Google's Panda Update
Brief description
A so-called Panda update that took place in late February 2011 and then in April - when it was launched in all English-speaking countries - was a big mistake of Google's engineers and architects of their ranking algorithms because this algorithm can be easily fooled, data can be manipulated and thus it can be misused to negatively affect rankings of competitors. Moreover the algorithm relies on human stupidity and laziness This article explains all details and reveals the truth about Panda - whether your site was affected or not, you must think about Panda every day. Because Panda now runs continuously and can demote your website without any warning. Here below is a short excerpt of what you can find on this page, I would like you to follow this page from first to last line - sometimes it might be too personal, however it's "as is" - accept it just like you accepted Panda.
- Important - Introduction
- The Trigger - Basics
- Panda Trigger as Math and Equation
- Behaviour Causing Panda Trigger
- The Panda Process
- New Definition of Duplicate Content
- Bounce Rate is Irrelevant - Sometimes
- The eHow Case
- How to Recover from Panda
- Panda Recognition Application Download
Important
My native language isn't English; It's Slovak instead. However most likely you wouldn't be able to understand Slovak, thus I write in English - not perfect, but at least I'm trying to do my best. If you feel it's raping the language, then am sorry - but you're here for other reason.
The Panda Trigger
Continue reading further after you read the following sentence, please. The answer is simple, the "Panda question" is answered: The Back Button - we can call it time one page before trying other search result or trying related, more specific search query. Now how the hell do I know? I asked myself a simple question; What do people really do on my website? So I created a tool that tracks how long each visitor stays on my website. Don't be fooled, I'm not talking about average length of visit - I'm talking about real time spent on a single page - which can be determined no matter whether a visitor clicked some link or just closed the page after 5, 60, 140 or 1788 seconds! This case is about analysing my largest website aqua-fish.net as it was negatively affected by Panda. The results of this internal analysis were amazing! Most visitors stay 135 seconds and longer. Total bounce rate of that website is roughly 70%, however many people spend over 20 minutes on a single page. Yes, it sounds suspicious - 20+ minutes on a page? - Try to read my article about Guppy and try to read the forum that is under the article too: http://e0.aqua-fish.net/show.php?h=guppy... and don't forget to start your stopwatch. If you're genuinely interested in learning about Guppies, then you will spend those 20 minutes there. Or even more. But how long do people stay on that page exactly? Here are the tables, the results are valid for this period: July 8, 2011, 11:53 pm - July 16, 2011, 4:20 am.
Traffic from Google - Google images traffic excluded
| 0-14 seconds = 221 times (23.36%) | 15-29 seconds = 55 times (5.81%) | 30-44 seconds = 45 times (4.76%) | 45-59 seconds = 59 times (6.24%) | 60-74 seconds = 30 times (3.17%) | 75-89 seconds = 31 times (3.28%) | 90-104 seconds = 29 times (3.07%) | 105-119 seconds = 24 times (2.54%) | 120-134 seconds = 22 times (2.33%) | 135+seconds = 430 times (45.45%) |
Traffic from Bing and Yahoo - images traffic excluded
| 0-14 seconds = 11 times (18.64%) | 15-29 seconds = 5 times (8.47%) | 30-44 seconds = 1 times (1.69%) | 45-59 seconds = 2 times (3.39%) | 60-74 seconds = 2 times (3.39%) | 75-89 seconds = 2 times (3.39%) | 90-104 seconds = 3 times (5.08%) | 105-134 seconds = 2 times (3.39%) | 135+ seconds = 31 times (52.54%) |
The difference and <14 (<29) seconds visits
The difference is clear; Visitors from Bing and Yahoo stay on that page longer. 7% is not much, but it's a difference! 7 out of 100 people, hell this means one thing only; Google doesn't send quality traffic. Why? Before I answer the question, did you notice that Google also sends more visitors who stay on that page shorter than 30 seconds? The difference is only 1%, but it still exists. Irrelevant you might say. OK, here's another article: http://e0.aqua-fish.net/show.php?h=angelfish. The difference here is greater; The difference between visitors who stay on that page 29 seconds and less is 3.01%. Yes, 24% of visitors from Bing and Yahoo leave the site in less than 30 seconds. Google? 27% of visitors don't stay on that page longer than 29 seconds. But how about 135+ seconds? Only 48.43% of visitors who came from Google stay on the page that long (some stay there over 30 minutes, it's a long article). Bing and yahoo? 58.7% of visitors from these sources stay on the page longer than 135 seconds!!! The difference here is over 10%! Bing and Yahoo send more quality visitors. I could continue this way and could compare almost every page that has received enough data to perform statistical analysis. So let's explain when a page is going to be affected by Panda - and when it's OK. This isn't a proof of Panda yet - the next paragraph is.
