 2, respectively).

As of today, Big G shows 11 results, of a total of 16 for Article 1 (so, approx. 4 supplementals). Article 2 fared better in the end, today displaying 16 results of 22 total (so, approx. 6 supplementals).

The [recently exported] PageRank for the top 10 results on each article range from 0-2, with the majority being 0 (and 2 N/As!).

So now some theories:

1. Article 1’s target phrase was more competitive than Article 2’s. My theory is that the more competitive an area, the greater number of filters (or in some cases, reviews) a page must pass to become part of the index. This is explained best in the theory of long-tail keywords, where phrases that don’t mean much in a marketing sense have a lot of impact on John Q. Searcher.

2. To compete with social bookmarking, Google needs to be buzz-aware. When a site creates a certain amount of buzz (linking, textual-references, etc.) Google needs to get in there and evaluate it for ranking. It will weight these sites with additional trustrank to get on top of the coming wave. A second (and potentially third) filter will later decide if the page is worth keeping in the index. Possibly by analyzing search volume for a phrase vs. the amount of “buzz”.

What might a takeaway be from this experiment? In my case, the combination of the “buzz” created with the article distro, plus the already-established authority (or Trustrank) of the site was enough to put the [brand new] pages I was targeting into the top 10 for their intended keyphrase.

As with most SEO activities, it is recommended to use this tool appropriately, and in combination with other tools.
Any thoughts?

Update: Looks like Aaron Wall and I may have been thinking along some similar lines. He just posted about new domains getting ranked in Google over old sites, and mentioned the following:

“Also think of the search business model as though you are a search engine. To them, being the first person to do something is a sign of quality because to be the first person in a market requires some market timing / knowledge / investment / luck.”

“Catch a wave” theory, explained in financial terminology.

I don’t want to blow the whistle prematurely on a respected website like this one, but Luke brought to my attention yesterday that they ranked pretty highly for the Google search [switch]. Now the page appears in the Google index as “Apple - Switch,” but the page title on the linked page is “Apple - Get a mac.”

Strange.

So, checking the cache date, I notice that the page is 7 days old. Okay, maybe they’ve changed the title. Nope, the title is the same on the cached page as it is on the live page today.

So I fire up the good old Lynx-Borland browser and navigate to the same page. The no-Javacript browser says the title is the same as the live version.

Maybe it’s a backlink thing? My backlink analysis shows that less than 3% of links to the page contain the word “switch.” I wouldn’t think that would be enough recognition to warrant an editorial title change by Google.

Checking DMOZ, Apple has over 900 entries, none of which seem to have this page with the ‘Switch’ title.

So the question is: how does Apple get a different title in their search results from the title their users see on their website, and from their DMOZ entries?