Link not working, even when I manually visit SEW.

Link apparently working now, in IE and FF, so let's give it a shot. For this topic, it might have been server overload.

I haven't read the article carefully enough to see if I agree with the methodology, but in the test run the conclusion is that the Panda dropoff, when it happens, is a steep dropoff, inbound link juice can help to a degree, and sections of a site were affected (as opposed to single pages).

What do they mean by "removing low quality pages". Complete removal, noindex, nofollow, blocked by robots.txt?

Too bad that the article is a bit vague. It would love to have seen some hard data to see how analysis was done.

Conclusions -
* Backlinks are impt
* Improve quality can bring back site
* Removal of low quality pages can bring back site
* Moving to a new domain can help

And sites with great content were affected because of some technical issues that caused additional duplicate content.

That was my problem in Panda 1.0. Thankfully found that problem and that's now fixed.

I have been also watching two of my competitors, both Pandalized, both with these issues.
Bet if they were to fix this they would come back. Perhaps. Shhhh....

That's one of the most important web pages of the year.

From our own sites, we remarkably found that the threshold that got us affected was different than the one required to get us back. Much more quality was needed to really prove we had bettered our lives.

Suspicion confirmed.

I just pulled over 200 pages (of 700) out of one of our worst hit sites. Waiting.....

From our own sites, we remarkably found that the threshold that got us affected was different than the one required to get us back. Much more quality was needed to really prove we had bettered our lives.

The article seems to indicate that the algorithm tolerates low quality until it gets bad enough to reach a threshold, but after that threshold is passed, from then on the site's reputation is tarnished (in Google's eyes) and there is less toleration.

I think that the fact that you have toleration in the US and we have tolerance in the UK is part of the problem!

@potentialgeek..That's what I've seen too on the sites I've looked at that have started to regain position..Panda was not designed so that webmasters could just "tweak" their way back..

Those who have been using other peoples's content especially are going to have to make their own quality content if they have been hit, not merely remove a few pages and shuffle a few tags as doing so leaves their sites substantially the same as before..

But then personally, I find that removing sites and their owners that use stolen and scraped content is a good thing, not allowing them back unless / until they've produced original content and quality, is even better..if they never come back ..is even better still..:)

It says low quality pages won't affect homepage rankings, is that the case?

i think you mean those that have been using other peoples content AND HAVE BEEN HIT, because a lot of people are sitting and looking at their own content outranking them on other websites.

To me I think that the take-home point is that once you are pandalized you have to over-correct to get back into the good graces.

That's one of the most important web pages of the year.

From our own sites, we remarkably found that the threshold that got us affected was different than the one required to get us back. Much more quality was needed to really prove we had bettered our lives.


Suspicion confirmed.

But..

Could that be because it takes time for the Panda effect to manifest itself? So, in the test the site may have already breached the threshold but by the time Panda took effect they had already reduced the quality further?

The combination of hysteresis and the comment about moving to a new domain is very interesting. Many of us have attempted to correct the issues that we believe got us Pandalized in the first place, but without effect. Presumably it is possible that such sites are in the middle ground, good enough not to be pandalized, but not good enough to recover. For such a site, moving to a new domain might be the answer.

Has anyone tried this? Was it successful?

I think that the fact that you have toleration in the US and we have tolerance in the UK is part of the problem!

Yes, people in the UK definitely have some confusionment about the word toleration...

Thanks a lot for posting this, Lenny.

I don't have that much to say, but the results don't surprise me. The test definitely wasn't scientific and the results are not confirmable.

Actually the article reeks of inaccuracy and tabloidism (in each meaning). Why would you conduct so many tests on such a large scale and produce such a short article with no examples (or even blurred screenshots)? Smells like linkbait to me.

Why would you conduct so many tests on such a large scale and produce such a short article with no examples

Probably has clients that gets the good stuff.

I agree that the test sample is probably too small to be conclusive. Nevertheless, none of the tester's conclusions contradict any of my own observations regarding my Pandalized sites.

a lot of people are sitting and looking at their own content outranking them on other websites

That statement and yesterday's announcement from Google about WMT now reporting cross-domain duplicate content are probably related.

Sorry to say so but I've red this report with skepticism. I don't believe any of stories about recovery with no supporting data.

I believe large sites are more vulnerable to Pandalization and large subdirectories, too. They more closely resemble profiles of auto-generated and other spam-like sites the algo was built to target.

Peter at Search Engine Watch wrote:

"Panda is a Machine Learning algorithm, which means it looks for patterns in a set of given properties. These properties are very likely to be factors like "% duplicate compared to other pages within the same section", "% duplicate compared to other pages in the website", "% duplicate content per sentence compared to other sites."

It makes sense that Panda would cause sitewide demotions if Panda finds signals of very similar content on other parts of your site.

A computer programmer should design a Sitewide Uniqueness Checker.

Peter also made this observation in his comments:

"On some domains just 20 low quality pages in a section with 100 medium/normal quality pages caused all pages to drop in ranking, but in most cases the amount of low quality pages needed to exceed the quality ones by tenfold."

Subdirectories are usually related content, so Panda would naturally scrutinize all subdirectories to see how much is exactly duplicated content and very similar content.

I was going to do some subdirectory pruning. If his data is valid, many Pandalized sites may need to delete/no index as much as 90% of their content.

I was going to do some subdirectory pruning. If his data is valid, many Pandalized sites may need to delete/no index as much as 90% of their content.

I've just done this on a site, 40k pages noindexed everything except 100 merchants with really high quality content.

Seeing if that does anything then I'll start opening the site back up again.

You may want to check the pages you plan to delete for backlinks before removing them. I'm doing that manually with Yahoo Site Explorer. I don't know if there's a Web Tool that checks all backlinks for a single directory or an entire site.

I was concerned about removing lots of pages until testing on a few sites. I removed everything on one site except the home page, and it still ranks quite well. On another site I removed everything except two pages. Rankings came back to page 2 for the main keyword and #2 for the second target keyword.

I think if all the pages left on a site have backlinks from various other sites, that's a fairly good sign to Google, i.e., you're not filling your site with stuff people don't want.

At the same time, I'm checking Landing Pages in Google Analytics for specific directories and pages within those directories. If a page doesn't have backlinks but does get lots of direct hits, and it's not super thin or completely useless, I'll keep it.

If this approach still isn't tight enough after the next Panda Update, I might remove those pages without backlinks.

If that doesn't work, I'll start removing the pages with weak backlinks.

P.S. In checking the backlinks, I noticed that one page on my Pandalized site that survived Panda has about 15 backlinks, including one from a PR7 or 8 site.