|What determines relevancy?|
With talk of relevancy all the rage, just what is it?
I've read through the posts and relevacny seems to be the new buzzword. I've read here that some webmasters have become so paranoid with regard to the topic that they have been emailing folks to TAKE DOWN links that are not "relevant."
But what exaclty is relevance and how is it determined?
What factors go into an algo to determine if a link or a page's content is relevant?
How does a compter model get its feel for relevance?
I've started this thead in the hopes of digging into these and related questions, and hopefully to get feedback of a technical or practicle nature.
Technical or Practical I'm not sure I am, but here's my hapenny-worth:
The most basic method of 'relevance' that I am aware of is the Naive-Bayes method which (generally, with lots of missing steps) looks at the probability of a keywords appearing in one of a number of groups, based upon this, can decide what it is about documents which differentiates them.
e.g. sites "cat lovers love to stroke cats" and "dog lovers love to stroke dogs"
Can be differentiated by "cat" and "dog"
Another 'group' "cats and dogs eat biscuits" and "dogs and cats eat cheese":
Can be differentiated from the first group not by 'cats' or 'dogs' (the probability being the same), but by 'love' or 'stroke.
Hence the relevancy of documents can be built as a 'score' of how much a document is similar to another.
So, in a very basic way, "cats dogs and chickens" is 66% similar to "dogs chickens and foxes", and 0% similar to "boxes, pens, and calculators".
Well, that's my input to help get this thread started, I'm really interested in what the clustering-guru's here will come up with...
Just bumping this up to the top as I am very interested in hearing more about this as well.
I seem to remember reading somewhere about theming being related to node theory. I'm not a real pointy head, but the gist seemed to be that the must authority documents have exponentially more links from other documents with the right on page factors / anchor text and all the relevant documents will only be separated from these by a couple of degrees of separation or something?
Confused - I am!
Not very helpful but I tried and perhaps it will encourage someone who really knows what's what to post, even if only to tell me I've got it all wrong!
>I've read here that some webmasters have become so paranoid with regard to the topic that they have been emailing folks to TAKE DOWN links that are not "relevant."
Then I'd say these webmasters are fools. First, Google wouldn't want to penalize a site for irrelevant links. If they did, a competitor could hose the competition's Google rankings by arranging a lot of irrelevant links to them. Second, it might be hard for an algo to spot relevant links. For example, a bunch of local music bands in a town might link to each other because they are relevant to users of their site (people in the same town may be interested in music bands in the same town.) The algo might miss the geographic connection.
Much more likely Google would give an algo boost to links that are identified as relevant. Thus, a link where no relevancy could be spotted would give little benefit, while one on a site on obviously the same topic (such as the linking site is in the same ODP category as the site linked to) would be given more weight.
We donít spend a lot of time contemplating the technical; how Google compiles data when it comes to relevant links, but we believe they do. We do spend time on the practical; finding relevant links as we believe these help a great deal.
We have found any links (almost) help, and relevant links help even more.
To be sure anyone could try to destroy my sites with tons of links, I wouldn't mind.
|Then I'd say these webmasters are fools. |
Which link is a more relevant link? (A high PR authority site aside). A link from page with one or more keywords in the meta title? One form a page with a high percentage of similar keywords in the content? Or one that links to a variety of authority sites about your keyword as well as to your site?
As relevancy has become more important (well as the rumor goes) are all those links pages becoming a serious liability? Does anyone think they should be scrapped? After all, there is not more greater source of distortion and irrelevancy than links pages if you're trying to have it that a link represents a vote for good content.
I'm not sure if I've ever read a Google postion on links pages. It seems if they want to protect Page Rank, which I assume weighting relevancy highly is intended to do, wouldn't it be much easier to just let it be known that Google frowns on haphazard links pages by putting it in their guidelines?
Anyone else have any input into determining relevnacy?
If you have a page about round blue widgets, do a google search for round blue widgets and there is your answer... a list of the most relevant sites starting with the most relevant and going to the less relevant.
If they aren't in your own results they aren't relevant.
You can go into semantically related topics and bleed it our from the main game, but I'd stick with the search results.
*The algo might miss the geographic connection.*
Think we'll be hearing a lot more about link location in the future...
Just thought I'd throw in a prediction for MMV ;-)
I am not sure that I agree 100% with this. I just picked up a Run of Site link on a newspaper site that was ranking #4 for part of my keyword. Then, just one day after my link went live, my google and yahoo rankings for my KW (which were in the top 10) disappeared. I'm not sure what to think of this.
