Throttling aspects of Panda seem inconsistent and strange

[Whitey]

How can it be that everyone's site drop's an even 50-60% with Panda. Is Google throttling traffic on targeted sites and calling it quality control?

- Google Traffic Quality Throttling [webmasterworld.com...]
- Better Ranking Brings 90% Traffic Drop - is this Traffic Throttling? [webmasterworld.com...]
- Google Traffic Throttling - revisited [webmasterworld.com...]

[tedster]

I've seen reports of anywhere from 8% drops to 90% drops from Panda, but it does seem like a lot of people here have reported 60%. Maybe there's something to it.

The other side is also worth noting. Gains from Panda tended to be around 8% to 18%, from what we can see.

[Whitey]

Maybe there's something to it.

My hunch is that this kind of behaviour ( if it is true ) is Google kind of sending a "warning shot", or we won't completely wipe you out kind of signal. I think there is something more by design than default going on here.

[walkman]

Whitey,
I always took the 60% as casually saying more than half. But
Is it 60% of Google's traffic?
Total search engine traffic?
Total site traffic?

The same applies for minor drops and recoveries. Was it even Panda? Was it just a couple of pages that bring most of the traffic difference? Pages change rank and for some sites that focus on a few keywords that might make the difference.

So unless we can see stats I'd take the numbers thrown with a tiny grain of salt.

[Whitey]

Saw a site yesterday, US based , Panda 1 , approx 50% drop from 3M UV's monthly - all mashable content ( very well done btw and useful, highly linked to , ).

I've seen another 30+ sites with analytics, with different owners in different categories, similar figures but in different roll outs.

I've seen specific trophy keywords hammered on e-commerce sites, other keywords left alone and reports of this on other threads are surfacing. Allysa's recovery comes to mind.

Mathematically, Google attacked the site heirarchy through category duplication and this equals, say, 50/60%, or, it stopped at killing off site traffic completely with an either/or scenario. Since I'm no mathematician this can only be a conceptual contribution to what i perceive.

[numnum]

Put me in the 60% pile. That's the decline in total traffic, not just from Google. But it came in two doses: about 35% in mid-April, the balance late last week.

[outland88]

With all this throttling going on does anybody think Google is trying to control bandwidth or power usage?

[Whitey]

Good point. I spoke with an engineer familiar with SE technology systems. His view was that this update was taking enormous resources. There could be a push on to conserve resources. Not sure if this coincides with the earlier threads rationale though or even if it applies - but a good point all the same.

[tedster]

I'm considering another possibility. Since Caffeine, Google has indexed a huge pile of URLs - so a lot of queries may have scores of results that are all potentially first page quality because they are separated by just the smallest smidgen of score, an amount that is below a statistical significance threshold.

Throttling was with us before Panda and what we see may not have anything to do with Panda. It may well be that Google is trying to refine their relevance algorithm with deeper statistical testing - and the webmaster just feels that as throttling.

[tedster]

Going further, we have been considering what mechanism (or mechanisms) could be creating this effect we call throttling, along with the strange WebmasterTools position data which often seems at odds with the actual SERPs that we see when we make that query.

Bill Slawski just published an article about a patent from last year that might be part of this picture.

Under this system, a query might be sent to a distibutor node which would decide between a number of producer nodes to send that query to...

...this system will attempt to predict how likely it is that relevant information may be found at a particular producer node, and may or may not take into account particular topics or subject matter that may be relevant to the query. Remember, this is a "prediction" before the query is processed...

How Google Might Classify Queries Differently at Different Data Centers [seobythesea.com]

The "predictions" would be part of a machine learning process that evolves over time, but the essence of the system is that sometimes a query doesn't get answered by the same data set.

There's no one or two sentence quote that does justice to the patent that Bill dug up - the article and the patent are worth some close study, because there are some big clues in there to help us understand some of our stranger observations.

[Planet13]

I don't know if this is related to throttling or not, but...

If I compare the traffic graphs for March - April - May of this year to last year, there is a striking similarity in the patterns of traffic, even though overall traffic this year is WAY UP over last year (nearly 50%)

Also, I made a significant change to my site by deleting similar content from a different site and redirecting it to this site. Which accounts for SOME (but not nearly all) of the traffic increase.

