Is there a pattern to Panda demotion?

[dibbern2]

I can't see any regularity to the depth of Panda's reranking of my pages. Some went from top 5-10 to somewhere in the twenties, another group slid into the fourties, and a third group got smashed, they are in the 100-300 spots. (There is also a small group with NO demotion, or perhaps one or two places.)

Is this typical of what others are seeing?

[andy_boyd]

Sounds very, very familiar to me.

[dazzlindonna]

Yep. From what I've seen, the best get demoted the worst, and the "thinnest" get demoted the least. Go figure. ymmv

[suggy]

Despite getting obliterated in 2.2 (after signs of improvement in May) one of my pages in particular remains in rude good health for a very big search.

I am beginning to wonder if there is a pattern, it's just not apparent to us because we're not staring at the data. Maybe this is micro not macro....

We are looking for similarities between pages that got panned, sites that got slung out, etc. But, what if google is looking at (nee learning) similarities between pages people like and dislike for specific searches or keywords or searches related to searches? We'd see no pattern at all!

Under these circumstances, there would be no apparent commonality at all at a macro level; only at search phrase/ cluster/ semantic level.

At it's most basic, this filtering process might mean ensuring SERPS do not contain 'informational' pages for searches Google has learned (or suspects) are transactional. (remember those groans when you landed on an article page instead of somewhere you can actually buy the thing!)

Kind of fits with decision tree and machine learning and user feedback suspicions. AND, more importantly, with the complete lack of common factors in what lost out!

[c41lum]

Op thats exactly what I'm seeing. I'm doing some large scale noindex tests, i will let you know about the results.

[Panthro]

Very nice suggy.

[johnhh]

there would be no apparent commonality at all at a macro level; only at search phrase/ cluster/ semantic level

that is possible, for example we have sections where every page (100%) ranks page 1 for the targeted phrases, and some sections where only 10% of the pages rank on page 1 of the SERPS. Yet there is no difference in any factor I can see, and no it doesn't seem to be related to the number of pages in a section, nor to page design or navigation.

Still way down on traffic though !

[PPC_Chris]

The way my ranking dropped, it looks like Google demoted all pages on my entire site by xx%. This wouldn't mean that all pages drop the same amount in ranking or traffic goes down by xx%... let me explain what I mean.

Let's say for each search query Google assigns a value to every ranked page. For simplicity, this value is 100 for the top result. Consider the following examples:

Keyword 1 (highly competitive)
Rank 1: 100
Rank 2: 80
Rank 5: 70
Rank 10: 50
Rank 100: 25

Keyword 2 (long tail term)
Rank 1: 100
Rank 2: 50
Rank 5: 20
Rank 10: 10
Rank 100: 2

Keyword 3 (branded term, "yourdomain.com")
Rank 1: 100
Rank 2: 10
Rank 5: 5
Rank 10: 2
Rank 100: 1


Now consider that Panda imposed a 50% penalty on your domain, you can see that depending on your ranking and the value Google assigns to you and your competitors this could drop your rankings tremendously... while an equally penalized page will not appear to be penalized at all (think of ranking #1 on your branded terms).

This is what I see with my Panda penalty... for brands-specific terms we are still in # 1 position - sitelinks and all with 0% loss in traffic. For our most competitive terms where we compete head to head loads of sites, our rankings took a huge hit with 90%+ loss in traffic. For less popular terms, where we compete with only a handful of sites, we dropped modestly and lost about 50% of traffic.

My theory is that the Panda penalty had the most effect on pages that Google considers to be close in value (in Google's algorithm) to the pages below in the SERPs.

[johnhh]

In our case we actually compete with the same sites for most phrases.

So for very competitive pages, in some sections we "win" in others we "lose". There is some sort of "is the site(s) above/below better ?" going on. The definition of better, IMHO , still leaves a lot to be desired.

I have a horrible feeling that the next roll-out will be announced as a Panda "improvement" but may well be, in fact, a reversal to a previous pre-panda algo, or at least part of it.