The newspaper site utilized a portion of my keyword throughout the Meta Title tags so I thought it would be relevent. It also ranked very well with a PR7.
I currently have a PR7 and am fairly well established.
There has been some discussion on the board about the effect you describe. Run of site links seem to trip a penalty / filter that, at the very least, appears to negate their impact.
here is a real high tech method for determining relevancy. Go look at a site. Does it seem related to your site? Will it be useful for your users? If so, give it a link.
*If* google is concerning itself with relevancy, then recognising this will be the goal that they are striving towards.
It doesn't matter all that much exactly how they do it. It just matters that as they get closer to the ideal, the better they will like your site.
"Run of site links seem to trip a penalty / filter that, at the very least, appears to negate their impact."
If this occurs what can you do to come out of penalty? Remove the links? Deopimize the page they were pointing to?
|Go look at a site. Does it seem related to your site? Will it be useful for your users? If so, give it a link. |
What if these highly relevant links would all be competators? I agree for hobby sites and informational portals, but the twist for many that provide business services is that what in reality might be a good link from a site that would generate actual click-through traffic, might be deemed by Google as irrelevant. In the case of one of our sites, our service is relevant to a huge number of businesses, yet other sites may not make any mention of any of the keywords on their site (and if we're lucky, won't link to any other business that provide similar services.)
Take a case like SEO (not our business), normally a link would be relevant and could generate clients virtually anywhere on the net (assuming that most business people don't know much about the subject and would like to generate online traffic), but would Google consider it relevant if it appeared on say, a website that sells ohh, wholsale goats milk and lithium?
|Run of site links seem to trip a penalty |
I'm not sure this is the case unless the source website is on your own network so it becomes extreme cross linking. Otherwise I've learned that only the source site may be subject to a penalty and not the target site. As Google says, there is almost nothing a competator (read: anyone except you and Google) can do to harm your SERPS. A lot of sites that offer this sort of plan have the links going though the cgi bin which makes the link watered down if not usless... (or am I wrong about this last part?)
|What if these highly relevant links would all be competators? |
Then link to them if you want. Don't link to them if you don't want.
In the Real World, almost any business will tell you who their competitors are. In fact, the best companies will even specifically send you to a competitor if they think that the competitor can serve you better.
If you cannot find what you are looking for at the nearest Lowes, just about any employee will give you directions to the nearest Home Depot or local plumbing supply store.
In the online world it is folly to think that you can keep a customer from finding a competitor. They probably found your site from a list of competitors.
|I agree for hobby sites and informational portals, but the twist for many that provide business services is that what in reality might be a good link from a site that would generate actual click-through traffic, might be deemed by Google as irrelevant. |
I sure hope you did not mean irrelevant. Google scores things, it is not a binary switch where something is either relevant or irrelevant. While they may decide that something is "might not be relevant" I am certain that they quite well understand that they cannot and should not jusdge something to be "irrelelvant".
By the way, GoogleGuy has posted that he likes off-topic links because of the way that thing occur naturally. A kayak school posts a link to his car repair place because of the quality of their work and their honesty. (and actual example and how I found the mechanics that I brought my car to till I moved.
Remember, Google employees are generally a lot smarter than the paranoids that sit around speculating on the SEO boards. they have spent a lot of time sitting around contemplating and researching the nature of links.
|but would Google consider it relevant if it appeared on say, a website that sells ohh, wholsale goats milk and lithium? |
Does it matter? *If* google even cares about relevancy, their goal is to understand and emulate what people considre relevant. Their goal is to be able to understand that it in fact does make a lot of sense for a kayak school instructor to recommend a Subaru repair shop. (start watching for kayak racks, what brand of cars has the highest percentage with racks?)
The point is that you shoud forget worrying about decoding what is probably a non-existant relevance algorithm, and work on getting relevant-to-a-human links. And if you are still concerned about it, build sites a "relevancy transition page" for them to link to. "Inventory management equipment for wholesale dairys" or something like that.
There is no real advantage to trying to figure out what google will consider relevant in any particular month, it is google's job to figure out what you consider relevant. The time you spend trying to figure it out, you could be working on building your site and getting relevant links.
Exactly what do you mean by "Run of site links"?
No single answer can explain relevancy but here are the names of some papers with clues.
