64 replies to this topic

[Michael Martinez]

Michael, you missed several key points of the test.

No, Shane, I did not. I understand how these sorts of test fail because I have looked at a lot of them through the years.

Just because you mis-spelled "resume writing" doesn't mean that Google and Yahoo! wouldn't combine the mis-spelled words with the proper spelling. They do that all the time. You didn't take that into consideration in your test.

You don't determine what the rules of relevance are.

Worse yet, Google and Yahoo! haven't revealed what the rules of relevance are.

So just exactly how do you calibrate a test that is supposed to ignore those rules and focus on the importance of linkage?

Your methodology doesn't even open the book on the problem, much less address it.

Your test and your conclusions are therefore unfortunately invalid. You're not the first person (or people) to make that mistake. You won't be the last.

[BathGems]

But I am not interested in ranking well for HOME REPAIR

I would not spend money buying that key phrase even if it were $.10 per click. It is just way too unfocused for commercial purposes. Of course if it were free through the organic listing I would take what I could get.

I only purchase targeted words like contemporary vanity or waterfall faucet or curved shower rod because that is what I sell.

Its hard to say how many sites are optimized for home repair but I am fairly sure how many are optimized for curved shower rod.

This should make a difference in this discussion. I think theory 3 only holds up for the general category type search rather than the product specific search.

Steve

[Shane]

Just because you mis-spelled "resume writing" doesn't mean that Google and Yahoo! wouldn't combine the mis-spelled words with the proper spelling.  They do that all the time.  You didn't take that into consideration in your test.

I did consider that. However, if they "fixed" the spelling for me behind the scenes, then these two things would have been true:

• They would have returned way more than 800 pages, since the search would have returned all resume writng pages and all resume writing pages.
• CareerBuilder's page would have shown up somewhere in the results for resume writng before the test since it did show up in the results for resume writing.

Neither of those were true, though.

[Jill]

Now, let's take look at Submit Express. Are they optimized? (i DON'T think so)...
However, they have overpowered "the system" by getting a trillion links.

That's all part of optimizing. So yes, they are indeed optimized.

But I do like Shane's point of how 1 really good link can make a difference. That's what all the link spammers and farmers don't seem to grasp.

[Michael Martinez]

I did consider that.  However, if they "fixed" the spelling for me behind the scenes, then these two things would have been true:

• They would have returned way more than 800 pages, since the search would have returned all resume writng pages and all resume writing pages.
  
• CareerBuilder's page would have shown up somewhere in the results for resume writng before the test since it did show up in the results for resume writing.

Neither of those were true, though.

They don't have to fix anything for you behind the scenes. All they have to do is associate one word with another. And, no, they don't have to return any specific number of pages. They can return a random number of pages every time.

I can get over 600 results for "resume writing" from CareerBuilder's Web site today.

I can't tell you when those pages first appeared on CareerBuilder, but at least one bears a 1999 copyright date, for what it's worth.

Now, if you point a link at CareerBuilder with the anchor text "resume writng", and if Google knows enough to suggest "resume writing" in your search results for "resume writng", how do you show that your test validates 1 link for "resume writng" as making a significant difference?

Today, I get 475 results for "careerbuilder.com resume writng".

And look what Google highlights for the one of the top ten listings (I don't do screen captures, so the results may wiggle because of the ongoing update process):

resume writung: new resume writung here, about resume writungCareerBuilder.com Jobs - The Largest Job Search, Employment.. ... resume ritung - resume witung - resume wrtung - resume wriung - resume writng - resume ...
resume-writung.resume-writing.new.aabto.com/ - 8k - Cached - Similar pages

Hey, "writing" appears in a URL, and Google pays some attention to URLs, although most of us (including me) feel that it's not a very significant factor in their relevance algorithm.

A little further down, another listing has both "writing" and "writng":

Here you will find more items under this Employment Category81 free resume and Job search workshops, including Cover Letter Tips, ... Cover Letters Writng Services. Cover Letter Writing plus resignation, reference, ...
www.canada.christfamilies.com/display1. asp?Category=Employment&item=Thank%20You%20Letters - 150k - Supplemental Result - Cached - Similar pages

To be honest, vetting a test you performed over a year ago is impossible. But my point is that you cannot simply pluck an expression out of thin air, create a single link, and then assume that it's the link which causes the change in rankings. Google certainly knows that "writng" is a mis-spelling for "writing". And (today) they clearly have data on at least one page that uses both "writng" and "writing" in its indexable text.

