A Simple Theory about Panda

[aristotle]

Here is a simple theory about Panda:

-- Panda evaluates the "quality" of a website mainly by analyzing user-behavior.

-- By user-behavior, I DO NOT MEAN BOUNCE RATE. Instead, I'm referring to more reliable indicators of quality, such as
----- User bookmarks a page as a favorite.
----- User saves a copy of the page on their hard drive.
----- User prints out a copy of the page.
----- User returns to the same page later.

-- Google mainly uses the Chrome browser to collect this data on user behavior. Tens of millions of people now use Chrome as their main browser. This is enough to allow Google to collect statistically meaningful data. And Chrome enables Google to collect data for ANY WEBSITE.

-- In order to evaluate a site statistically, Panda needs a minimum number of user-behavior data points. Thus, the data must be collected over a period of time. As new data is collected, the oldest data can be discarded, but enough must be kept to enable a meaningful evaluation.

-- At the present time, for some sites Panda could still be using data that was collected as far back as last year, because it is still needed for a statistically-meaningful evaluation. This could explain why people who made big changes to Panda-affected sites still haven't seen any major ranking improvements.

[Andem]

>> Bottom line: it was /is based on 'site' not on individual pages.

Walkman is right in my opinion and based on some experiments I've personally conducted over the past several weeks. Namely scraping my own content and republishing it elsewhere.

[tedster]

But look at the history of major algo changes and you'll see that they tend to focus on one issue, or group of related issues, at a time

But Panda is not like any historical algorithm change. It's an immense and brand new undertaking - introducing a measure of "quality" to stand side by side with the measure of "relevance".

From the beginning of Panda, Google engineers gave us a verbal blueprint [webmasterworld.com] of how they used machine intelligence to build the new algorithm. This was a major undertaking that was in development for more than a year before its initial launch, and that will remain Google's focus for another year at least.

If we insist on looking for a "simple" model, then we are doomed to spin our wheels. In fact, I don't think we CAN reverse engineer this thing from the outside - maybe the humans at Google aren't even clear about the fine details. We can take actions based on some of the basic advice (content to ad ratio, for example), but having done that I think I'd focus on exactly what Matt Cutts advised last fall: chase what your visitors want, not the Google algorithm.

And I do know that's cold comfort for those who hope to "recover". But just look at all the resources people have spent trying to "undo" what Panda did to their traffic. We're not going to undo it with any simplistic approach, or technical/mechanical approach. I think it's time to swallow that pill, and work to develop our business models and cultivate a following.

[balibones]

Tedster,

While I agree with what you've said I do think it's important to know more about what Panda was about so that we know what to expect going forward. Most of us have done all of the things Google has asked of us and more, but our traffic still isn't back, or even moving forward enough to give us hope that what we're doing is being noticed by Google (content to ad ratio) being one example you gave.

If this is about user feedback data (to simplify it for the purpose of discussion) then that tells us there is still hope for our domains and that we should work on getting traffic from other channels to get feedback from new visitors into the algorithm faster, if possible. If this is about something else (like ?) and our sites aren't still back then some of us may want to consider starting from scratch with a new domain and new brand. I would usually only do such a thing with a tough-to-shake penalty, but here we are with an algorithm change that acts very much like a penalty.

Again, I agree that we need to swallow the pill and start focusing on a following and making that following happy. But the question for me is "should I do that with the current site/s, or should I just start over?" My answer, based on what I believe to be the case, is to wait for another iteration and see if all of the changes I made have made my post-Panda visitors happy.

[tedster]

It probably depends on what the business model of the existing sites is. If it's something like "get rankings then show them ads" - switch to "build reasons to get return visitors". If it's "get rankings and show them affiliate links" it's "give people a good reason to remember you and shop with you"... and so on.

In many ways, Panda spelled the end of a Google-powered bubble.

[walkman]

This still does not account for one thing: sites that don't really have any social signals (impossible to get excited about) and are way up. Next in line?

- Some sites don't inspire a 'bookmark or comeback,' they can be great but the info is a one time use only.

- There's is more than the first page or first place

- Vocal minority might hate you but are that bad? Sites with 1-2k visits might never reach that point.

Key is removing that site-wide penalty (even if they don't call it that) and let the site rank on its merits. It is as the sites hit hit on 2/24 have a higher burden.

