|Does anchor text work to the 2nd generation?|
I'm beginning to suspect that a keyword in anchor text can exert an influence over two "hops", and not just one. In other words...
A -keyword-link-> B
B --no--kw-link-> C
...results in C being helped on the keyword search. I don't have enough data to quantify this one, but I'm wondering if anyone does. I keep seeing things that are very suggestive.
Yes, it's been that way for quite a while - at least in the instance that first brought it to my attention. In fact, I've mentioned it in posts a few times. It's definitely got nothing to do with on-page factors in that particular case, because there *was* nothing on the page. I mean *nothing* - not a speck of text.
Found one - Message # 22.
Still sitting at #1 & #2 indented out of 2,610,000. And still nothing on the pages.
[edited by: Marcia at 7:09 am (utc) on Sep. 11, 2004]
You know, now I remember reading your previous post. Guess it was only an abstraction to me at the time, but now that the evidence shows itself in front of me, it all seems so different.
I don't have any examples that are so very clear - so it's good to hear some back-up for my gut feeling. I've been working on a group of related sites, and I sure thought I saw keywords having this kind of rolling influence. But it can be hard to sort it out from other algo tweaks that are also influencing things.
[edited by: tedster at 7:12 am (utc) on Sep. 11, 2004]
this is the first time I have noticed some one acknowledge this.
This is definate evidence to support theme-ing in the algo.
I think that this theme-ing factor has been turned up in the algo lately. I have a lot more PR than my competitors but very little is from links from on topic pages and my position has suffered. I'm interested to know other peoples experinces.
I have a specific question. Does this affect occur when two of the three pages are on the same domain.
eg. page A.something.com links -> "bananas" to page B.something.com links -> "anything" to page C.other.com
or were all three on seperate domains?
In the cases that brought it to my attention, it was three different domains every time. Different owners, c-blocks, the whole nine yards different.
It would make sense for link text to have an effect across two hops. Out of the different characteristics a backlink page - or any page - can have, inbound link text is one of the most important ones.
Other factors being things like the keyword being found in the anchor text, partially found in the anchor text, in the page title, page URL, H1 tags, being within a few characters of the link. These all seem to cause pages to show up in allinanchor searches.
Since Google highly values inbound link text in comparison with on-page factors, it would seem likely it would also be used when evaluating the relevancy of a backlink to a keyword.
Same domain Matt, but I've also seen something else. With a C -> B -> A link where the two-word phrase in question was not in anchor text, since B was relevant to the phrase and the phrase appeared in proximity to the link from C -> B it gave A an unshakeable boost for a few phrases that include that two-word phrase. The link at site C happens to be from an authoritative source beyond a shadow of a doubt.
Things like this are anecdotal and by no means indicate what we could call a "themed algo" but certain elements that go into defining theming can come into play. It's a matter of isolating individual factors, and then testing and/or applying them, which is the harder part.
Am well down the road in testing this.
Will wait a while longer to ensure a few shuffles in SERPS.
Not quite clear what is being suggested here...
If I link to Tedsters page on widgets with "blue widgets", and he has a link with "green sprockets" to Marcia's page on sprockets, that page could rank for both terms?
I think that the implication is that Google gives weigthting to the grandparent backlink. In order to get the best opimization for your pages you must consider this new (for me)development.
Seems to me like the suggestion is that anchor text may not be simply page specific, but can transfer, much like PR.
It seems pretty simple to me:
Anchor text of links TO a page are considered part of that pages properties, similar to it's title body text and so on.
When considering the ranking of a page, it's citers, i.e. pages linking to it, are considered, what are it's topics, keywords, what do they rank for. Since anchor text is a property of a page, it must be considered if parent pages are considered. tehrefore links to parent pages are being considered. I'm not sure that this owuld be a new development.
> Does anchor text work to the 2nd generation?
I believe it probably does, and thanks Marcia and Tedster for sharing your experiences.
I built a tightly themed pyramid style website a couple of years ago (with the help of a lot of WebmasterWorld folks here - thanks!) and the site gets top ranking for dozens of keywords, including #1 spots many competitive 2-word phrases.
This 2nd hop influence would explain a lot IMO.
Also, in this month's Planet Ocean SEM newsletter, they announced the results of a very relevant in depth study. The big conclusion of the study was this: The number of links pointing to a page linking to another page has a huge impact on the ranking of the target page.