There is no way to know how much (if at all) Google's under-the-hood stuff affects the search results when one more piece of data is associated with a Web site that already has a LOT of data associated with it.

You're adding a feather to the top of an already huge pile of feathers. All feathers carry some weight, even though it's largely negligible for each feather. Add them all together, though, and you'll eventually cross a threshold where your collection of feathers tips a scale one way or another.

How do you show that the scale didn't tip because of the collective weight of all the feathers in your test?

You made no provision for that.

Hence your test is invalid.


I assure you, Shane, this has been tried many times through the years. Everyone makes the same mistake. You have to look at ALL the data, not just what is visible in the results of whatever queries you run.

[gbot]

That's all part of optimizing. So yes, they are indeed optimized.

Yes... but as far as "on-page optimization", they don't even compare to you or Bruce.

Michael- I understand, and agree with your point about content. For non/semi competitive phrases, you can get ranked high quite easily with very few, if any links.

Luckily SEO isn't 1, 2 or 3 things. It's over 106 (for Google), so it's a collective combination of many things. With that being said, I do believe you can overpower "the system" like Submit Express. There page isn't bad, but, it doesn't compare to Jill's front page or Bruce's web_rank page.

Steve- Your keyword phrases are somewhat competitive. Following a search for allintitle: curved shower rod, Google said it had 481 Web sites that had "curved shower head" in the title. However, allintitle: curved shower rods (rod with an "s") said that there were 1,180 people with the search phrase "curved shower rods" in the title. These are more than likely your competitors. While 1,180 isn't the highest number I've ever optimized for, it also isn't the lowest. Therefore, you will need to incorporate as many SEO components as possible in order to get ranked high. One component (content, links...etc) just won't cut it.

As far as having 50 sites "tied" for first, I don't thing that's possible. With at least 106 variables, you couldn't have two sites the same.

My two lil pennies,

gBot

[BathGems]

While 1,180 isn't the highest number I've ever optimized for, it also isn't the lowest. Therefore, you will need to incorporate as many SEO components as possible in order to get ranked high. One component (content, links...etc) just won't cut it.

As far as having 50 sites "tied" for first, I don't thing that's possible. With at least 106 variables, you couldn't have two sites the same.

Lets differentiate between two definitions of "tied". One is numerical score that Google assigns to these 1,180 websites. Lets assume that 900 of them are actually pages that sell the curved crescent shower curtain rod and the rest are other things like SEO attempts at creating inbound links to various pages. Lets further assume that Google's ranking is a number between 1 and 2, 2 being the most relevent. Further Google may use one of two methods to rank the relevence of this universe of pages between 1 and 2. One is to make relevence relative and evenly distribute these 1,180 pages between 1 and 2. The other way is to make relevence absolute where the very most relevent pages might be ranked 1.99999876, 1.99999874, 1.99999866, etc.

Either way, the top 200 are very likely very well formatted sell copy containing the obligatory curved shower rod in the copy, title, description, and meta tags(yeah yeah yeah we all know, it does not matter). Anyways, these 200 pages are likely all in very nice e-commerce website where you can make the purchase right away.

They all probably have links in from paid submissions in directories.

So while there is a algo difference because by the very nature of ranking there is a continuum, there is no real difference in relevence to the consumer who typed in curved shower rod, looking to purchase one in the next 20 minutes. Besides there being no real difference in whether the consumer is finding what they are looking for, there is also hardly a difference between the usability of these sites so that the Google measured difference is little more than arbitrary statistical differentiation. Arbitrary because the things measured have little to do with any human perceptible differences.

So thats what I want to say about that.

Steven

[Michael Martinez]

Michael- I understand, and agree with your point about content. For non/semi competitive phrases, you can get ranked high quite easily with very few, if any links.

And for highly competitive phrases, too.

