Panda - how low can you go?

[whatson]

(maybe we need a new sub-category for panda posts)

So anyway, we hear about people getting hit by panda 1, 2, 2.1 and 2.2, some sites getting worse with each update. So how low does it go?

All of my sites that got hit suffered from panda 2, were untouched in 1. Most of my sites have remained consistent since then, except one that was affected even worse by 2.2, as in went from 2500+ visitors/day to under 500.

However, I had one site "recover" from Panda 2, in the Panda 2.2 update. The traffic went from 2000/day down to 1300/day on Panda 2, but from Panda 2.2. it went back up to 2000/day, perhaps even slightly more visitors than Panda 2. So yes, I do have an official Panda recovery story, with evidence.

HOWEVER, I didn't touch this site, I didn't do anything to it since Panda, in fact I haven't done anything for over a year with the site. So I can't claim that I tweaked my site and did the right thing to get out of it.
On top of that my rankings are still not back in full, my main keyword that I had been #1 for years, is still down at 5, which is what Panda did to it. So it's not a complete recovery.

I believe that some people think they have got out of panda by tweaking their sites, but it was actually Panda that changed. Which just makes it harder to really know what the hell is going on.

[BillyS]

There are sites that come out of Panda, we just did. We were hit with Panda 2.2. On June 17 at 1800 (eastern time), Analytics showed me a clear drop off in traffic of 45 to 55%. It was all Google. Very obvious. Over the coming weeks, this was evident in GWT, where I could see ranking decreases of 3 to 10+ positions on thousands of keyphrases.

On July 12, at nearly the same time, Analytics showed me a very clear increase in traffic of 120 to 130%. Again, this is very obvious because it's a mid sized site that went from 545 visitors to 1,253 (week over week - actual data for the hour 21:00 last night).

Nothing else happened in this timeframe, we have about 1,300 pages in Google's index, and that number was very steady over the last four week.

Again, when you go from 25,000 visitors a day to 11,000 you know something happend.

That trend continues this morning, back to pre 2.2 levels (perhaps doing slightly better, way too early to tell).

[almighty monkey]

Can I ask BillyS - do you keep track of your rankings independent of GWT, and if so, did you see a consistant drop across the board, or was it just a select proportion of keyphrases that were hit by a temporary drop.

Congrats on escaping Panda, btw.

[walkman]

@BillyS
congrats, less than e month and you're out.
What did you change, if you can tell us?
And when did you finish the changes.

I guess Panda run again yesterday

[indyank]

Has anyone who dropped on Feb 24 or May 12 come back? Most of the reported come backs are from people who were hit later.

I also suspect that most of the sites that people are reporting to have recovered are smaller ones (in terms of traffic) like Billy has referred.

billy, Do you have any other that received 25K visitors per day?

[viggen]

@BillS i could swear i heard short lived recoveries just before the new Panda Version came live before here on Webmasterworld, (of course i am not hoping that you get even hit harder after Panda 2.3, just saying...)

[indyank]

Viggen, Are you sure that the next version hasn't yet been rolled out (or settled down)? Are you waiting for July 16?

[followgreg]

I'm seeing a complete rollback, pre-panda type of SERPs, for 3 days now.
Things started to move on Sunday then I noticed that we were back almost exactly to where we were before the first Panda.

To be honest, and don't kill me for this, I started liking Panda last month as our conversion were at the highest ever, even though the traffic was lower.
We were getting more traffic to real good pages but much less overall, which turned out to be more beneficial for us.
Now we are back with more traffic and less conversions.

Was Panda a 5 month long joke?

anyone else sees the same thing?

[indyank]

Honestly, I am not seeing any rollback to pre-panda SERPS. I am sure it just cannot happen.

[BillyS]

Can I ask BillyS - do you keep track of your rankings independent of GWT, and if so, did you see a consistant drop across the board, or was it just a select proportion of keyphrases that were hit by a temporary drop.

I only use GWT for rankings. I could see the rankings drop in GWT across thousands of keyphrases. I could see the decrease in traffic confirmed in Analytics and my web logs. There was red on nearly all phrases, very few exceptions (maybe 1% increased, 99% went down - very dramatic).

What did you change, if you can tell us?
And when did you finish the changes.

