64 replies to this topic

[Michael Martinez]

The question wasn't about which sites will rank better for secondary phrases, it was how SE's determine rankings between 75 to 100 pages that are all equally well-optimized for one specific phrase. ...

Which can still be a secondary phrase for many other pages.

And, no, it does NOT come down to off-page factors.

It only comes down to off-page factors when everyone forces the issue. There are relatively few search expressions where that is the case.

[BathGems]

Why would you need curtain magnets when you have this wonderful curved shower rod that bows away from you, creating an additional 25% space? The curtain is so far away from you that the upward air flow physics fail to pull the curtain close enough towards the user to be a nuisance.

Why do you say links are not the issue? What is your reasoning?

[mcanerin]

Maybe I'm going against the crowd (or moving things off-topic) here, but frankly, threads about links are starting to bore me...

What I think is much more interesting is the question by the original poster of what a search engine is likely to do about this issue?

There are several options. First, they can stop worrying about it, since no matter what they show, as long as the top 50 are all good results, then they don't have to spend the resources worrying about sorting it much.

After all, at a certain point, who cares (aside from the owner of the website in question)? If I get a great list of sites in the top 10, do I really care if the ones in the top 20 are good too? And do I really care what the order they are listed in at that point is? An SEO might care, a website owner would care, but the users would not.

We see this effect with DMOZ, where the editors care a lot less about adding a website to a sub-directory with 100 other good sites on the exact same subject than they are with adding sites to areas with only a couple of good results.

Once you achieve this, then as a software engineer you have to make a decision. "Do I continue to refine this result, spending more and more resources on detecting smaller and smaller differences that the end user doesn't even care about, or do I spend those resources in areas that the user does care about"?

On the other hand, I know a lot of engineers. They often don't have a "it's good enough" mindset, but rather a "it can be better" mindset, particularly if they are young and ambitious, like Googles engineers. Especially if they have lots of powerful tools to work with.

So what would happen if a search engine decided that there really was a difference in there someplace, and that difference mattered? You would end up seeing either 1) smaller and smaller differences having a larger and larger influence on the SERPS, or a movement towards completely different or new measurement tools that offered the ability to measure things that were not, before.

The reason I wonder about these things is because this creates 2 possible scenarios in an highly competitive SERP (which I tend to be in, lately).

1) "Once the top 30+ are all passing the quality checks, we don't really care what order they are in". The result in this case is likely to be sorted out via the proverbial "butterfly effect", named after the effect where a single flap of a butterflies wings could, in a complicated self referencing system like a weather system (or search engine), cause a hurricane to occur in another part of the world.

This could result in what I'll call "Butterfly SEO", where, once you get to a certain level of optimization, the things that affect your rankings are things that are less and less obvious, and more and more technical. Technicians (and spammers) love this. I know for a fact that in certain SERPs you can see this effect, where something that traditionally isn't an problem, suddenly makes or breaks your rankings.

2) "Since we are having a hard time figuring out which sites are better than others at a certain point, we need to start measuring criteria other than the traditional ones" This is interesting, because instead of needing to get pickier and pickier about links, content, etc, the search engine begins to look at areas that are not normally looked at, and thus more likely to show meaningful and measurable differences in the sites listed. This implies that a more holistic, less regimented approach to SEO would work better than just pushing the same traditional buttons harder and harder, over and over again. Or at least a change in SEO tactics that also addresses the new criteria. I see indications of this type thinking in the aging delay and other similar issues.

These are 2 totally different approaches to SEO for highly competitive results. I have some ideas on the actual plan, but I'd be interested in hearing what other people think.

My opinion, as usual.

Ian

Edited by mcanerin, 24 January 2006 - 10:44 PM.

[Michael Martinez]

Why do you say links are not the issue?  What is your reasoning?

Well, hopefully I won't bore Ian out of the discussion, as I like what he wrote. But my reasoning is quite simple: I know better than to believe all the nonsense that gets posted about links and linkage in various SEO forums and tutorials.

The people who write about links today were not around when I got involved in SEO in 1998. By 1999, I understood pretty well where links fit into the big picture. I didn't understand how staying on the good side of influential people mattered a great deal.

I know better, now.

In the meantime, I've spent years playing both sides of the fence: links here, content there. Content wins every time because, frankly, I don't have the resources to play with the links. I do have to go up against a lot of sites that have several times more links than I do.