..finally Google has killed productdomains.com, brands rank now. Search for something you wear, use to make toast, make ice cream, cook etc. That's of course tweaked manually.

p.s. Personally I have zero problems doing what Tedster said "give people a good reason to remember you", but going all sitewide credit is bad because you can come back for x when most of my pages are for y.

[internetheaven]

Google mainly uses the Chrome browser to collect this data on user behavior. Tens of millions of people now use Chrome as their main browser. This is enough to allow Google to collect statistically meaningful data. And Chrome enables Google to collect data for ANY WEBSITE.

That's like saying Alexa's information is useful because so many people use their toolbar ... but the majority of users are SEO's aren't they? No-one NORMAL (i.e. the people that I try to get to most of my sites) will consider using Google Chrome. All Google is tracking is how geeks, google fans and SEOs treat the web.

I have Chrome. NO-ONE else I know has Chrome or has even heard of it except other webmasters.

[walkman]

@internetheaven
Bottom line google can do it even if it's not (or seem) fair, and of course they might change it later on, etc. Sucks for those caught in the middle but what can we do.

Something has happened and we can't quite get it. Deserving sites have been pandalized. But very good sites with communities and with repeat visitors have also been slammed. As have sites with very good content. In a thread on another site a Google engineer reviewed a Pandalized site and said he will research presumably using G tools to see why. Answer? I can't tell you why it happened.

Moderator note: Here's a link to that thread [news.ycombinator.com].
The Google search engineer is Ryan Moultano, the same person who
shared early messages about Panda just before it was rolled out

[edited by: tedster at 2:56 pm (utc) on Jun 2, 2011]

[wheel]

I don't know anything about Panda specifically. But as I noted, nothing changed in my niche recently.

Thinking about it, I would expect that user interaction in my niche is pretty low. However if there are any measures, then I'd have expected my site to gain noticeably - and it didn't. Articles in my niche are short and terrible for the most part, and there's not much going on to keep a user engaged. However I've got some long articles I know people read (much longer than any of my competitors) and I've also got some calculators that I know folks spend like 5 minutes playing with. Calculators my competitors don't have. Some percentage of visitors to my site must stay longer than the norm in my industry.

Yet nothing changed. Whatever that means - probably nothing.

[rowtc2]

aristotle, thanks for thread.

In this area I am looking right now. I suspect the signal of low quality can be when user search again the same term after he visit one page.

I do not think Google make algo based on Facebook data or another company, maybe just if they will buy it. Until then i do not think a penalty can come from here or for number of bookmarks.

[londrum]

i suppose, if you think about it, it makes sense that google will run panda once and then leave it a while.

if google decides that your site has half a million "low quality" pages, and you change them all within a week, the chances are that they will still be rubbish, because you cant write half a million decent pages in that amount of time.

so rather than running panda on the whole net every once in a while, maybe google takes the size of your site into your account (and how much of it is "low quality") before they come back and run it again.

eg. if 50% of your site is deemed to be "low quality", but that only amounts to 10 pages, then maybe you could expect panda to be run on your site every week. but if 50% amounts to half a million pages, then maybe google will make you wait a month or two.

[wheel]

Londrum, you're assuming that Google is encouraging you to clean up and do better. They may not care - they may say here's your penalty, we're done here. No reason a website needs to get a second chance.

Either way is possible - I doubt they even gave this consideration. More likely they just did whatever was convenient.

[rowtc2]

Also, even you have the best and useful article on the Web, if you put aggressive ads in top, user enter, click the ad and return to Google to search again his answer. And Panda could hit this.

[aristotle]

If this theory about Panda has some validity, then perhaps there is a simple reason why people who re-worked their Panda-affected sites haven't seen any significant rankings improvement -- the reason could be that Google will need time to collect enough new user-behavior data to reflect the changes, and meanwhile some of the old data is still being used. Thus, even if Panda is re-run, there still might not be enough new data to overcome the old data.

[walkman]

Londrum, you're assuming that Google is encouraging you to clean up and do better. They may not care - they may say here's your penalty, we're done here. No reason a website needs to get a second chance.

Either way is possible - I doubt they even gave this consideration. More likely they just did whatever was convenient.

3 months have passed and I have yet to see one site come out of it (CultofMac and Digital something excluded) and prove it with traffic charts. A few that said they've come back [webmasterworld.com...] , haven't, and traffic stats show it.