[johnhh]

Edit above - having thought about it we are actually into the realms of game theory here - where someone wins at anothers expense

[tedster]

similarities between pages people like and dislike for specific searches or keywords

Excellent - and that could line up with the comment from Google that further iterations of Panda go deeper into the "long tail".

[suggy]

Hi Tedster, but the key issue is that it's related to what's being searched. Hence no universal formula or solution.

We used to rank well with long content in a sea of superficial commerce sites. They just couldn't be bothered to write more; we did because it 'worked'. Not since Panda. And, rewriting twice as long and even more 'useful' sunk those pages further.

I just don't think google wants to show what appear to be articles anymore for these big 'buy' searches. The SERPS where our long content pages lost out certainly back that up.

[tedster]

That's very interesting, suggy. Google has long measured and assigned "query intention" taxonomies - and they adjust them on a regular basis to match up with page-type taxonomies - pure and mixed taxonomy matches are both possible. Any correlation of ta queery taxonomy change with Panda Updates would be quite a new observation.

You saw this phenomenon in exact coincidence with the Panda Update dates?

[tedster]

I've been thinking about this idea for the past few hours, and it really holds water. If Google were trying to find shallow content - something that lies between spam and valuable content - one of the natural signs would be that the URL is already ranking for a decent query phrase. Any URL that isn't ranking for something strong would naturally be tossed out, because it's a waste of computational resources.

Then, as Panda evolves, the definition of "a decent query phrase" would slide further down into more long tail queries.

I find it curious that, according to your observation, results for some query terms now represent less diverse types of content. I'm going to ponder that some more, and also keep an eye out for any patterns like that,

[suggy]

Tedster -- it also means that Google doesn't need to collect lots of user feedback data about individual pages (where sample size would be a problem outside big searches). It doesn't need to know that YOUR page has a high bounce rate or low CTR, only that it looks like one that would.

After reading a paper about how CTR is a pretty good predictor of bounce, I think Google feels it can afford to be fairly bullish/ aggresive about this classification and filtering.

As far as diversity in search phrases goes, it's pretty clear from the suggest tool and their rewriting of search terms, that Google's getting pretty good at 'understanding' what we want (or at least, thinks so!).

[tedster]

It doesn't need to know that YOUR page has a high bounce rate or low CTR, only that it looks like one that would.

Yes - and that's the essential idea behind machine learning. That is the approach Google has preferred for many years, rather than human composed "guesses" at would be good algo factors. The idea is to use a hand-chosen seed set and then isolate the factors that describe just those URLs and not others. Then those factors can be used to build a predictive algorithm. Panda does it for "quality" but "trust" has long been done that way - building out from a seed set.

The early TED interview about Panda 1.0 [wired.com] essentially describes the process of machine learning in this case. I've returned to that interview many times in recent months.

[suggy]

So, basically, we're saying the learning process is nothing new, but maybe customisation of factors for individual searches or clusters of related searches is? Such that searchers get returned page suggestions, not that Google thinks they prefer generally, but that they will prefer for this specific search?

If so, it's a damn clever way of driving article spam out of SERPS it's just not wanted on. And, also, ultimately, you can see it could kill off thin (low value add) affiliates, unless they are able to disguise themselves well enough to fool the algo!

Even as I teeter on the brink of financial ruin due to Panda, I am starting to like it (if it is what I think it is). Because it spells the demise of endless low cost, low value add, slippery (as opposed to sticky) SEO-driven competition in commerce searches!

[tedster]

Right - essentially, the seed set was chosen for a different set of criteria. And then a lot of crunching over a wide set of data. As you probably know, Google collects a lot of data signals that are not currently active in the algorithm. So there is a lot to set the learning program to work on.

And then there's the massive job of building the predictive algorithm - essentially a massive and complex decision tree based on all the isolated factors. It was a breakthrough by Mr Panda in this area that made the new algo possible.

it spells the demise of endless low cost, low value add, slippery (as opposed to sticky) SEO-driven competition in commerce searches!

I think you've got it. Anyone who's been playing that game should be smelling the coffee by now.

[suggy]

I also like that fact that it appears we're no longer required to write lots of nonsense, just to compete with other people's nonsense!