Panda Trigger as Math and Equation
The Angelfish and Guppy article weren't completely demoted - they still receive reasonable traffic from Google. But here's one page that was affected recently - a little - http://e0.aqua-fish.net/show.php?h=swordtail. It used to rank #1 for swordtail and #2 or #1 for swordtail fish. Then one day it dropped to #3 or #4 position for these search terms (sometimes it shows #2 for swordtail fish). So what's the problem here? It's a long article with forum, pictures and links to other interesting sites. The problem is this: 16.93% of visits leave the page in 14 seconds, 8.99% leave the page in 29 seconds, 6.35% leave the page in 44 seconds and 5.29% leave the page in 59 seconds. All in all 37.56% of visits leave the page in 59 seconds while only 41.8% stay longer than 135 seconds. (interesting, Bing + Yahoo send 45.45% visitors who stay longer than 135 seconds). So there's a high ratio of visitors who leave the site in 59 seconds - this is the trigger. I can't tell you the exact formula, but the more visitors leave your site fast, the greater is chance they're going to click on other site in their search result. Here's the table that might help you understanding the whole problem.
--Bear in mind that the "Panda trigger" isn't only 0 or 1. I believe the scale is wider, but only Google engineers know this range exactly.--
| URL | % of visits leaving in 14 seconds | % of visits leaving in 59 seconds | % of visits staying 135+ seconds | quality links pointing at the URL? | triggered panda? |
| http://e0.aqua-fish.net/show.php?h=corydoraspanda | 3.45% | 6.9% | 75.86% | no | NOT |
| http://e0.aqua-fish.net/show.php?h=corydoras | 10.71% | 26.79% | 53.57% | not much | NOT |
| http://e0.aqua-fish.net/show.php?h=bluegourami | 11.82% | 32.72% | 51.82% | only a few | NOT |
| http://e1.aqua-fish.net/show.php?h=platies | 14.94% | 44.83% | 45.98% | only a few | YES |
| http://e0.aqua-fish.net/show.php?h=discusfish | 18.00% | 30.25% | 58.95% | yes | YES |
| http://e0.aqua-fish.net/show.php?h=aquariumfilter | 18.09% | 38.49% | 39.14% | yes | YES |
| http://e2.aqua-fish.net/show.php?h=pufferfish | 18.97% | 38.35% | 41.24% | yes | YES |
| http://e0.aqua-fish.net/show.php?h=flowerhorn | 22.82% | 44.08% | 43.40% | not much | YES |
| http://e0.aqua-fish.net/show.php?h=oscarfish | 24.05% | 45.13% | 38.65% | yes | YES |
| http://e0.aqua-fish.net/show.php?h=cheapfishtanks | 39.86% | 66.67% | 13.77% | no | YES |
The Result
A document gets some score, say "Panda score". It's clear that pages with less than 15% visits leaving in 14 seconds and with over 50% staying on a page/site for over 135 seconds shouldn't be affected by Panda. So the equation could look like this (sorry I won't write the following in PHP or so, I want everyone to understand the equation) - bear in mind it's not 100% mathematically correct but you get what I mean:
CASE 1)
IF ( page_left{<14s} < 0.15 ) AND IF ( page_left{>135s} > 0.50 ) THEN PANDA_SCORE{URL}-->0
CASE 2)
IF ( page_left{<14s} > 0.15 ) AND IF ( page_left{<59s} < 0.40 ) AND IF ( page_left{>135s} > 0.40 ) THEN PANDA_SCORE{URL}-->(0-1)
CASE 3)
IF ( page_left{<14s} > 0.20 ) AND IF ( page_left{<59s} > 0.30 ) AND IF ( page_left{>135s} < 0.40 ) THEN PANDA_SCORE{URL}-->1
User Behaviour as Trigger?