Ironically, we had been updating all content starting in August 2010. By that I mean we updated the data (analyses, insights, statistics) where possible. That effort should be done later this month (150 pages to still review). The only other "Panda" penalty change we made was to remove some ads on thin content pages. Out of the 1,300 pages on the site, around 200 would be considered thin (less than 300 words). Most pages average closer to 1,000 words. The other scheduled changes we made were to check for outdated links (outgoing). We also did some clean up based on the GWT suggestions (changes to robots.txt and such).

billy, Do you have any other that received 25K visitors per day?

I run only one site.

@BillS i could swear i heard short lived recoveries just before the new Panda Version came live before here on Webmasterworld, (of course i am not hoping that you get even hit harder after Panda 2.3, just saying...)

If that happens, I will report back here.

One thing I did notice over the last week was our rankings were very slowly returning. For example, right after Panda 2.2 you could see the across the board decrease. If I compared the last two weeks of available data in GWT, I was seeing about 25% green (maybe a 5 - 10% increase in Google traffic). It's tough to say if that was real because the July 4th holiday might have changed the traffic pattern a bit.

[HuskyPup]

@nickreynolds

If I build new sites with the same analytics/adsense code, will they in turn be affected by the Panda "virus" by association?

I've been constructing a new site through all this debacle and my experience is that it has not been affected at all, in fact it's doing extremely well.

The concerning factor for me is that this site is brand new on an example.eu, I also have an example.asia, one of the pages is very, very similar except that one production supply is from Europe the other from Asia.

The new .eu is ranking in G.com at #12 and the 5 year old .asia page at #88...in my professional trade opinion the .asia site is the better page...just how has the .eu been able to gain such good traction so fast?

Is G also targetting tlds?

[indyank]

my pandalized sites were relatively unaffected in some of the previous panda iterations (May and june) but this one seem to be killing them so far.

[walkman]

Monday and yesterday I got the biggest traffic ever for my three non-panda sites. Very, very weird. Obviously have not change a thing in years there.

Do people see the new Panda (2.3?) as done or are the SERPS shifting?

[freejung]

I'm not seeing much movement in my niche. These are definitely not pre-Panda SERPs.

No I meant running the panda program, like the one they run on 2/24.

Presumably about a month, but that's missing the point. If my hypothesis is correct, the Panda algo that was run on 2/24 only evaluated a small fraction of the web. The Panda that was run during the following month also evaluated a small fraction of the web, but it was a different fraction. Only a small percentage of the sites evaluated in Panda 1 were reevaluated in Panda 2, and so on for the subsequent Pandas.

How long would it take to run Panda over the entire web? I'm guessing that would be measured in years, not weeks or months.

That means they have to prioritize what to score. The first priority is to evaluate the sites that rank for super high-traffic keywords. The second is to evaluate sites that rank for medium-high traffic keywords. The third is to evaluate all of the sites that rose up to take the place of the ones that were pushed out, and I'm thinking that for the most part that's still what's happening. Only after all that is done will the original set be evaluated again.

As BillyS points out, there have been exceptions. They would want to test the effects of slight changes in the algo, and see what would happen if some sites are rescored, but that would only require rescoring a small random sample of the original set -- enough to get statistically significant data on the effects of changes, but not enough to bring improved sites and false positives out of Panda en masse.

[walkman]

How long would it take to run Panda over the entire web? I'm guessing that would be measured in years, not weeks or months.

So in essence you are saying we might have an effective 12+ month penalty. Nothing from Google surprises me anymore. After reading last night, it turns out G does have a huge issue with computers, engineers even begging for time to test their ideas. Several other Googlers also chime in. Not sure if I can post the link so search for this sentence:
"What Larry Page really needs to do to return Google to its startup roots"

[freejung]

So in essence you are saying we might have an effective 12+ month penalty.

Maybe. Or maybe they'll never run it over the entire web, most of which doesn't rank for anything of any significance anyway. I expect that sooner or later they'll hit the point of diminishing returns with applying Panda to new parts of the web, and they'll run it over the previously-hit sites again. When is that? I honestly have no idea, but I admit it would seem unreasonable to go for as long as a year with no rescoring.

[Shaddows]

Maybe as the gradient on the learning curve starts to level out, resources devoted to learning can be reassigned to actual implementation.

In addition, as more sites are given an assigned value, further calculations will be incremental and thus less resource hungry.

In effect, Panda will be less of a resource hog over time.