I can live with being in the top five for targeted search expressions. I'll accept being in the top twenty for experimental expressions.

Ian is right when he says that you can reach a point in your optimization where the obvious stuff is balancing out. Links are obvious stuff. When everyone has 20,000 inbound links, another 1,000 links aren't going to help you. Then you need content.

Content, as some people have noticed, can be both on-page and off-page. I prefer working with on-page content, but there are people who work with off-page content.

Off-page content includes (but is not limited to) link anchor text (I've used as few as ten links -- out of thousands -- to adjust rankings in highly competitive expressions simply by changing their anchor text), directory listing titles and descriptions, RSS/XML listing titles and descriptions, and text surrounding links. I've also experimented with text in which I've embedded non-linking URLs with inconclusive results.

You won't find any linkmeisters propounding their theories today doing anything that I haven't done a lot of in the past. When they catch up to where the curve is today, the curve will have moved on.

And that is strictly by their own choice.

When you compete against people who are every bit as knowledgeable as you about search engine optimization, who have at least equal resources to bring to bear, if not more, you will fail every time if you depend on links.

It is that plain and simple. You cannot rely on links when you go up against the heavy hitters. I'm not talking about spammers. I'm talking about the huge honkin Web sites with 100,000+ pages of content, where the title tags are unique, where they have every imaginable type of content in play, where their inbound linkage comes from across the board, where they get major media exposure.

Amateurs think of links in situations like that, and I assure you none of the linkmeisters are successfully competing in these arenas -- unless it's strictly by fortuitous accident, and they won't admit to that.

You have to tweak anything and everything at your fingertips. What works today won't work tomorrow because there is someone on the other side of that SERP looking at the same stuff you are and tweaking everything at his fingertips. Sometimes, there are ten other guys doing it, and they're all experimenting as furiously as you, and they're all reading the SEO forums like you to see the "weather reports" (the screaming that accompanies major updates, the gentle upsurge in queries that accompanies minor updates).

The real masters of SEO rarely speak up, and when they do they say things so profound most people barely react. You get wrapped up in the "Is to/is not!" arguments over links-versus-content and you don't learn from the sweet gleams of inspiration that squeak through the clutter of noise.

I don't read Matt Cutts because I believe everything he says is exactly like things really are. I read his blog because the most valuable information he discloses is the incidental stuff that people rarely respond to. They're watching data centers update. They're looking for vindication of the Unified Link Theory. They're exchanging meaningless pitter patter and flames with each other and anyone they think is vulnerable to their bullying.

Don't do that. Don't follow the crowd. Be the leader. Look ahead. Take some risks. You'll be wrong more often than you're right. But learning from your own mistakes will teach you what works faster than learning from other people's mistakes, just as learning from other people's mistakes will teach you faster than your own.

Like links and content, they go hand in hand.

If you got no content, you got nothing to link to.

[BathGems]

The reason I wonder about these things is because this creates 2 possible scenarios in an highly competitive SERP (which I tend to be in, lately).

1) "Once the top 30+ are all passing the quality checks, we don't really care what order they are in". The result in this case is likely to be sorted out via the proverbial "butterfly effect", named after the effect where a single flap of a butterflies wings could, in a complicated self referencing system like a weather system (or search engine), cause a hurricane to occur in another part of the world.

This could result in what I'll call "Butterfly SEO", where, once you get to a certain level of optimization, the things that affect your rankings are things that are less and less obvious, and more and more technical. Technicians (and spammers) love this. I know for a fact that in certain SERPs you can see this effect, where something that traditionally isn't an problem, suddenly makes or breaks your rankings.

2) "Since we are having a hard time figuring out which sites are better than others at a certain point, we need to start measuring criteria other than the traditional ones" This is interesting, because instead of needing to get pickier and pickier about links, content, etc, the search engine begins to look at areas that are not normally looked at, and thus more likely to show meaningful and measurable differences in the sites listed. This implies that a more holistic, less regimented approach to SEO would work better than just pushing the same traditional buttons harder and harder, over and over again. Or at least a change in SEO tactics that also addresses the new criteria. I see indications of this type thinking in the aging delay and other similar issues.

These are 2 totally different approaches to SEO for highly competitive results. I have some ideas on the actual plan, but I'd be interested in hearing what other people think.

Well the first theory simply creates an upper class that anyone can enter through SEO meritocracy. Get yourself into the top tier, be it 10 or 50, and then tweak away.