[dibbern2]

Even as I teeter on the brink of financial ruin due to Panda, I am starting to like it

I'm looking at all those ahead of me now --since Panda-- as I ponder one of my best performing pages, and how it has lost place. And I'm thinking, "These are darn good sites, they BELONG up there."

So yes, it hurts sometimes, but I have to admit I like it also.

[MrSavage]

Wow I thought I've heard it all until now. Oh well. I think there are bumpers stickers that relate to this.

But is there a pattern to the demotion? The pattern that I see is that Google now doesn't take your search inquiry as so literal. Put it in quotes and although those should take back some of your personal control over the results, I see that not happening so much. I sense that the control over what I'm looking for a being more Google driven than driven by myself. As in taking more assumption on what I want based on their own data.

So is there a pattern? Sure. If you relied on specific keyword searches for traffic you're going to see less traffic in my opinion. Not unless of course you have a sizable site with a great following and a lot of popularity. Of course to have those variables you can't be a new site that's for sure. In general I'm thinking of subject that are popular. Really niche areas? I'm sure you get some crumbs there because those big sites don't give 2 you know what's about bottom feeder search traffic.

So the pattern really is this. If you're a smaller site and seeing some good traffic still, then my explanation is this. It's only because what you're covering isn't popular and isn't high on the radar. It's not bubbling with activity perhaps. But if that topic or subject get popular, expect to be drowned in a Tsunami of website titans. I'm speaking more from technology perspective subjects. You're only getting traffic now because nobody really is looking for that information. If CNET starts covering that information, your little website is done. Finished. Afterall, it's all about quality, links, great writing and daily content. If you're small you're not competing. That's my take.

[suggy]

Mr Savage

Even as I teeter on the brink of financial ruin due to Panda, I am starting to like it

...was said in the context of no longer having to compete with every two bit websmaster on a cheap hosting plan who buys a keyword 'rich' domain, throws up a Wordpress site, writes a few articles or swaps a few links, in the effort to rank for solely for one keyphrase (albeit a sizeable one). There were so many they were like weeds. All these webmasters rinsing and repeating. The level of mediocrity of the sites was staggering, because all they cared about was SEO and not the user. They were noise, a wasted click, an unwanted annoyance, an obstacle. Good riddance to that game, because under Panda it can't last.

I am looking forward to being able to focus on the real business of being in business: satisfying visitors. Panda should deal with the rest!

BTW, I don't think you're 'getting' Panda if you think it's favouring big over small/ brand over independent/ anti small guy/ etc /etc

[MrSavage]

With respect, deep respect, it's too early for you to say if Panda is about favoring big over small. I hope I'm wrong about that, but what I'm seeing in the areas that matter to me, it's what I see. On the surface, philosophically speaking, it only makes sense to have the big sites on the front of the index. Remember their priorities? Kind of hard to get +1's as an example when you are an upstart or not a big website. People can't find you? You aren't going to get those followers.

If you don't have keyword rich domains that's your problem. Nothing more to say on that. They outrank you? I guess you need to learn more. So I guess you're suggesting that SEO or being smart is tricking the system and must mean the quality of content sucks. I suppose because it's free traffic nobody has a right to complain also. Ok.

I hope this turns out well in the end. I will say sure, in the obscure subjects it's not going to matter. You can still get traffic. Just hope that the CNET's of the world don't start covering those subjects becuase if they do, you better hope you've garnered lots of +1's to compete when that day comes.

[johnhh]

all those ahead of me now --since Panda-- as I ponder one of my best performing pages, and how it has lost place. And I'm thinking, "These are darn good sites

not in my niche - and I am the first to say "thats a great page" to my competitors

so if you are searching for "red widgets information" and get a telephone directory or a "we dont have any info on red widgets" page - well IMHO thats not great pages to find

[walkman]

Good riddance to that game, because under Panda it can't last.

I am looking forward to being able to focus on the real business of being in business: satisfying visitors. Panda should deal with the rest!

BTW, I don't think you're 'getting' Panda if you think it's favouring big over small/ brand over independent/ anti small guy/ etc /etc

They don't have to last, a month is enough, and each month a few do it, pushing you back even more.