Honestly, I've never closed a page in 14 seconds after it was fully loaded. Bear in mind that my script is loaded as the last element of the page - so it is quite accurate. Let's analyse this cricital interval. The visitor doesn't like the result? I attached real data above for more than just fun. If you open these pages, you will see a lot of content. Especially the Discus fish article, Puffer fish article and the Aquarium filter article contain plenty of answers - answers asked by visitors sent to my site by Google (pre-panda mostly). So why do people leave these pages that fast?
- People don't get the answer! Seeing that those pages contain thousands of words and all human-created content, it is almost impossible they won't find the answer on my website. In addition I make relevant text crosslinked with relevant articles or other pages. So even if people aren't able to find the answer on a particular page, they can click links and even ask the staff (the form under the articles/forums). I even offer links pointing at external domains just in case people are looking for something else. So in my case I wouldn't think people are given irrelevant result by Google.
- People don't like the layout/design! Hm, I never closed a page because of layout unless it was black/blue text on dark background. In some cases I even turned off styles and viewed a page as black text on white background if I assumed the information I'm looking for is there. We can say that too many ads, too confusing navigation, too much time required for loading a page fall under the condition called People don't like the layout/design.
But How do Visitors Actually Behave?
Google staff disclosed a list of questions that every webmaster who has been affected by Panda should answer honestly in order to understand the reason of penalisation. Be honest, be open to criticism and you'll figure it out. Perhaps they haven't asked searchers these questions:
- Do you know that CTRL+F exists in internet browsers?
- Do you know how to search accurately?
- Do you really know what you're looking for when searching?
- Do you know that terms that consist of 3-5 words usually return results that are closer to your query than a single word query?
- Do you have a masters degree or are you just a pupil or do you even know alphabet?
- Do you search just for fun because there's nothing better to do or are you looking for serious information, products, services?
- Are you lazy (honest answer required)?
- Are you stupid (honest answer required)?
Most people don't know that the combination CTRL+F exists. Whenever I land on a page with plenty of content I press CTRL+F. Many forums are just like that; They contain useful information, however sometimes it's a little difficult to find - but it's there anyway. Searchers are LAZY indeed. Searches are often STUPID. This is why cultofmac.com got penalised - Imagine you just opened http://www.cultofmac.com/tag/ipodwizard. I guarantee you that approximately 70% of visitors WILL NOT CLICK "Read the rest of this post »". I know this because my aqua-fish.net was offering comments as separate pages. They weren't monetised and they were getting some traffic too. However bounce rate reached almost 100% - it was alarming. Until I made the text Click here to read the entire article and other comments more visible by making it look like h1 header. Then the bounce rate was reaching only 30%-50%. As Panda went live, I decided these were "low quality" pages and they're permanently redirected - Now I know most of them weren't causing the problem. Cultofmac.com got whitelisted while thousands of other quality* websites weren't. Quality* is a relative term, how can Google determine quality of an ecommerce site? Imagine there's one shop having low bounce rate and high conversion rates with poor customer service - Panda won't affect it. Now imagine there's a shop causing lazy people to bounce away because of colours, layout or navigation - but with excellent customer support. Who is better in terms of Panda? You know the answer, the second ecommerce site will have to find other ways of receiving traffic.