So soon, something will become clear and no Google spin doctor can spin it. We'll just believe our lying eyes instead.

[Whoa]

In many ways, Panda spelled the end of a Google-powered bubble.

I would agree with this.

Unfortunately for Google, while popping so many bubbles of small businesses and web publishers, they popped their own bubble in the process: since Panda's debut in late February, their stock has been swirling downward -- from 611 to 520.

That's a nearly 15% drop in the company's valuation. It's a loss of nearly $30 BILLION DOLLARS in just a few short months.

I believe this is attributable, in large part, to a very influential group of people who abandoned Google post-Panda. In one fell swoop, they flipped the switch from Hero to Zero, turning many Google Lovers into Google Haters. Bing search engine usage has boomed post-Panda as well, and that's no coincidence.

It's just a matter of time before Wall Street's financial analysts realize what Panda did to the Google franchise and convert their Buy/Hold ratings to a recommendation to Sell. I think we're still in the early days of the Panda story.

Of course, I still like you, Google. I'm just saying what I've heard from others, you know.

[Tallon]

This still does not account for one thing: sites that don't really have any social signals (impossible to get excited about) and are way up. Next in line?

It's my belief that only a small part of the web has been evaluated by Panda (based on the amount of google traffic or maybe even the amount of high volume keyword traffic/site). The rest of the sites are untouched by Panda (currently) and are floating up to fill the void. It's a theory so do with it as you will.

That's like saying Alexa's information is useful because so many people use their toolbar ... but the majority of users are SEO's aren't they? No-one NORMAL (i.e. the people that I try to get to most of my sites) will consider using Google Chrome. All Google is tracking is how geeks, google fans and SEOs treat the web.

ITA. As far as my core audience, they know chrome on cars but don't have a clue about some web browser. This audience responds much differently to a site like mine than savvy techies and webmasters would and I build according to their taste.

But very good sites with communities and with repeat visitors have also been slammed.

That describes mine, hit by Panda 2.0. I started adding social buttons to the site and realized that the main domain count facebook likes were already in the thousands and were thousands higher than many big brands in the niche (mainstream magazine type sites with millions of visitors a month). These likes were completely user driven from traffic levels and a user base much lower than those brands, and given without codes in place for reminders. Where the site sucks: twitter...my niche isn't there (just fellow webmasters, bloggers promoting each other and brands). I'm not saying Facebook likes are proof positive that a site rocks, but it does give a strong signal IMO that when done organically (like in this case), it's definitely a signal that there's something "right" about this site.

If this theory about Panda has some validity, then perhaps there is a simple reason why people who re-worked their Panda-affected sites haven't seen any significant rankings improvement -- the reason could be that Google will need time to collect enough new user-behavior data to reflect the changes, and meanwhile some of the old data is still being used. Thus, even if Panda is re-run, there still might not be enough new data to overcome the old data.

That's my current feeling.

[walkman]

the reason could be that Google will need time to collect enough new user-behavior data to reflect the changes, and meanwhile some of the old data is still being used. Thus, even if Panda is re-run, there still might not be enough new data to overcome the old data.

I thought so too initially but then sites like askthebuilder.com and Daniweb still have a lot of traffic and more than enough to quantify the changes.

If we don't automatically believe what Google tells us we might even figure it out.

[aristotle]

I thought so too initially but then sites like askthebuilder.com and Daniweb still have a lot of traffic and more than enough to quantify the changes.

That's a valid point. Unfortunately, we don't know what Google considers the minimum amount of user-data needed to get statistically reliable results. Also, we don't know how often Google re-runs Panda. Maybe some sites will fully recover on the next run.

I think it's too soon to reach final conclusions about any Panda theory. Most likely We need to watch what happens for a while longer.

[brinked]

3 months have passed and I have yet to see one site come out of it (CultofMac and Digital something excluded) and prove it with traffic charts. A few that said they've come back [webmasterworld.com...] , haven't, and traffic stats show it.

So soon, something will become clear and no Google spin doctor can spin it. We'll just believe our lying eyes instead.

One of my websites did in fact recover better then pre panda levels. I am starting to believe however it was a different non panda related penalty. I have not seen any other site under my guidence fully recover from panda as of yet. The website which recovered could have also been a benefactor of a slight algo edit.