Again, I don't know if this points to any sort of throttling, and I understand that there are going to be seasonal fluctuations from year to year, but to me, the spikes and valleys seem to mirror one another a bit too much.

[onebuyone]

There might be other explanation for throttling problem. Could you tell how much % of your visitors come from search engines? Is it somewhere between 70 and 80%? For my sites(medium+ traffic), which show symptoms of having it's organic traffic from SE throttled it's the only common denominator. I also think that it is calculated not for entire domain, but for subsets of pages.

If this is the real reason for throttling then perhaps getting more traffic from other sources could be some kind of solution. I mean visitors, who have nothing to do with Google like i.e. facebook users.

Did anyone try to do that?

[tedster]

@onebyone, it seems to me that only search traffic could be throttled.

[onebuyone]

Our website didn't have any problems with traffic throttling by Google when we were on 50-60%, but problems started once it reached 75% and more when we lost part of our non Google traffic for other reason. Now we can't go below or above 75% no matter what we change on page. Obviously we would like to go back to our old traffic numbers(up to 100k uu per day). :)

There also can be other explanation for this. If large part of your traffic comes from Google then probably there is something wrong with your backlinks, like no one is clicking on them, which may mean that they are paid links, however this is not the case for us, because we didn't ever buy any links.

Seems like the only way to get out of this is by investing a lot of money in getting traffic from outside Google and we are trying to do that now.

[Whitey]

The thought occurred to me that there really is no reason for Google to limit a fall to 40-60%. Why not go the entire way?

Probably because Google didn't want to completely eliminate websites and some consideration was given to providing a chance to siteowners to address the issues, or at some point completely perish.

There may well be some mathematical explanation that is better than this, but i thought it worth tabling this possibility.

[tedster]

A number of sites seem to have taken an 8% to 20% hit when Panda first came around. In some cases, because they weren't supported by ad income, the owners didn't even register that they WERE hit for a while.

[Whitey]

I wonder if Google gave this any commercial consideration from the siteowners viewpoint before roll out.

Would they have been aware that some sites could not possibly afford to write new content in large quantities fast enough, take an income hit and further adjust other aspects - all not knowing whether it would work or not. The better part of me says they did and thats why we don't have total removals of sites, the poorer part of me says they didn't consider or care about this.

All in all, from the feedback I am getting no large sites have recovered. Please correct me if that feedback is wrong.

[DirigoDev]

We log all session starts and page loads to a SQL database using a custom tracking system that we wrote in 2003. The system is super accurate and is not dependent on JS. We built the system back in the GoTo days to track PPC - before GA. The system has evolved over the past 8 year and is far more useful for us than Analytics packages.

I have reporting that trends each hour (where we should be) for each day of the week. I track organic search engine traffic from Google, Yahoo, MSN and Ask separately. We pinned our losses to ~60% just after Panda 2. Our real numbers are below.

We have just over 8,000 organic KW phrases as indicated by GWT. The patterns of traffic these days is pretty consistent. Far more so than several years ago when we saw lots of variation in our daily numbers. When we have a great day early-on the traffic slows in the evening so that we always fall into the predictable pattern.

Here is my quick and dirty comparison using two data sets. Our business is not seasonal - very steady. Panda has cost us about 20%. We are making up part of our loss via other means (we are not supported entirely by organic traffic).