To put this another way, let's say I have a choice of getting a link to my page A from either page B or page D. And lets say that each page B and D have multiple unique inbound links pointing to them. You could express these two options as 500 > B > A and 25 > D > A, where 500 and 25 are the number unique inbound links pointing at each page.
Also, presume that for both page B and Page D each has similar topics, a similar number of outbound links, and a similar PageRank. The conclusion of the Planet Ocean study is that the link from page B would be much more valuable than the link from page D due to the number of links (500) pointing to page B.
I that think this is a very significant finding, and dovetails in with your anchor text 2nd generation effect finding.
>Yes, it's been that way for quite a while - at least in the instance that first brought it to my attention. In fact, I've mentioned it in posts a few times. It's definitely got nothing to do with on-page factors in that particular case, because there *was* nothing on the page. I mean *nothing* - not a speck of text.
This theory would tend to explain a real world case I know of. I run an amateur site on widget safety. "Widget" is part of the site name, and thus I have lots of links with widget in the anchor text. There is another site on widget safety with lots of anchor text with "widget", and we both dominate the SERP for widget.
There is yet another amateur site about widgets, and they also rank well. This site just happens to use a brand name, and just about everyone links to them with that. They have not a speck of text on the home page. However, both my site and the other one dominating the SERP link to them on the home page. The theory that anchor text can carry over an extra hop would explain how a page with not a single instance of widget visible can rank well. There are other possible theories why this site could rank well, such as they are listed in the widget ODP category the along with mine. However, other pages on that ODP category don't rank well for the widget SERP, and neither me nor the other site dominating the SERP link to those. The most simple explanation for this site ranking well is that anchor text from sites linking to them is carrying over.
|...it must be considered if parent pages are considered. therefore links to parent pages are being considered. I'm not sure that this owuld be a new development. |
But if the Google algo is way and has been the case for a while, then such a multi-generation influence must also be either:
1) cut-off absolutely at some specific number of generations
2) damped (maybe a PR-like function, as mentioned above)
Otherwise, every page on the web would end up being about everything. I had always assumed without looking very hard at it, that there is a cutoff after one generation of influence. But now I very much suspect that is wrong.
It makes sense in a crazy kind of way....
We know from the "similar pages" fiasco they expected linking to be topical/relevant, so on that basis they could assume anchor text could be relevant X steps down the linking line....
>>Otherwise, every page on the web would end up being about everything
I don't think it's any kind of a magic formula that could stand on it's own, it's more like one part of the equation with other factors entering into it.
This is an interesting read related to successive iterations in linking:
Experiments in Topic Distillation [almaden.ibm.com]
On linking. I have performed a series of tests of my own over the past season and noticed that themes are clearly carrying more weight than in the past. It appears that most of the SE algo's are learning in this direction now, or at least trying to.
Another thing I have noticed and found a little odd was spotting a high SERP result on a contact page that was/is supposed to be a SSL page? Yes, I was perplexed as I believed it would never even be spidered. Surprised I was! It beat every other page I had going for a client and to this day I am still puzzled by it. Ironic. What we want and what happens are sometimes contrary.
I wonder when google will sit still on changing its page ranking algo long enough for anyone to actually fully grasp it! Seems it has changed recently more times than I can count..
|I don't think it's any kind of a magic formula that could stand on it's own |
I agree - it may not even be an explicit part of Google's special sauce, but just a side effect of several other interacting factors. Still, this seems to be a good way to think about things, whether there's precise math there or not.
One of the places I plan to work with this is where a page has acquired a strong link citation from an independent site -- whether from a content page, link page or directory category page -- and that linking page is very weak in "potency" for an important theme or keyword.
Such a page may be outside my direct influence. Perhaps some directory editor perversely avoids using a particular keyword, or a website is authored by someone who is not willing to communicate, but has given my page a strong, inline link. These can be the most valuable and trusted kinds of links for Google: unaffiliated (in Hilltop terminology) with the target page and therefore an "independent" citation of relevance to a particular topic.
So here's my test. If I link TO that page, becoming my own grandfather if you will, I may be able to increase the weighting of a particular theme for its links.
It's like saying to Google, "Look, I know this specific word set isn't on my target page, and maybe your semantic algorithm still hasn't establish a connection between these particular word sets, but this page really IS about "keyword".