We are probably looking at a time frame between 3-6 months for a website to be re-evaluated and recover if it indeed change enough to get out of panda. If it was a quick and easy thing to recover from panda, it would kind of defeat the purpose of panda.

[HuskyPup]

Stop theorising and walk away...it's pointless...Google's relevancy is completely broken but, unfortunately, until the user's mindset is changed to a.n.o.t.h.e.r. SE, basically, those of us affected are screwed.

I do not know how to change something that is not wrong...period!

Pointless, I cannot change something that is NOT incorrect.

[walkman]

We are probably looking at a time frame between 3-6 months for a website to be re-evaluated and recover if it indeed change enough to get out of panda. If it was a quick and easy thing to recover from panda, it would kind of defeat the purpose of panda.

For sites with 6 million junk pages maybe, but for sites with now under 1000 pages how hard can it be, if content matters? Not all sites pulled a Mahalo or another content far, many small sites were done in by tags or search pages. Even if their purpose was to be vindictive sc*mb*gs, 3 months is still enough.

Daniweb now has over 2.5 million visitors a month according to quancast and after Panda they still kept 50K US users a day. So even if 5% of them had a Google toolbar or Chrome, Google would still have over 120,000 separate users /signals to judge her. In reality probably 20+% of them have one or the other since they're a tech site. Given that she deleted and noindexed so many of her pages isn't odd that she didn't move up quite a bit just for that? So it's almost certain that Google is lying.

Askthebuilder still gets over 10k a day and he is 'brand' as well. If they don't have data to judge a site that gets 10,000 users a day, something is seriously wrong. These are the two sites that went public and made several changes, I'm sure there are many other popular sites in the same spot.

[aristotle]

walkman -I was thinking That high-traffic sites like those you mentioned could still have some individual pages that don't get many visitors, so a longer period would be needed to collect user-data on those pages. Maybe the overall evaluation has to wait for all the pages to be evaluated individually first.

[walkman]

Maybe the overall evaluation has to wait for all the pages to be evaluated individually first.

Under that theory it means that they penalized these domains for ever. For your theory to be true each page must have about 50-100 visits that Google can track within a month or so, and that's impossible for 99.99%, especially since information also changes on those pages.

The simple explanation is that Google is lying through their teeth and that the sites hit on 2/24 have a special penalty. Considering how sensitive they are by rushing Matt Cutts to put down fires, I am more inclined to believe it. If you believe that Google, is capable of lying or misleading, then you can easily see it.

Remember, Panda has not been a success, contrary to Google's claims. During Panda, Google has lost market share each reporting to Microsoft and Panda has produced a lot of bad press for Google. To make matters worse, a lot of webmasters are now sworn enemies and Panda has probably hurt their earnings too, at least short term. For every friend, like Macys or Walmart, Google has upset probably 10,000 small business owners. Not a good formula, especially when WSJ called the Panda results “more and more polluted” [online.wsj.com...] (true or not they called that.)

Then there's a theory that Panda is not going as planned. When not even Google engineers can tell you why your sites was pandalized, then something must be up.

Worth noting that Jc Penney and Overstock were caught buying /manipulating links and their penalties were removed 60-90days later. Many sites were pandalized for listening to Google's advice on tags /dupe content and nothing yet. Shows how they work.

[tangor]

WSJ called the Panda results “more and more polluted” [online.wsj.com...]

Requires subscription.

Panda is not the problem. Google's growth for investors and shareholders is the problem. As long as their "product" is "us" and they can find ways to ratchet the dollars [webmasterworld.com...] and webmasters allow it...

We live in interesting times.

[walkman]

Google cache is your friend and someone's enemy: [online.wsj.com...]

7:09 pm: Walt: I find my Google search results to be more and more polluted with junk, at least for certain kinds of topics. Is there an opportunity for someone to come in and do to you what you did with Alta Vista.

Schmidt: We've looked a lot at this. An awful lot of people try to, essentially, game the system. He notes recent changes that the company made to affect low-quality content that was gaining ground in the search results.

The latest changes affected about 12 percent of the results, which Schmidt said is an indication of how widespread it felt the problems were.

7:12 pm: Schmidt said that social signals are important, but that there are others, such as location information

Walt: Has Bing done anything that impresses you?

Schmidt: There's a set of questions where Bing has done a better job, in a couple narrow and vertical cases. We have a couple of strategies to address that.