Panda as a Process
Since Panda runs continuously (please, don't ask me for a proof of this - simply accept it), you never know when your site is going to be punished. OK, with my tool you can determine whether it's going to happen or not. Don't hurry... This also means your site CAN and WILL recover in case you eliminate the number of short visits with further click on other link within search results. It's not that easy though. In case you have a 50-pages website, this is not going to be a hard task. In case it's a website with thousands or millions of pages, it will be harder - but not always. As time goes by, sites that "suit" the Panda trigger get their rankings lowered. New sites gain rankings and show instead of those penalised. But no-one said these new sites are of top quality; Google's algorithm simply doesn't have enough data to tell whether they're good in terms of Panda or not. This is the case of many sites that gained as first and second Panda took place. These sites gained better rankings because their competitors lost them. But not for always. Bear in mind that data have to be collected before rankings can be lowered. This is why many low-quality sites occupy top rankings for relatively non-competitve terms. Google's algorithm hasn't received enough data to punish these websites yet. And this is also why many webmasters whose websites were "rewarded" with first/second Panda rollout took a dive later. They simply got more hits from google, but these hits were low-quality ones (short visits with click/s on other search result immediately), so in the end these "rewarded" websites lost. One thing I know for sure is that by making small improvements you can get better or worse rankings - however it's necessary to wait for the Panda to collect enough information. A page with 10 hits/day (from google text search) will recover much slower than a page with 500 hits/day (again from google text search). A part of the "Panda" process is assigning value to each URL; I'm guessing only, because I haven't collected enough data for this, but I assume that if at least 40%-50% of your website's pages fall under the "low quality" category, your site is likely going to be punished. I am not going to theorize and write conspiracy theories on this; But I feel it's a good assumption. In this case your site is going to lose approximately 50% of traffic from Google. In case 70% of your website's pages are considered crap, you're going to lose even more hits. In general we saw a 40%-50% cut in traffic - most sites, but we also saw websites with 90% cut in traffic - this is why I'm convinced there are several intervals for various penalties. Traffic of some websites was cut by 20% only - approximately 20%. Thus, you don't need to wait for a recrawl - you need to wait until the Panda algorithm collects enough data from visitors!
Let's Re-define the Term "Duplicate Content"!
Up until now we considered duplicate content as copy&paste. Now the term duplicate content has totally different meaning - it also includes the term rewritten content. Here we come to the reason why so many "news" and "celebrity" websites were penalised; If I take an article about Bill Clinton from say CNN and rewrite it, for a limited time period my site can even outrank CNN for some certain search terms (less competitive phrases that consist of 4+ words). However people are not totally retarded and most of them will leave my article after a few seconds because they've seen in on CNN - or they even read that article on CNN.com. In this case Panda works correctly and I must admit it's a great idea. No-one is interested in articles that were rewritten million times. However now imagine that you write unique content, but your website somehow shows bad statistics - many short-visits because you're not a top-notch designer nor you have thousands of pounds (or dollars, euros, ...) to pay for great design. A competitor doesn't consider your site to be low-quality one, so he/she uses your articles, rewrites them - and has a website that doesn't trigger Panda. I have an example below.