[walkman]

Frankly the more I think and read about this the worse I feel. If they had a sense of decency they'd come forward and tell the truth, or least some of it. When you control such a huge market share you have to be responsible too

[freejung]

What Larry Page really needs to do to return Google to its startup roots

OK, I read that -- most of it is internal culture stuff which isn't strictly relevant, though it makes me sad to hear... but Google's biggest problem has always been one of computing power and scale. Remember back when the basic PR algo was this huge beast that could only run once a month? They improved that with a combination of applying math tricks to reduce the load and scaling up their physical infrastructure, to the point where they can basically run incremental PR calculations in real time. So...

In effect, Panda will be less of a resource hog over time.

I'm sure that's correct. They are very good at finding more efficient ways to solve hard math problems.

Frankly the more I think and read about this the worse I feel.

If I'm right, I feel much better about it. I'd way rather be sitting in a queue waiting to be rescored (at which point I think I'll probably "pass"), than to have been rescored several times already and still deemed to be spam.

[Shatner]

@BillyS Don't count your chickens before they hatch. Give it about a week to see if the recovery lasts before you start counting this as a recovery.

Many people have seen what they thought was a recovery, only to see it completely vanish after 5 - 7 days.

[netmeg]

You may or may not be interested in reading Vanessa Fox' thoughts on Panda in this interview.

[stonetemple.com...]

[BillyS]

@BillyS Don't count your chickens before they hatch. Give it about a week to see if the recovery lasts before you start counting this as a recovery.

I already promised to report back if anything changed. Quite frankly it's comments like these that drove me into retirement from this forum two months ago.

Best of luck to everyone. I will keep that promise and report back in a month (or sooner if the doom and gloom is correct), but back to retirement until then.

[HuskyPup]

I was reading on one of those "other" type of forum boards about a proposed Titan Panda regarding digital licencing from Google for detecting genuine/original content. Has anyone else heard about this or is it BS?

[whatson]

I haven't heard about it, but it might make sense to offer such a tool to webmasters. They clearly have that technology and it is implemented in the Panda "algo". All sites have some level of duplicate content too, so it would also depend on what level of duplicate content is offensive.

[koan]

Thanks netmeg, I'm half way into the article and it's top notch.

[MrSavage]

Ditto on that. Thanks for the link. It's a must read. I'm a nobody but wow there is some points there I need to challenge. Glad I have a site to do that, and no, it's not here.

[BillyS]

@BillyS Don't count your chickens before they hatch. Give it about a week to see if the recovery lasts before you start counting this as a recovery.

I already promised to report back if anything changed. Quite frankly it's comments like these that drove me into retirement from this forum two months ago.

Best of luck to everyone. I will keep that promise and report back in a month (or sooner if the doom and gloom is correct), but back to retirement until then.

Not only did our traffic return on July 13th, but it's continued to climb each week since then. I noticed the other post about Panda 2.4 and it appears we're getting another 15 - 20% bump from that change (English site too). I was thinking about posting a picture of the dramatic drop and sudden increase in traffic. Both the decrease of 50% and the increase of 100% occur over a matter of 2 hours. But those of you that are hit by Panda know what it looks like - a knock out punch.

I'll close this out with two things I do know:

1 - It is possible to get out of the grip of Panda - especially if it looked like a mistake to begin with...
2 - Nearly every guess I've seen posted about Panda did not apply to our site (Bounce Rate, Adsense, Social Popularity, +1... the list goes on.)

The only post that made sense to me was Vanessa's. Before we put any new content up, we look at the "competition." If we can't do better (provide meaningful or unique insights), then we don't bother trying. We're alway thinking about visitors and providing them with something that will truly be helpful.

[Lapizuli]

Is anyone using google analytics seeing this? on the dashboard, for any chosen date, I am seeing an "avg. time on site" which is much lower than what is reported as "avg. time on page" when you drill into content Overview -> view full report.

Yeah, I noticed it this morning and thought I was going crazy. Then I checked and saw that it's retroactive for previous days. I'm still not sure it hasn't always been this way...! (I'm one of those suggestible people who would have been a perfect subject for those memory tests in classrooms - you know the ones where somebody bursts into the classroom, then leaves, and the teacher says "Hey, students, what color beard did that guy have?" And everyone says "black" or "brown" or "red" or whatever, when really the guy didn't actually have a beard? I'm the person who's sure there was a beard.)

If this is Google's way of telling us that average time on site is more important than average time on page (which didn't average in zero-time-on-page bounces, right?) then it kinda contradicts what tedster said the other day about bounce rate not being all that important as long as time on page is good.