tedster - 7:45 am on Jul 31, 2009 (gmt 0)

I read Google Patents often, not because they use everything they put into a patent. They don't - and especially not at first. But the patents offer clues about what MIGHT be coming down the road. And a couple years later, sometimes they give us nice clues about ranking puzzles that have begun to surface.

With the current update apparently doing "something different" with backlinks, I went back for another reading of the 2007 patent application Document Scoring Based On Link-Based Criteria [appft1.uspto.gov] and I found a bunch of interesting points. It's not necessarily the clue to this update, but more along the lines of explaining some other observations.

If you read the entire patent, you don't end up with a clear-cut list. Instead you get "sometimes this way, sometimes that way, and it all depends". Most interesting to me was the way some back link factors can work differently (and even work the exact opposite way), depending on the query terms.

The patent mentions some of the standard factors we already talk about - trust and authority of linking sites, spikes in back link growth, spikes of similar anchor text, and so on. But a closer reading brought me some other goodies. Here is my paraphrase for some of the paragraphs I found interesting:

PAGE SEGMENTATION and RATES OF CHANGE
[0051] Here Google defines a factor called UA [update amount], and it can be a factor that they weight differently for different segments of the page. Not only the back link juice itself is weighted differently, whether it changes is also given a different weight, depending on where the link appears on the page.

PAGE CHANGES CAN IMPROVE OR LOWER RANKINGS
...it all depends on the query terms!
[0052] Pages that show an increasing rate of change might be scored higher than pages for which there is a steady rate of change.

Now contrast that paragraph with this:

[0055] For some queries, content that has not recently changed may be a better result. So ranking factors can work one way for one search term, and the opposite way for another.

PARTIAL INDEXING OF PAGES
[0053] This paragraph deserves some exact quotes:

In some situations, data storage resources may be insufficient...search engine may store "signatures" of documents instead of the (entire) documents themselves to detect changes to document content. In this case, search engine may store a term vector for a document (or page) and monitor it for relatively large changes. According to another implementation, search engine may store and monitor a relatively small portion of the document.

And so we hear "why can't I find my page for an exact phrase search." And we also have a hint that sometimes Google may not have enough storage all the time.

RANKING FOR SEVERAL SEARCHES
[0063] How often a page appears for different searches can help boost rankings across the board. So maybe optimizing a single page for several different terms makes some kind of sense, eh?

RANKING CEILINGS, TRAFFIC THROTTLING and the YO-YO EFFECT
These two paragraphs deserve to get bumped together:

[0075] A spike in BACKLINKS can mean two things - a suddenly hot topic, or an attempt to spam.
[0102] A spike in RANKING can also mean two things - a hot topic or spam.

Now here's where it gets interesting: According to [0102], Google may allow a ranking to grow only at a certain rate, or apply a certain maximum threshold of growth for defined period of time. This might well account for the pain of "I've hit the ceiling" that we sometimes feel.

Even beyond those painful ranking ceilings, I've seen analytics that show amazing Traffic Throttling [webmasterworld.com]. The daily traffic graph looks like a barber comes in at 2pm every day and gives a buzz cut. And in order to throttle traffic that effectively, the only way I can see is Yo-Yo Rankings [webmasterworld.com].

This patent suggests that if a site experiences an extreme throttling of its traffic, (or a yo-yo between page 1, page 5, page 1, etc) then the site probably had some suspiciously spiky growth in back links -- spikes that couldn't be explained by a Hot Topic suddenly popping up for the general public. And so, Google put the site on their traffic regulator.

That lines up exactly with the cases I've worked with. And members here first noticed the yo-yo (traffic throttling) in 2008 - more than a year AFTER this patent was filed.