IR colleagues and marketers are now reading the IRW 2007-06 issue wherein I elaborate on The User-Machine Relevance Perception Gap. So far, I have received some feedback from these. As part of the process, Ben Pfeiffer from Ranksmart asked me a valid question:
“Hi Dr. Garcia,
I just finished reading your relevance gap report. Very good, thank you for writing this.
Can you clarify on the comment about SEO myths in regards to “repositioning of text with CSS improves relevance – not necessarily”
In what case would CSS repositioning improve relevance?
Sure, Ben. The IRW issue discusses cases wherein CSS not necessarily improves relevance, but does the opposite. However, there are cases wherein CSS can be used to improve relevance, at least as perceived by users.
In Chapter 8, page 188 of his book “Speed Up Your Site, Web Site Optimization” under the section that reads “Raising Relevance”, Andrew King explains how this can be done.
He uses a happy medium for designers to follow: try to place important content first in the code and also to appear prominent in the layout as perceived by users.
In his example he uses a three-column layout and CSS. In Chapter 4, page 82, he describes a similar trick, but entirely with tables.
In reference to what was discussed in IRW 2007-06 I can restate Andy’s approach as follows:
Place the most important content (or the one intended to convey an important message) as earlier as possible in the pseudo document (text stream) and then apply CSS repositioning to make it appear prominent in the layout. However, don’t go overboard with this, as not always an important message has to appear earlier in a text stream. One just needs to ponder the odds on a per case basis.
Overall, I suggest to avoid what many web designers and SEOs seem to do. They use CSS to make content appear relevant in the layout and ignore how it appears in the source code. Others do the former only to later struggle with the source code. The odds are that this is quite backward.
As a cardinal rule, don’t go overboard by nesting HTML elements just to make a layout “beautiful”. You would be surprised, for example, to see how content placed in beautified over-nested divs look like in the pseudo document after linearization and tokenization.
When it comes to optimization and CSS positioning, its about time for SEOs/designers to think in terms of a pseudo document (linearized text stream) to be optimized. It might be a good idea to have the notion of a linearized text stream “burned in the mind” during both web design and optimization. My personal experience is that this strategy minimizes parsing and indexing issues that one might find at a later time during a Relevance GAP Analysis.
Now the following is for SEOs or designers that want to monetize the above:
The steps involved during document indexing can be used to conduct a GAP Analysis, to check parsing/indexing issues in a web page, and to examine the apparent vs current state of the distribution of words.
I always try to do such Gap Analysis before and after any SEO work so I can spot indexing issues. When doing so, I conduct several tests with the target search engine so I can have a feeling about its parsing behavior. SEOS can do this all this as well and charge for it; that said, provided they know what they are doing.
I have written a parsing tool that does this, precisely because I did not want to use third-party parsers wherein who knows how good or efficient the parsing rules programmed are. For those that are IR- and programming-challenged, I suggest to try an open source parser or text miner and hope for the best. Currently there are plenty of text mining tools available for SEOs to use. As an added benefit, some of these can be used to improve both copy style and indexing.