Post PandaG - - - - - - - - Pre PandaG - - - - - - - Loss
Mon - 6/27/2011 - 10,006 - Mon - 4/11/2011 - 15,829 - 58.2%
Sun - 6/26/2011 - 6,981 - Sun - 4/10/2011 - 10,891 - 56.0%
Sat - 6/25/2011 - 6,503 - Sat - 4/9/2011 - 10,170 - 56.4%
Fri - 6/24/2011 - 8,229 - Fri - 4/8/2011 - 13,669 - 66.1%
Thu - 6/23/2011 - 9,389 - Thu - 4/7/2011 - 14,002 - 49.1%
Wed - 6/22/2011 - 9,638 - Wed - 4/6/2011 - 15,644 - 62.3%
Tue - 6/21/2011 - 9,975 - Tue - 4/5/2011 - 17,900 - 79.4%
Mon - 6/20/2011 - 10,104 - Mon - 4/4/2011 - 18,451 - 82.6%
Sun - 6/19/2011 - 6,641 - Sun - 4/3/2011 - 10,856 - 63.5%
Sat - 6/18/2011 - 6,271 - Sat - 4/2/2011 - 9,531 - 52.0%
Fri - 6/17/2011 - 7,924 - Fri - 4/1/2011 - 12,452 - 57.1%
Thu - 6/16/2011 - 9,223 - Thu - 3/31/2011 - 13,990 - 51.7%
Wed - 6/15/2011 - 9,435 - Wed - 3/30/2011 - 15,033 - 59.3%
Tue - 6/14/2011 - 9,331 - Tue - 3/29/2011 - 15,889 - 70.3%
Mon - 6/13/2011 - 9,504 - Mon - 3/28/2011 - 15,979 - 68.1%
Sun - 6/12/2011 - 6,642 - Sun - 3/27/2011 - 11,266 - 69.6%
Sat - 6/11/2011 - 5,891 - Sat - 3/26/2011 - 10,064 - 70.8%
Fri - 6/10/2011 - 7,297 - Fri - 3/25/2011 - 12,929 - 77.2%
Thu - 6/9/2011 - 8,370 - Thu - 3/24/2011 - 15,344 - 83.3%
Wed - 6/8/2011 - 8,947 - Wed - 3/23/2011 - 15,576 - 74.1%
Tue - 6/7/2011 - 9,147 - Tue - 3/22/2011 - 15,521 - 69.7%
Mon - 6/6/2011 - 9,325 - Mon - 3/21/2011 - 15,602 - 67.3%
Sun - 6/5/2011 - 6,542 - Sun - 3/20/2011 - 10,828 - 65.5%
Sat - 6/4/2011 - 5,769 - Sat - 3/19/2011 - 9,599 - 66.4%
Fri - 6/3/2011 - 7,350 - Fri - 3/18/2011 - 12,373 - 68.3%
Thu - 6/2/2011 - 8,588 - Thu - 3/17/2011 - 13,913 - 62.0%
Wed - 6/1/2011 - 8,965 - Wed - 3/16/2011 - 15,364 - 71.4%
Tue - 5/31/2011 - 9,318 - Tue - 3/15/2011 - 22,636 - 142.9%
Mon - 5/30/2011 - 7,235 - Mon - 3/14/2011 - 15,440 - 113.4%
Sun - 5/29/2011 - 5,940 - Sun - 3/13/2011 - 9,544 - 60.7%
Sat - 5/28/2011 - 5,739 - Sat - 3/12/2011 - 9,250 - 61.2%
Fri - 5/27/2011 - 7,106 - Fri - 3/11/2011 - 11,274 - 58.7%
Thu - 5/26/2011 - 8,491 - Thu - 3/10/2011 - 14,617 - 72.1%
Wed - 5/25/2011 - 9,007 - Wed - 3/9/2011 - 14,837 - 64.7%
Tue - 5/24/2011 - 9,473 - Tue - 3/8/2011 - 15,444 - 63.0%
Mon - 5/23/2011 - 10,223 - Mon - 3/7/2011 - 16,111 - 57.6%
Sun - 5/22/2011 - 6,907 - Sun - 3/6/2011 - 11,944 - 72.9%
Sat - 5/21/2011 - 6,059 - Sat - 3/5/2011 - 10,049 - 65.9%
Fri - 5/20/2011 - 7,705 - Fri - 3/4/2011 - 13,015 - 68.9%
Thu - 5/19/2011 - 9,021 - Thu - 3/3/2011 - 14,709 - 63.1%
Wed - 5/18/2011 - 9,932 - Wed - 3/2/2011 - 14,894 - 50.0%
Tue - 5/17/2011 - 10,411 - Tue - 3/1/2011 - 15,421 - 48.1%

[edited by: tedster at 3:30 am (utc) on Jun 28, 2011]
[edit reason] enhanced the formatting [/edit]