The second theory might mean that the SE sees that the top 50 are equally relevent, and lets the market decide. Whichever gets a higher clickthrough ratio is the most relevent according to the relevent market. In this case the rich get richer.

[Jill]

Wow. 2 great posts.

So what would happen if a search engine decided that there really was a difference in there someplace, and that difference mattered? You would end up seeing either 1) smaller and smaller differences having a larger and larger influence on the SERPS, or a movement towards completely different or new measurement tools that offered the ability to measure things that were not, before.

Don't do that. Don't follow the crowd. Be the leader. Look ahead. Take some risks. You'll be wrong more often than you're right. But learning from your own mistakes will teach you what works faster than learning from other people's mistakes, just as learning from other people's mistakes will teach you faster than your own.

Thank you Ian and Michael.

[laertes]

The reason I wonder about these things is because this creates 2 possible scenarios in an highly competitive SERP (which I tend to be in, lately).

1) "Once the top 30+ are all passing the quality checks, we don't really care what order they are in". The result in this case is likely to be sorted out via the proverbial "butterfly effect", named after the effect where a single flap of a butterflies wings could, in a complicated self referencing system like a weather system (or search engine), cause a hurricane to occur in another part of the world.

This could result in what I'll call "Butterfly SEO", where, once you get to a certain level of optimization, the things that affect your rankings are things that are less and less obvious, and more and more technical. Technicians (and spammers) love this. I know for a fact that in certain SERPs you can see this effect, where something that traditionally isn't an problem, suddenly makes or breaks your rankings.

2) "Since we are having a hard time figuring out which sites are better than others at a certain point, we need to start measuring criteria other than the traditional ones" This is interesting, because instead of needing to get pickier and pickier about links, content, etc, the search engine begins to look at areas that are not normally looked at, and thus more likely to show meaningful and measurable differences in the sites listed. This implies that a more holistic, less regimented approach to SEO would work better than just pushing the same traditional buttons harder and harder, over and over again. Or at least a change in SEO tactics that also addresses the new criteria. I see indications of this type thinking in the aging delay and other similar issues.

These are 2 totally different approaches to SEO for highly competitive results. I have some ideas on the actual plan, but I'd be interested in hearing what other people think.

I'd like to suggest a third theory, to muddy the waters even further, i.e.

3) "Since we have so many totally relevant sites for this keyword, lets choose a few of each type. We can show the searcher an e-commerce site or two, an academic .edu site, a large contnet site, a hobbyist site, a high TrustRank site, etc. And we'll vary the %age of the mix now and then to keep SEO types guessing."

Is there anything that supports this theory? I don't know. Just thinking out loud. I didi a search for "home repair":

1. doityourself.com/
2. homerepair.about.com/
3. www.hometime.com/
4. www.lowes.com/lowes/lkn?action=home
5. www.msue.msu.edu/msue/imp/mod02/master02.html

were the top 5.

Any thoughts?

[Jill]

Welcome laertes!

I think it's a good theory. I've often thought that if I were a search engine, that's how I would do it. It wouldn't surprise me at all if that was one of the goals they're working towards.

[gbot]

I'd like to suggest a third theory, to muddy the waters even further, i.e.

Good theory! You should get a job at google and show them how to do this!

Here is what I see.

Not a theory or a guess; but what I see... with my own eyes.

Ok, lets look at the search phrase "search engine optimization".

Who are the top 3?

1. Submit Express
2. High Rankings (go Jill!)
3. Bruce Clay

Ok, would everyone agree that Jill's Web site is optimized? (i think so)
Would everyone agree that Bruce's site is optimized? (i think so)
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.
Why are they ranked first on quite possibly the most competitive search phrase? (i don't know, but "links" is looking like a good answer)

I know you guys are tired of "links, links, links". But they are a big piece of the puzzle. (along with 100's of other factors)

gBot

[mcanerin]

lets choose a few of each type.

That's certainly something that happens in some SERPs, as far as I'm concerned. I don't know if I was the first person to propose that (I don't think so) but I remember quite clearly being roasted publicly by Danny Sullivan for proposing it.

I still think he's wrong and I'm right.

I suspect it's more likely to happen in vague, high traffic SERPs, where you have a lot of people looking for something, but the search engine isn't totally sure what, than as a sorting function.