An example where Panda failed
As I described above and as you can see in disclosed statistics, visits that last less than 14 or 29 seconds are alarming. I haven't mentioned the following idea before, but imagine this scenario (hopefully someone from Google is READING this!): There is a page devoted to the species called Agamyxis albomaculatus; The page is this: http://www.aqua-fish.net/show.php?h=thornycatfish. Now we have another page that rewrites a part of the content; It's this page: http://en.wikipedia.org/wiki/Agamyxis_albomaculatus. The reference is 100% clear. And now imagine someone performing this search query: http://www.google.com/search?hl=en&q=Agamyxis+albomaculatus&btnG=Search. On July 17th 2011 wikipedia's page was ranked 1st while aqua-fish.net's page was ranked 4th. I don't mind watching wikipedia being no.1, however it's absurd. What's behind this search result? Definitely my website's rankings were demoted due to these short visits - and rewritten content now ranks better. But this is not all! Moreover someone who clicked the first result also visited my website - because it's mentioned as reference. Good, a hit for me. The visitor leaves my website and continues clicking the result no.2, no.3 - all in a reasonable time, not causing Panda to be triggered on Wikipedia's page. Now the searcher comes to the fourth result and clicks it - Oh, it's a website that I visited just a couple of minutes ago, let's leave it and move to another result. Is it clear enough? I have a short visit, my page is not quality enough in terms of Panda. By the way, that search term isn't very popular, but that fish profile on my website received two visits from search engines since I launched monitoring - on July 8th 2011 (until July 17th 2011).
User Behaviour Again
Definitely Panda has reduced tons of low-quality results and if new results are recognised as low quality ones, they'll be removed-will be penalised sooner or later. The smaller website you have and the less trafficked it is, the longer it will take for Panda to punish it. But we're facing user behaviour again and no-one is going to convince me that an average internet searcher is next Einstein who knows how to search, how to use different websites, how to react on different layouts. Most people don't know how to use websites - Google should accept this instead of forcing webmasters to perform ridicolous modifications.
Small Recovery after Modifications
The image below shows a small demotion and small recovery too. Approximately 14 days prior to demotion (marked as point 1) ) I moved AdSense ads just under the h1 heading - so most people saw header, h1 heading and ads. Searchers are lazy to scroll down! The same day (point 1) ) I moved adsense ads down a little - immediately people stayed longer on pages and traffic got better a little. Not much, the difference is only 0-300 visits/day after 2 weeks, however I know that google needs more data. Also there's no guarantee that after partial or full recovery rankings of your website stay "as is" - such a website can be demoted again in case people leave it quickly after clicking the search result link!
The Misunderstandings in User Behaviour
- Ads above the fold - This can be actually beneficial, because if you're trying to force visitors of a particular page to click an ad and if it's possible to reach CTR of 30% (trust me, it's possible if visitors are in buying mood), then you don't have to worry about length of visit. On the other hand if remaining 70% of visits leave that page because it's overly filled with ads, then you have a problem! In this case it's necesary to understand how visitors behave - do they mind or do they read content that can be found under the ad? If only 10% leave that page because of an ad, then it's completely OK! Again on the other hand this cannot be applied site-wide. Each page attracts different visitors - some pages may attract more men between 15-20, some pages are liked by women between 30-40. Different people, different opinions, different behaviour. Read further, there is a tool for measuring behaviour at the bottom of this page! But read first! :)
- Screens, resolutions, operating systems - I'm using a 1600x1200 display since 2005. That time I thought the 1024x768 displays are going to be extinct in 4-5 years. Unfortunately, they aren't and perhaps there are more 1024x768 screens than in 2005. Why? Tablet computers is the answer. While my 1600x1200 screen offers enough space in order to tell whether a page is good enough or not, a 1024x768 screen shows something totally different. You must use eye-catching design in order to keep the visitors on your site longer than 30 or 60 seconds. Again, what is eye-catching design? You must target your website - know who are the visitors and what they want.