[Jill]

Yes... but as far as "on-page optimization", they don't even compare to you or Bruce.

But on-page optimization is just one factor in rankings. You have to always take both on and off page factors into consideration. Especially for highly competitive phrases, neither one alone is going to cut it.

[Michael Martinez]

...Lets assume that 900 of them are actually pages that sell the curved crescent shower curtain rod and the rest are other things like SEO attempts at creating inbound links to various pages....

Lets further assume that Google's ranking is a number between 1 and 2, 2 being the most relevent....

Further Google may use one of two methods to rank the relevence of this universe of pages between 1 and 2....

Either way, the top 200 are very likely very well formatted sell copy containing the obligatory curved shower rod in the copy, title, description, and meta tags(yeah yeah yeah we all know, it does not matter)....

Anyways, these 200 pages are likely all in very nice e-commerce website where you can make the purchase right away.

They all probably have links in from paid submissions in directories. 

...there is no real difference in relevence to the consumer who typed in curved shower rod, looking to purchase one in the next 20 minutes.

...there is also hardly a difference between the usability of these sites so that the Google measured difference is little more than arbitrary statistical differentiation.

Steven,

Assumptions are not the best fuel upon which to run a marketing engine.

When dealing with search engines we have to make some guesses.

But the difference between a guess and an assumption is that we are more easily divorced from the guess than from the assumption. The consequences of testing a guess and relying upon an assumption really differ only in that we make corrections more rapidly as our guesses prove to be incorrect. Quite often, when people are relying upon errant assumptions, they find themselves in forums declaring what their assumptions are only to have someone else challenge those assumptions.

In practice, I've seen people run from forum to forum, looking for validation of their assumptions, always leaving when people point out they made the wrong assumptions.

In this field, hypothesizing is unavoidable. Assuming is self-defeating.

[BathGems]

I was actually drawing on my economic backround by making simplifying assumptions to analyze a theory.

You assumed I was was using the term assumption in a different way. I was only married to these assumptions so far as they allowed this discussion to move forawrd without getting tangled in the side issue of what it means to be "tied".

Steven

[jehochman]

Anyways, these 200 pages are likely all in very nice e-commerce website where you can make the purchase right away.

Exactly.

In such a situation, a bunch of sites at parity, you have to invest in advertising or SEO to get to the top and stay there. Whoever has the best business model, providing the most value to their customers and earning the most money, will be able to afford the top ranking.

As a business owner you need to consider:
+ When a prospect calls, what's your close rate?
+ What's a customer worth to your business?
+ Do you generate add-on sales and follow-on sales?

Online business eventually depends on these age-old methods. SEO is a tactic that you can use if you have the resources to invest. If you aren't creating enough value, SEO won't solve your problem.

[Michael Martinez]

I was actually drawing on my economic backround by making simplifying assumptions to analyze a theory.

You assumed I was was using the term assumption in a different way.  I was only married to these assumptions so far as they allowed this discussion to move forawrd without getting tangled in the side issue of what it means to be "tied".

Steven

Touche....

[gbot]

But on-page optimization is just one factor in rankings. You have to always take both on and off page factors into consideration. Especially for highly competitive phrases, neither one alone is going to cut it.

I agree. But do you think it's possible to "overpower" the system? Like with Submit Express, there on-page optimization isn't bad. But I don't think it compares to yours. So... You have more of a "complete" package, while Submit Express has a couple of good SEO components, with one big "overpowering" component. Put it this way; if you had the same amount of links as Submit Express, would you be ranked 1st or 2nd. In your "honest" opinion you would be ranked 1st, in your humble opinion you wouldn't really know. Am I right?

[Jill]

But do you think it's possible to "overpower" the system?

Certainly you can, but that generally comes crashing down on you eventually.

I wouldn't say that S.Exp. is doing that though. They seem to certainly be about search engine optimization, even if I don't necessarily agree with submitting to 75,000 sites and probably other things that they might do!

I think that all comes back to the stuff about G trying to show a variety of types of sites for any given query.

I've long since given up trying to figure out what makes one page number 1 and another number 2. I personally can rarely find any rhyme or reason for it.