An Insight Into On-Page SEO Factors

[Martin.Avis]

When designing web pages, we try to incorporate as many things that the search engines like as possible. This is called On-Page SEO. But the problem is that nobody knows for sure exactly what factors the search engines actually do like. And they are not saying!

We can 'reverse engineer' the search engines to some extent - my own Keyword LSI Spy tool, and others, do a pretty good job of analyzing the top ranking pages for any given keyword or phrase, but there is only so far you can go because every page is different, and so many factors can come in to play.

What we need is a means of looking at pages where only a small amount of what is on page changes from one page to the next - effectively allowing us to reduce the variables and find out what the SEs really do with a small set of data.

But how do you go about finding a whole lot of pages that you can compare against each other in Google (for example) where we know that only a limited number of things change between them?

It is something I've puzzled over for a long time, but then I had a brainwave.

Forgive me if this is blindingly obvious to you. It is to me NOW, but you, like me, may not have made this connection yet, so the technique I'm about to share may be very useful.

EzineArticles is an online article directory that has many hundreds of thousands (millions probably) of articles. And the great thing from our point of view is that each one is displayed in a standard template. So, effectively, the only things that change on the page from one article to the next are a couple of meta tags and the article body itself.

All the other stuff on the page, particularly for articles aimed at the same keyword, remains virtually identical.

In Google we can search for pages within a website. So by using the search string: "acne treatment" site:ezinearticles.com (using whatever keyword phrase we like between the quotes) we are given a ranked list by Google of articles that they consider to match our keyword from the EzineArticles site.

Because of the sheer size of EzineArticles, you'll be hard pressed to find a niche that doesn't return a good selection. My acne treatment example pulls up 16,500!

Now we can look at the articles that Google has selected for us and see what, if anything, makes the ones nearer the top of the pile stand out.

Here are a few suggestions for factors that you could check out:

* Meta Title tag
* Meta Description tag
* Meta Keyword tag
* Article word count
* Article keyword density
* Use of LSI words in the article
* Use of synonyms in the article
* Keyword density in the article title
* Keyword density in the first paragraph
* Keyword density in the last paragraph
* Keyword density in the middle paragraphs
* Average sentence length
* Average paragraph length
* Links and link anchor text
* Recency of article
* Style of article
* Use of subheadings

EzineArticles have software in place to prevent automated scraping of their content, so it isn't possible (by me anyway) to write a script to do this analysis in a php program, but with some patience you can easily get some good indications manually.

Try it out - and let me know if you find any surprising new insights!

[Jim Burney]

Hi Martin

Thanks for sharing your insights into seo.

Just printed it out to have it handy.

Will get back to you with feedback if I find anything significant.

Jim

[Greg Cooksley]

Hey Martin,

That's absolutely brilliant.....

Now just have to figure out the significance of all those terms:

Meta Title tag
* Meta Description tag
* Meta Keyword tag
* Article word count
* Article keyword density
* Use of LSI words in the article
* Use of synonyms in the article
* Keyword density in the article title
* Keyword density in the first paragraph
* Keyword density in the last paragraph
* Keyword density in the middle paragraphs
* Average sentence length
* Average paragraph length
* Links and link anchor text
* Recency of article
* Style of article
* Use of subheadings

Regards

Greg

[Kurt]

Instead of using Exine articles, I suggest setting up and using Google's site search on your own domain, which is what EA uses for their search.

This gives much more control. And, it's possible EA has set up some preferences that would act as a bias.

It costs $100 a year for up to $5000.

The question is, does Google's site search rank pages in the same way it's main SE does? I'm sure there are similarities....But how much?

I tried to find out how often Google site search spiders and re-indexes pages, but couldn't find it. I "think" it's 24 hours, but not sure. This would allow tweaking and changing pages and checking any changes in the SERPs.

Good idea Martin...I'd just take it one step further if you want to get the most accurate results possible.

Google Custom Search Engine - Site search and more

[ExRat]

Hi Martin,

Won't you also need to check if the articles have off-site anchor links pointed at them, in order to avoid totally incorrect results from your analysis?

I'm noticing in my research that many people have adopted the tactic of pointing anchor links from squidoo/blogger etc and all sorts of other places directly at their ezinearticles articles.

[Martin.Avis]

I agree. It's very difficult to exclude all outside factors. That's why I suggested that this could give an insight rather than a categoric answer.

EzineArticles seemed to offer an immediately accessible source of material to study.

Kurt's idea of putting control copy on a domain of your own would be good to explore further. That way you could be almost certain you'd excluded inbound links as a factor (although it is imposible to be 100% certain). And as Kurt says, you could create a very simplified and consistent template to put a series of articles into.

I might play around with that idea some more.

Martin

[Michael Mayo]

Hi Martin,

Here is something I think you might find interesting from the big G if you haven't seen it before. Google Spam Guidelines

Have a Great Day!
Michael

[dburk]

Hi Martin,

While it's obvious that you have put a lot of thought into what you are doing, there is a major flaw in your experiment.

You can easily do all that your are suggestion on a domain that you control. By using EA you introduce so many uncontrolled variables that there is no way you can get close to having valid results.

All you need to do is publish your own articles on your own site. You need to make the links to those pages easy for a spider to crawl but hidden away from real visitors. You also want to stick with keywords that are not likely to be searched ,to reduce or eliminate the effect of organic backlinks to those pages.

Backlinks and users search behavior are major factors in Google's algorithm. You need to isolate these factors from your experiment if you want to have anything close to valid results.

By the way, many people have conducted similar experiments and posted their findings. The search engine algorithms are constantly changing so it's a good idea to run new experiments to test current assumptions.

[theGypsy]

Yeah, the test is flawed and won't produce meaningful results... one of the more glaring things missing is the links (internal and external) as well as historical ranking factors... Oh and the ranking flux that occurs based on; geographic, behavioral, browser type, personalization and other factors....

And can I just give a quick WTF to the whole LSI bit? Why not phrase based indexing and retrieval? Or HTMM or PLSA? Using the term LSI is nothing more than snake oil salesmanship and if you were not aware; it has NOTHING to do with Google or any other search engines organic algorithms.... More here; Stay off the LSI bandwagon - Search and Internet Marketing rants, news and views - the Fire Horse Trail

Anyway, you have kw density as well as semantic equivalents - which is it? These days I would look more at 'phrase density' as search engines are into categorization and theme data.

Unfortunately, while I applaud the curiosity, the testing you propose would not give anything but dirty signals and the potential for false assumptions...

[websitetraffic]

That's really encouraging, besides, it is a fact that SEO is very vital in terms of making your website popular and be on the top rank on search engines.