The Anatomy of a Large-Scale Hypertextual Web Search Engine
Authoritative Sources in a Hyperlinked Environment
I would like to believe relevance is an overalls stategery of unique content, related keywords, related topics, and quality themed links. A paper on stemming would probably have some good info on relevancy - anyone have suggestions?
|Exactly what do you mean by "Run of site links"? |
Usually links which run across the bottom of every page on a site, e.g. sponsorship of a forum.
It's similar to ROP (run of press) in print advertising, it essentially means that they'll put your ad wherever there's space. It's usually the cheapest form because you can't target it. Alternatively on the net, it could conceivably be used to mean your link will appear on every page (or every page within a section depending upon how large the site is) because websites aren't like print... a web page can physically become larger to accomodate content.
|In the online world it is folly to think that you can keep a customer from finding a competitor. |
I appreciate the long post Dave, but I think it's a bit off point. It's not about keeping them from finding a competitor, its about not giving a competitor a boost at your expense.
|If you cannot find what you are looking for at the nearest Lowes, just about any employee will give you directions to the nearest Home Depot or local plumbing supply store. |
Would the Lowes employee redirect the customer to Home Depot if they CAN find what they are looking for at Lowes? In my case, I assume we're selling the same "pluming supplies" as our competators... but thanks for the input.
[edited by: Rollo at 9:22 pm (utc) on Dec. 8, 2004]
*Remember, Google employees are generally a lot smarter than the paranoids that sit around speculating on the SEO boards.*
To be honest BD, I might like to think that, but I'm yet to be convinced,I suspect most GEs would be horrified at the detailing some of us paranoids get up to.. :-
|I appreciate the long post Dave, but I think it's a bit off point. It's not about keeping them from finding a competitor, its about not giving a competitor a boost at your expense. |
Funny, I still think that it is completely on point. You just miss the point.
I never said that you had to put a link right on your PVC schedule 40 2 inch tee fitting you your competitor's page for the exact same item.
I was just pointing out that it is not necessarily bad business to be willing to point out your competitors.
The problem with most people in the retail trades is that they do not even know what they are selling. While your customer may be buying your products, what you are really selling is feelings about your company.
I just do not see where a link to a good competitor of related products is in any way giving them a boost at your expense, as long as you have confidence in your own company.
Yeah, you can play the PR argument, but I just have not seen that make a difference in practice. In fact, I have seen it go the other way in several instances.
In one industry that I am very familiar with, there are two big players. They are the only ones that work on an industrial scale. Everyone else is in the one to five employee size.
The top three sites for the highest traffic keyphrase are all little guys that actually link to all their competitors (except for one that is literally a delusional paranoid) with the explanation that they are prowd of their own products, but they recommend that you check out the designs by these other fine companies to make sure the customer gets the perfect product for their needs.
Believe it or not, this does build the trust of your customers.
I don't think that it is appropriate for every online company. But I do think that you should not dismiss it out of hand either. It is a great way to build customer trust, and very few of them will look elsewhere when they find what they want at a company that they trust.
|Would the Lowes employee redirect the customer to Home Depot if they CAN find what they are looking for at Lowes? |
No. And that has almost nothing to do with what I suggested. There is a huge difference between redirecting someone and providing them with useful information.
And I would highly recommend that you serve up your own product pages rather than setting up a redirect to your competitors. You should at least give them a chance to buy from you.
For evey page on the internet there are othe pages that are related to it that are not necessarily competitors. If you are just looking for keyword matches, you might have more trouble finding them. If you follow my advice, I suspect that you will find a lot more of them.
And if you think my posts before was about linking to your competitors, that was only a sidebar, and you missed the point of my post about what is relevancy.
|To be honest BD, I might like to think that, but I'm yet to be convinced,I suspect most GEs would be horrified at the detailing some of us paranoids get up to.. :- |
I would use the word ammused myself. They would be impressed when you get it right, and fall on the floor laughing at some of the insane theories.
