Forum Moderators: Robert Charlton & goodroi
I'm trying to optimize my link structure so more of my pages appear in the SERPs.
The original pagerank formula would say that each link gets an equal split, but that's clearly not the case as many have tested and discovered.
One shift from the original PR formula has been that links are weighted differently according to which segment of the page they appear, with the main content area voting the most juice. I'd also say that a link in the main menu votes more juice than a footer link, although I haven't nailed that factor as solidly.
So how does a second or third link on the same page to the same target page work? Some have claimed that only the first anchor text "counts" - if so (and I'm not yet convinced) would that also mean only the first link transfers any PR?
I don't think so, mostly because there is a signal here, and Google is not likely to ignore any clear signal. At the same time, I'm also sure that multiple links on the same page to the same target page will not all get the same weight.
Does PR flow out of that page at a rate of 2x or 3x, but only into the linked-to page at a rate of 1x?
PR doesn't "leave" a page. A better (but still not perfect) model is that a page "votes for" other pages by dividing its own PR through all the links it holds. The imperfection in even that model is this - it ignores the fact that PR is an iterative calculation.
In my opinion, the best thing you can do is to minimize the total number of links on a page - keeping the template links WELL under 100 and placing other content links for the benefit of helping your visitors find the most related pages. This also tends, naturally, to structure the site so that Google has a clear signal for what the best entry pages are.
My sense from working with major websites is that this also works well for whatever Google is doing today with PR calculations behind the scenes.
Whatever the case, a page linking to itself certainly certainly a poor user experience to click on a link and go... nowhere.
Some have claimed that only the first anchor text "counts"
I personally don't think it makes sense to do that because a large document that references (parts of) another large document would naturally use different anchors.
Some have claimed that only the first anchor text "counts" - if so (and I'm not yet convinced) would that also mean only the first link transfers any PR?
I thought this was interesting and ran a few tests a short while back. As far as the testing went - yes - only the first link passed any anchor text, in two independent tests.
As far as the PageRank flow, I have been discussing this extensively with some friends. We have tried to test it, but found it difficult/impossible to measure the results. The best we could come up with was that putting 3 links to the same page out of a total of 5 would pass more than just putting 1 link out of a total of 3, whilst diminishing the overall value of links on the page.
How much more value for 3 instead of 1? Marginal at best, in the test case.
Again, the test was for in-text links only, with little or no other factors... and not really all that helpful.
putting 3 links to the same page out of a total of 5 would pass more than just putting 1 link out of a total of 3, whilst diminishing the overall value of links on the page.
So duplicate links do pass additional PR, and they do diminish the value of each link on the page. Did your test indicate if the value of each link on the page was diminished at a reduced rate due to the duplicate links since they don't pass a "full vote"?
To the best of my knowledge - yes they pass PR, but I don't like to think of it in terms of "additional". The diminished rate was something we discussed at length, but ended up with two different theories and no consensus. Testing was inconclusive on this point.
I want to stress that it was difficult to measure the results. I don't have any black and white facts saying that 2/5 is better than 1/3. Actually, I am pretty confident that the 1/3 is the better way to go... unless you want to give other links on the page an overall lesser value, as per: n/3 > n/5. Again here, putting diverse links on that page will dilute juice flow just as much, but pass it on to other pages, which makes more sense.
This is a very fuzzy area imo, and when making choices about putting duplicate links I tend to go with what makes sense for the user. Best advice I can give.
1. Every one of the links with the same target page gets a full share of the PR vote.
2. Only the first anchor text passes influence - with the exception of stop words such as "Home".
3. If you use nofollow on even one of the multiple links, they all stop passing PR.
The "full share" in #1 surprised me. My sense of it has been that the PR transfer from the extra links is damped. But Bruce uses sound testing procedures, so his results are worth letting others take on board for consideration.
[edited by: tedster at 2:19 am (utc) on Nov. 12, 2009]
Duplicates must contribute to the 100 links per page maximum then?
