54 replies to this topic

[Raphael]

It's just that tables weren't designed for layout, and nesting tables can cause serious cross-browser compatibility issues. But as I said, none of that matters to search engines.

Heading slightly off topic here, and I *am* (now, at least) a big proponent of the CSS way of diong things - But a nested table based layout, written correctly is significantly easier to make cross-browser than a CSS-based layout because you don't suffer from the IE box model issues.

In a nested table layout (or any table layout), you'll often use spacer gifs to aid your layout, and IE can't render a 10 pixel image at anything other than 10 pixels, because that's the size of the image.

Browser compatibility is not why I stopped using nested tables. In fact if that was my only issue, I'd still be using them. However, I have to worry about code maintenance, and since I don't write HTML anymore, I write code that writes HTML (Or I write code that writes code that writes HTML), keeping my HTML itself as simple as possible and putting all my design elements into a separate file is more important to me.

[qwerty]

Spacer images...

So much code just to create empty space. Ah well. Chacun a son gout

[DanThies]

As I said sometimes things on here conflict with things that they taught and whilst I will always make up my own mind, I am of the school that I need to know 'why'.  Hence my question to you.

If you'd like to know why I don't agree with Jill & Scottie, aside from actually testing this stuff which is easy enough, you may want to read my article on how search engines work. Pages 5 & 6 will be of particular interest, as is Dr. Garcia's discussion of the vector space model, if you don't mind getting tired head from looking at equations.

What Jill & Scottie are suggesting is that search engines would analyze the following two snippets of text and find them exactly equal:

Home About Us Products Services Privacy Subscribe To Our Newsletter! Name Email Copyright 2003-2005 MegaGlopTronInc, Inc. Search term here bla bla bla...

Search term here bla bla bla... Home About Us Products Services Privacy Subscribe To Our Newsletter! Name Email Copyright 2003-2005 MegaGlopTronInc, Inc.

I just don't believe that's true. In fact, it is demonstrably not true.

[Jill]

Please show us your evidence, Dan, and then I might believe you.

I've looked at and optimized enough sites to see it NOT make a difference, which is why I will personally stick with that conclusion until I see it actually make a difference.

[DanThies]

Happy to oblige, Jill, test pages are online here and here.

Watch this SERP for the results.

[Jill]

Awesome, thanks Dan. I will definitely watch that spot and also make sure that the way you're testing it makes sense in order to draw the appropriate cause and effect conclusion.

[Jill]

One comment on this after reading both your pages. My contention is more that assumes your keyword phrases for the page are used throughout the copy. Not at the top, not at the bottom, but interspersed where it makes sense.

My beef has more to do with people who think they have to put the body of their page before their navigation.

Will your test do anything to prove this either way? Your test simply looks like it shows if a word is sooner in the document than one that is later, it might rank before the other one.

I'm not sure I've ever really disagreed with that one in theory (although one could probably find posts of mine that may seem like they do). But you don't seem to be testing what I have been arguing about with people.

[SmellieNellie]

Your test simply looks like it shows if a word is sooner in the document than one that is later, it might rank before the other one.

This is what I meant when I said that I had been taught that tables can "push the content down the page". I wasn't even thinking about navigation.

Unfortunately, I have not tested it because all the sites that I get involved in developing from scratch do not use tables. Those that I have inherited I don't want to mess with and I simply have not had time to create test pages myself though I do see that this is something I should possibly invest time in doing not just for this issue but for any issue that I want to test results on.

So Dan, I shall be watching your test with interest, though common sense would point to the fact that the sooner a search engine picks up a keyword phrase, the better.

Does common sense enter into SEO?

[Sleeve]

I have built a rather large CSS website with meaty content that I'd like to think is well written. Recently I moved a paragraph closer to the top of the page because I felt that the page made more sense for my visitors. (I like to think of paragraphs on a page as "word pictures". The overall meaning changes depending on what you 'show' them first.)

Interestingly, the page moved up in rank considerably for a keyterm used in that paragraph. Since then I've done it with other pages and its resulted in a similar effect. Maybe it was a coincidence but in this case, it would seem text placement has some validity in keyterm ranking.

[DanThies]

One comment on this after reading both your pages.  My contention is more that assumes your keyword phrases for the page are used throughout the copy.  Not at the top, not at the bottom, but interspersed where it makes sense.

If you want to send me some copy to play with, that you think represents a reasonable test, please do.

My beef has more to do with people who think they have to put the body of their page before their navigation.

In terms of code organization, or in terms of visual presentation?