[chrisv1963]

Askthebuilder still gets over 10k a day and he is 'brand' as well. If they don't have data to judge a site that gets 10,000 users a day, something is seriously wrong. These are the two sites that went public and made several changes, I'm sure there are many other popular sites in the same spot.

My main website gets 60k worldwide visitors on an average weekday. Several changes have been made since March (layout, cleaning up code, reducing number of ads, reducing outgoing links).

I did almost anything I could think of and now I started sending out DMCAs like crazy to "clean up" illegal copies of articles and images. I have the feeling that Google is punishing sites because content has been copied. It's like Google can no longer detect the difference between the original and the copy and simply punishes both the owner and the thief because of duplicate content + degrades the trust of the entire site. I still rank pretty well for pages that have never been copied. My rankings dropped dramatically for pages copied by Blogspot, Wordpress, Facebook, ... users.

There's a tiny recovery of traffic, but that could be seasonal. I see the same small recovery for my main compatitor's site.

[suggy]

chrisv1963 -- my experience concurs with yours, but on a site level basis. And, this is why I think this update is about content/ site architecture and NOT user feedback (in addition to the many good reasons mentioned above!)

I had some spammy door way pages on my 200k/ month commerce site. These pages had been written precisely because we could rank top ten for just about anything we cared to write about within our niche.

When Panda hit, I started fixing them and deleting others. I targeted those worst hit (greatest loss in rankings).

A week or so after making changes, pages started to drift back up to being only slightly lower than previously (just outside the real money!), but within top 10.

Then, on some pages I made a mistake in the CMS, which lead to duplicate versions of some improved pages being on the server and linked. Well, they got indexed and, guess what, my improved pages disappeared into the long grass again!

After a week or so, I spotted my mistake and deleted offending duplicates and links to them. Bingo -- today my pages are back again!

The other element at play here is site architecture. Google prefers to rank specific pages over the home page (that happened a while back), but prefers pages well linked within the site. It's like google's looking for your core essence, based on what's most linked.

Why do I think this. Well, if you stop daydreaming about user feedback and look at what's actually ranking, in my niche at least, smaller, tighter sites that were nowhere (due to lack of scale!), got a serious boost. Tiny players on page 1 now! And that's because they have few pages well linked (ie in sitewide nav.) and it's those pages that are doing well. Similarly, for the bigger sites doing well, I noticed most have noindex or a canonical to page 1 on each page in a store category beyond page 1! These sites might be 10,000 pages but google only indexes/understands 10% of that as the content. That has the subsidiary benefit of making all pages indexed more likely to be sufficiently unique!

Seems to me that tight and unique now wins out over gigantic, unless you're Amazon or Ebay.

[mcolom]

My theory about Panda is simple, and not technical: Reputation. You need reputation to rank well. If you're a big brand, you've got reputation already and you're safe. If you're an small website operation, you'll have to work hard to demonstrate your reputation to us (google). And we'll make sure there are not "silver bullets" to get there mechanically.

This I think is the strategy that motivated their panda update. This is why hundreds of thousand of webmasters have gone from loving to hating google. Panda has made them "guilty by default".

But, Google has underestimated the influence of thousands of angry webmasters. It was IT-savyy ppl that made google popular. Have you forgotten that, google? Have you forgotten also that it was a finnish student and a legion of geek followers that demonstrated that Microsoft was vulnerable?.

I'm quite sure bing is gaining marketshare partly because there are thousands of angry webmasters actively recommending bing to anyone that wants to hear them. We should not continue catering to the desires of a company (google) which thinks that "bigger = better". That might be a suicide for small operations.

My innocent, completely legal and honest ecommerce site was incinerated by panda. I'm quite sure my only crime is not being a big brand. In return , I recommend bing, yahoo and even yandex to anyone that asks for my opinion, because I know the more ppl doesnt use google, the more benefits for me. Bing already sends me more traffic than google, and I'm very happy about it.

[johnhh]

suggy : On the basis of what I can see and understand, your post here is probably the most important Panda related post I have read.

[walkman]

suggy,
if you got back with a week it's probably not Panda.
I have seen a huge increase in my non-pandalized sites, as much as 300%. Only one got penalized by the way but it's my $$ site.

My point is that we may be talking apples and oranges and what works on one niche /non-pandalized might not work on others.