True to that but under Panda they are SITEWIDE demotions /drag-downs, only exceptional pages rank and worse pages can easily outrank a very good page from a pandalized site. If Google doesn't think you're a top-whatever-site you may very well be panadalized.

On the third point, if things go like this, you will be disappointed. It's a matter of time before Panda touches every niche and 'high quality sites' will soon add pages and products they will very easily ranks for, simply becuase they are 'high quality sites' according to Google. It never fails and if your hopes are that Google will eventually penalize them...good luck waiting for years.

Edit: and I forgot Google's own products. They are entering virtually every niche so count on your site being a few more paces back.

[suggy]

all those ahead of me now --since Panda-- as I ponder one of my best performing pages, and how it has lost place. And I'm thinking, "These are darn good sites

I don't agree with this. There's some utter junk. But, I suspect Panda will fix that soon enough as it learns.

I really can't be bothered to debate any of the other nonsense, since we've been down the road before and the end has been nigh since 2003 (when I first joined).

[johnhh]

@suggy: April 2005 for us - I joined in the May - all other updates passed us by - in fact we always came out stronger.

I can't be bothered either now - I have so many A/B tests running I have forgotten what they are ...

I have a couple of throws of the dice left to do tomorrow - after that, well, at least our direct traffic will be pleased with the new content we are adding.

and its nothing to do with backlinks ( we have substantial backlinks from national organisations), or advert placements, or article length, or bounce rates, or domain names with keywords in, or css putting content first, or, IMHO, site structure...

basically after nearly 3 months ( hit 12th April) I don't have a clue

[dickbaker]

Mine was one of the first sites hit on 2/24. I lost 40% of Google traffic. Panda took nearly all of the pages that ranked on page one for various phrases and moved them to page two to page five or twelve or worse.

It seemed at the time that the pages that were affected the least were those that had been the strongest pre-Panda for their phrases. For example, one page had been at the #2 spot for years for a particular phrase. That page got dropped to page two for the phrase, but later came back to #9 on page one.

Some pages still stayed in their #1 to #5 positions, but dropped in the last month or so to #8 or #9 or to page two or even to page three or four. These pages should have been the strongest of all, as they're completely original, focused and are the main purpose of my site. They're pages for individual retail stores in my niche, each with descriptions written by me or the store owners.

These would be found by searching for "(insert city or state name here) widget stores".

I'm thinking that I didn't have enough of these stores for particular states, and so all of the store ads for those particular states were hit. There's also the Google localization factor, as Google has its own results for searching for "(insert location here) widget stores".

It was definitely a sitewide demotion, as I have almost nothing left on page one.

Here's where Suggy's theory makes sense: my site had a low bounce rate. It had an average of four to five page views per visitor. Average time on site was one to two minutes. IOW, it kept at least certain visitors engaged.

If it meets the pattern that Suggy is suggesting Google is looking for, then that would make sense. Even though it was a popular site that kept most visitors, it met the "loser site' profile.

[suggy]

These pages should have been the strongest of all, as they're completely original, focused and are the main purpose of my site. They're pages for individual retail stores in my niche, each with descriptions written by me or the store owners.

Surely a classic Panda realignment?

In the "good old days" pre panda, articles/ content could rank for anything -- even when people wanted to shop, not read. Big retailers wouldn't clutter up their slick stores with tonnes of content just to compete, so content sites did disproportionately well -- even when they were just in the way of searchers' real intentions. Not under Panda.

Now post-panda, if searchers want to eat and all you can do is suggest some places to eat, read the menu or offer a review, then you're probably toast. And, it don't matter how good your review, menu-reading or suggestions are.

What can be better for searcher satisfaction than giving users what they want, not an article about it?!

[walkman]

Now post-panda, if searchers want to eat and all you can do is suggest some places to eat, read the menu or offer a review, then you're probably toast. And, it don't matter how good your review, menu-reading or suggestions are.

What can be better for searcher satisfaction than giving users what they want, not an article about it?!

No, that means you cannot (IMO, and should not have) outranked the store. How about #2, #3...#10 and other ranks?