Your test simply looks like it shows if a word is sooner in the document than one that is later, it might rank before the other one.

Not might better, "will" rank better. This is what happens when designing with tables forces a designer to put navigation and other non-content ahead of the [url=http://searchengineland.com/070531-115312.php]Real[i][/i] Content[/url] of a web page.

But you don't seem to be testing what I have been arguing about with people.

I'd be happy to construct another test set, if you can explain what you want to test. Maybe we can just take 100 words (and a bunch of links) of navigational jibba-jabba, then a few paragraphs of text, and compare that vs. a page with navigational jibba-jabba at the bottom.

You want to take this offline and bring it back to the forum when we have a plan for a good test?

I'm not trying to suggest that everyone go out and redesign their websites, because there are better ways to spend resources, but unless it's actually more economical to use tables, why not take control of the order of the content? I don't claim to be a CSS layout expert, because I'm not, but I'm also not claiming to be a web designer.

[Scottie]

Cool test! Looking forward to the results.

[Jill]

It is my belief that search engines understand that a good portion of websites are designed using tables, with navigational tables generally coming before the content.

I also believe that they take that into consideration, and therefore "table tricks" are not helpful.

I seem to remember Qwerty testing this once and coming to the same conclusion (although I might have dreamt that). I believe he said he used the table trick and did not notice any difference in rankings.

I feel you would have to do something like that as a test, use a real site, with real tables that has been established, etc. and then do the table trick and see what happens. Then of course duplicate it across a number of sites so as to make sure you're not seeing the "cabbage on the monitor" factor.

I really believe that the engines understand what navigation is, how it's on every page, and blah blah, and treat it accordingly whether in the first table or the last.

[Ron Carnell]

A real test, I think, would take some time and effort to conduct.

I don't think it's enough to put up two pages to test navigation placement, for example, because it is the consistency of navigation that helps a search engine differentiate between links that are content and links that are navigation. Give me six web pages using the same template and I can write a program in thirty minutes that will separate [url=http://searchengineland.com/070531-115312.php]Real[i][/i] Content[/url] from systemic elements like header, footer, and navigation. Give me only two pages, using two different templates, however, and my program would be lost.

I believe the placement of words within the content is important in determining relevancy for the same reason your ninth grade English teacher taught you to write a topic sentence for each of your papers. Good writers are going to do that any way. I also believe, however, that today's search engines are VERY good at paring content from supporting structure. Just like your teacher, the engines generally know where to look for that topic sentence.

We all know you can put 10K of JavaScript at the top of a document and still rank well. The engines will ignore it. Similarly, I believe they will ignore a dozen table tags and the 2K of spacer images used to manage those tags when all that supporting code precedes the content. They have something like eight years worth of documents in their database that have been following pretty much that exact pattern, and programmers are darn good with patterns. I've seen evidence that points both ways, but I'm nonetheless convinced that standardized navigation, much like script and styles and table tags, will be parsed and considered quite apart from [url=http://searchengineland.com/070531-115312.php]Real[i][/i] Content[/url]. Today's algorithms, heavily influenced by link structure, almost demand it. Such being the case, it won't matter if your navigation precedes or follows the content, as long as it remains apart from the content.

It's going to take a lot more empirical evidence than simply two web pages to prove anything useful, though. Heck, we already know one has to rank higher than the other even if both are identically relevant, so the first test is just like flipping a coin to see if it comes up heads or tails. One flip, however, isn't going to tell you if the coin is weighted to always land the same.

<edit>I was still typing, it would seem, while Jill said much the same. </edit>

[DanThies]

Ron,

This isn't about code - javascript, css, html tags, etc. That's all got to be stripped out by the parser at some point, before a page can be indexed.

It's certainly possible that a search engine could attempt to analyze the page, and index something other than what's been presented in the original source code.

I won't try to guess whether search engines might re-order the content that's presented, but I am willing to make the bold assertion that if any search engine does this, whatever they are doing will not be 100% accurate.

We have done far more rigorous testing than this, but to go through all that again, just for this... I'm too busy.

[qwerty]

I seem to remember Qwerty testing this once and coming to the same conclusion (although I might have dreamt that). I believe he said he used the table trick and did not notice any difference in rankings.

Well, not exactly. It was absolutely-positioned layers that put the content at the top of the <body> and elements like the banner and navigation below it. But it wasn't exactly a test, as I didn't have an alternate version to compare it with. I just didn't see any evidence that it helped.

I did switch the site after a couple of years to a more fluid layout, and in doing so the content got moved down in the code, and I didn't see any changes after that happened either.