Then there would be a certain amount of bewilderment at what some peole seem to spen their time worrying about. About the only thing that would horrify them would be when a incorrect rumor gains legs and people with good sites, start rashly changing their sites because of that rumor.
|The point is that you should forget worrying about decoding what is probably a non-existent relevance algorithm, and work on getting relevant-to-a-human links. |
Hi again Big Dave,
I did get this point about relevance, I just don't quite agree with it. Your point is to essentially to ignore SEO and focus on the customer/visitor and that Google will follow suit and reward you? I'm not sure I agree with that. Imagine a new webmaster that thinks it is useful to clients to link every page of his somewhat off topic sites to every other page. Totally innocent, but what he or she doesn't know just might kill them. There seem to be too many distortions, too many fine sites getting sunk by not understanding what they are doing or at the very least not getting anywhere. "Build it and they'll come" doesn't apply online. I'm not suggesting building a website entirely for Google and ignoring real visitors, but the two aren't mutually exclusive. Since most webmasters must rely on trading links (or even buying directory or other links) rather than waiting for them the natural way, the relevancy issue gets right to the heart of link building and therefore SERPs. We're all here I think becuase we want to learn more.
Judging from the number of posts you've made here, it seems like you for one are doing anything but ignoring seo ;-)
Thanks again for the input Big Dave.
Anyone else have any theories/info about relevance? I think we have one resounding vote for stop wasting our time with it from BD.
|Your point is to essentially to ignore SEO and focus on the customer/visitor and that Google will follow suit and reward you? |
Nope, that is not it at all.
My point is that you already *know* what is relevant. It is a complex process in your brain.
Googlers know that determining relevance is a complex process, and they will try and take as many of these factors into account when trying to program that.
So it quite clearly is not as simple as matching keywords.
You are suggesting ignoring the most incredibly complex and accurate piece of input you can get simply because it came to you with no effort, in favor of a little checklist that will require a lot of effort and produce inferior results.
|Imagine a new webmaster that thinks it is useful to clients to link every page of his somewhat off topic sites to every other page. Totally innocent, but what he or she doesn't know just might kill them. |
No one actually believes that they are doing something like that for their users. They are boing it for themselves, without regards to what the users would really want.
And if they design a sites like that "for the users" then hopefully their sites will not rank high, because as a user, I really hate coming across sites like that. If they get dinged, then google is doing their job.
|Anyone else have any theories/info about relevance? I think we have one resounding vote for stop wasting our time with it from BD. |
Nope, you definitely did not get it.
I told you the simplest, and by far the best way to determine relevance. I answered your question exactly. From my first reply:
|here is a real high tech method for determining relevancy. Go look at a site. Does it seem related to your site? Will it be useful for your users? If so, give it a link. |
Where do I say to stop wasting your time considering relevancy? I suggested what I consider the best and most accurate way to determine relevancy that you could possibly ever come up with.
You asked questions and I answered.
You can feel free to ignore my advice, but you should at least try to get an accurate understanding of what I have said before you do so.
Ok... anyone else care to jump in? My fingers are starting hurt.
"By the way, GoogleGuy has posted that he likes off-topic links because of the way that thing occur naturally."
Ancient history. That concept is ludicrous today. I can't see anyone disputing that the vast majority of offtopic links on the Internet today are wholly unnatural and useless.
On the other topic though, it is a sure sign of strength to link to competitors, and weakness not to. With widget selling sites all they need to do is link to dmoz or Yahoo categories to send the same topical message. if the idea of that scares you, build a better site.
|I can't see anyone disputing that the vast majority of offtopic links on the Internet today are wholly unnatural and useless. |
I will. The vast majority of the web is non-commercial, whether web professionals want to admit it or not.
Personal sites like blogs and personal webpages almost always link freely to whoever they want on whatever subject interests them.
Though, in reality there is no such thing as a natural off-topic link. It was relevant to the person that put it there, and that is what counts, so I guess that I agree with you, and I will in fact go one step further "ALL off-topic links are unnatural, because natural links are by definition relevant to the person in control of the site."
And I'm not sure how you have a right to declared GoogleGuy's year-old opinion is "ancient history". Has he told you that his opinion has changed? Do you know something that the rest of us don't?
I was just pointing out that at least one person at Google understands that sometimes there are good reasons for what might appear to be off-topic links. And if someone is of that opinion, they are at least going to argue against declaring a link to be "irrelevant" when they have meetings to discuss what to do with the algo next.
Obviously off topic links sometimes show merit. Duh. Just as obviously the Internet is awash with literally millions of blog spammed garbage that dwarfs the legitimate blog linking.
"I was just pointing out that at least one person at Google understands that sometimes there are good reasons for what might appear to be off-topic links."
You disputed that "the vast majority of offtopic links on the Internet today are wholly unnatural and useless", which clearly assumes "sometimes".