This is not to say it might not be a good tool in the toolbox, and help alieviate the overflow in some SERPs, but once you get to a very specific search, like "buy HP 116CX free shipping" then it obviously would not be helpful to show museums and directories. Even many less specific searches would not be improved.

It would certainly help for vague searches though.

Ian

[Michael Martinez]

Ok, would everyone agree that Jill's Web site is optimized? (i think so)
Would everyone agree that Bruce's site is optimized? (i think so)
Now, let's take look at Submit Express. Are they optimized? (i DON'T think so)

Well, in looking at the cached text version of the page on Google for Submit Express, I would say that, yes, they did indeed optimize it.

Whether that was all they needed is anyone's guess, but I would be willing to side with you on the links-carry-more-weight for this search because it's hyperoptimized.

If your purpose is to argue that links count more than other stuff, you have to come up with searches that have not been hyperoptimized.

So, you cannot use "search engine optimization", most commercial travel related searches, most real estate searches, most poster store and similar online commodity searches, pharmaceutical searches, ring tones, etc.

It doesn't take long to get past the list of hyperoptimized search expressions, but sadly too many people in the SEO forums think they have some sort of relevance to the question of how much links count. Solely because these are hyperoptimized searches, they don't have any relevance at all.

It's equivalent to injecting a rat with about 1200 soda cans' worth of some chemical every day for a year and concluding that drinking soda is unhealthy for you because the poor rat dies.

[Martin C]

lets choose a few of each type.

Well I think the only problem with this is that is looking at it from the search engine's point of view.

People searching don't want to be given a selection, they just want the exact thing they are looking for.

The more relevant the information is that is returned the more they will use the search engine. You could argue that by giving them a cross section of results they might be happy but we are a fickle bunch and will just complain that not enough of the 'right type' of results were returned.

You can see a lot of the SE are trying to make it easier for people to be more precise with their search phrase, Google occasionally put in the 'See results for ....' when you put in a very generic term, Yahoo have a similar 'Also try......'.

At the end of the day the SE are expected to be mind readers and know not only the exact thing it is that people are looking for but know if they are searching in 'business' and 'leisure' mode.

The SE have become a victim of their own success where people now actually expect to find what they are looking for by entering one word search terms and most the time the SE manage to find what it was they were looking for.

Occasionally it is good to step back and not take this stuff for granted.

[Shane]

If your purpose is to argue that links count more than other stuff, you have to come up with searches that have not been hyperoptimized.

Again, though, you have to go back to the original question, which was: given 50 to 75 sites all equally well-optimized, how does a search engine decide who gets the top spot? The conversation has drifted a bit, though, and because I do think that links count more than other stuff--in both Google and Yahoo, and probably MSN as well--I'll tell you why.

A little over a year ago, we did a test to try to ascertain whether Google and Yahoo put equal weight on links and link text. I didn't expect them to.

So, we took a phrase ("resume writng", intentionally misspelled) that had enough results in both engines (almost 800 each) that we would actually be able to determine something from any movement we saw.

We then put a link on the homepage of Jobs.net to CareerBuilder's "Stellar Resume" page with link text of "resume writng" (without the quotes). CareerBuilder's page did not contain the phrase "resume writng," nor had it ever contained it, and thus didn't appear anywhere in the 800 results on either engine.

Since both engines had the same number of results and, theoretically, basically the same set of pages, if CareerBuilder's page moved to #200 in Google but only #400 in Yahoo, we could conclude that G put more emphasis on links and/or link text.

The link showed up in the cached version of the Jobs.net homepage on both G & Y on Friday, October 22. By Monday, CB's page had gone from not being ranked at all for the term to being #1 for it. By the next Friday (Oct. 29), it was #1 in Yahoo, too. When we removed the link, the page again disappeared from the results completely.

(Since that time, MSN has rolled out their own search engine which, from everything I've seen, is even more susceptible to link spam than either G or Y simply because they don't have anywhere near as sophisticated filters in place.)

I drew from this simple experiment that both G & Y placed a great deal of emphasis on link text. Going from unranked to #1 in both SERPs by only changing a single link seemed to show that. Which one placed more emphasis (the original purpose of the experiment)? We really couldn't tell since the page went to #1 in both engines. All we succeeded in showing was that they both loved links.

Something else that I draw from this now, though, is that links are actually the most important element. CareerBuilder's page was able to leap over many pages that had intentionally been well-optimized for this misspelling, but being well-optimized didn't help them against one page that had only a simple, well-formatted link going for it.

Now, it's entirely possible that I've completely misinterpreted the results of this experiment and drawn conclusions that don't actually follow from the evidence. So, I would love to hear what other people think. Does this indeed seem to indicate that, at least 14 months ago, both G & Y valued links above all else? Or are there other factors at work here?

[Michael Martinez]

Again, though, you have to go back to the original question, which was: given 50 to 75 sites all equally well-optimized, how does a search engine decide who gets the top spot?  The conversation has drifted a bit, though, and because I do think that links count more than other stuff--in both Google and Yahoo, and probably MSN as well--I'll tell you why.

And I'll tell you that 200 words and 100 words can both be "equally well optimized" for a targeted expression, but one set of words will be deemed more relevant than the other. Which set? It depends on how relevance is measured.

That means it could be either set, and therefore you cannot accurately show that either set should be ranked first on the basis of links because links don't determine relevance. link text can be used (if it is associated with the target document's text) to assert relevance. But in that case you're no longer dealing with either 100 or 200 words of text. You're dealing with (100 words of text plus the text of all the links pointing to this document) and (200 words of text plus the text of all the links pointing to this document).

Now you have no way of knowing whether the documents are "equally well optimized", because the optimization has to be measured against all the text associated with the documents.

Many people have done "tests" through the years. All the tests are flawed, however.

....We then put a link on the homepage of Jobs.net to CareerBuilder's "Stellar Resume" page with link text of "resume writng" (without the quotes).  CareerBuilder's page did not contain the phrase "resume writng," nor had it ever contained it, and thus didn't appear anywhere in the 800 results on either engine.

This search shows why your test was not valid:
http://www.google.co.....sume +writing

There are thousands of links pointing to CareerBuilder from documents that include the words "resume" and "writing". Putting that expression on CareerBuilder's page just made its document more relevant, but the relevance was already there. So you didn't start out with an equal base of test documents (the 800 which originally showed up in the search results may or may not have as many inbound links from similar documents).

Note that I am not talking just about link anchor text. Google and other sources say they are now looking at the documents that link to a given page, not just their link anchor text.

It's not so much that you misinterpreted the results of your test as you really didn't do anything to vet the base dataset. You had no way of predicting how your words should have affected the test because you were working with an incomplete data model.

These kinds of tests are misleading precisely because the data models are not well defined. Without a valid data model, you cannot validate the results of your test.

The CareerBuilder page could have benefitted from more than one factor beyond your control as well. In order to eliminate other factors, you should have run similar tests on other sites in different industries for different keywords. And you also should have tracked at least an equal number of search expressions for similar document sets where you did not alter any content to see how stable their rankings remained during the same period.

The bottom line is that to do a reliable test, you have to track a LOT of search expressions and you have to divide them into two groups: groups where one document is changed and groups where no document is changed. Then, when you compare the results, if you have no statistically valid discrepancies, you can weigh the results and draw a reasonable conclusion, such as that links make the difference or that they don't.

Even then, you might be misinterpreting results.

[Shane]

Michael, you missed several key points of the test. The phrase wasn't resume writing and the page we were linking to wasn't the CareerBuilder homepage. Had either of those been true, yes the test would have had too many uncontrolled variables.

We picked a phrase that there were already pages well-optimized for. We then linked to a page that wasn't optimized for that phrase at all. That page then jumped immediately from nowhere to #1. To me, that clearly shows that both G & Y considered one link from a trusted (read: well-linked to) page to be more reliable for determining relevance than any other on-page factors.

So how does Google decide what 10 pages to put first?

...and therefore you cannot accurately show that either set should be ranked first...

The first quote is from the post that started this thread. The second is from yours above. How Google should rank pages and how they do rank pages are two different discussions. One deals with drawing conclusions based on observed cause and effect. The other is largely a philosophical discussion.

Ian's earlier post is a great post related to the latter (I don't know that I've ever read anything from Ian that wasn't a great post), and I believe you could make a good philosophical argument that SE's could rank pages while giving only very minimal weight to links. I've just never seen any evidence that would lead me to believe that they do, though. Every single thing I've ever seen leads me to believe that a page with just a few good links will rank higher than a page with a ton of great content every single time. In other words, even if you knew exactly how the algorithms viewed on-page content and modified your page accordingly, I could get mine ranked higher with just a handful of really good links and zero on-page work.

I would love to be shown that that's not the case, though