- Search terms - Again, plenty of people search just for fun and they're not looking for anything particular. Imagine they're watching TV and something sounds interesting - hurry up, open google.com and search. Google shows wikipedia top for many-many search terms. So our searcher opens article on wikipedia and keeps on reading. 2-3 minutes are gone, let's go to the next search result, perhaps something interesting is there. Obviously there is a great chance that wikipedia outranked document that it took information from! Searcher opens a page and finds information - starts reading, leaves the page because wikipedia's article wasted all willingness to read (why to read if wikipedia contains information that is already on pages that rank worse than wikipedia? remember, wikipedia has no unique content - it's all about rewriting content from other sources and giving them a nofollow link - no value at all!), perhaps next search result shows some image or something really cool. Flickr.com for example. Immediately after Panda was launched, my site was outranked by Flickr for this search term: http://www.google.com/search?q=jack+dempsey+cichlid&hl=en&prmd=ivnso&ei=mN0iTsfvG4inhAfz7LWcAw&start=20&sa=N and also by pbase.com's page. Both are ranked worse now, I couldn't find pbase's page in the top40 on July 17th 2011. This case nicely shows how rankings evovle depending on user behaviour.
So the content doesn't have to be original - it can be copied, it can be rewritten as long as people stay on your website longer than on the original website. This is the reason why many scrapers survived all Panda's. This is why you can find your own content ranked better on websites that stole it from you - without permission, without a link back. While this doesn't happen too often for competitive search terms, nothing is guaranteed for less-competitive terms. Try this search query for example: http://www.google.com/search?hl=en&q=suriname+eartheater&btnG=Search. Immediately after first Panda took place, elite-pets.narod.ru/fish_br329.htm outranked my www.aqua-fish.net/show.php?h=surinameeartheater even it's a partial copy of my page, however I assume elite-pets.narod.ru got punished later - because of poor user experience. My page performs quite well with 33.33% visits staying on the page longer than 135 seconds, I assume elite-pets.narod.ru's page performed much worse as they copied only a part of my page and thus have less text and no "call to action" - which means short visit.
About Bounce Rate
Bounce rate is a percentage of visitors who entered a website, didn't click any internal link and left the site. Bounce rate isn't exit rate; Exit rate is a percentage of people who left a website after visiting a particular page - if page A1 is a checkout page or a "goal" page, and if visitors aren't forced to perform further click on internal link, then they leave the page A1 - but this page doesn't have to be an entrance page, so in an ideal case the page has been used 0 times as entrance page, and all people left the page after, say, checkout - the bounce rate of A1 is 0% and the exit rate of A1 is 100%.
Example
As it has been described above, bounce rate is not your enemy unless you want visitors of your website to perform "call to action". According to Alexa, mongabay.com has a bounce rate of 70%, average time on site culminates around 2 minutes per visit and average pageviews per visit is around 1.5 only. See http://www.alexa.com/siteinfo/mongabay.com for more details. The mentioned data were available on 17th July 2011. On the other hand we have bizrate.com which is a comparison shopping website. According to Alexa they're getting more pageviews per visit (roughly 2.7) and bounce rate is only 42.50% (data valid on July 17th 2011, http://www.alexa.com/siteinfo/bizrate.com). So why was bizrate demoted while mongabay stays "as is"? As you can see bizrate did a comeback although it not 100% recovery yet. In mid-March they managed to reduce bounce rate and increase pageviews per visit. It took almost 2 months until this change got reflected in rankings. Back to the question: If you click on "go to the store", a javascript-resized window will open while the bizrate's site is still open. What will an ordinary visitor do? If he/she is not happy with the destination URL, then the window is closed and again sees bizrate.com. So let's check other result on bizrate... This simple technique lowered their bounce rate by 10%! Currently I noticed there is some pop-up window opened when I visited bizrate - it will reduce bounce rate too because even if 5% of visitors were interested in that pop-up and followed the offered link, it's another 5% cut in bounce rate. And it's also time+ in terms of average length of visit - which after all means smaller chance of returning to google's search results.
Small Screens and Laziness
As I stated already, small screens bring problems. Visits that last only a fraction of time (less than 15 or 30 seconds) are caused by misunderstanding of website, laziness and stupidity or total ignorance. Take a look at http://www.homeconstructionimprovement.com/ which was affected by Panda too. According to Alexa that site gets approximately 40%-50% bounce rate which isn't that bad, average time one site exceeds 5 minutes (I think it's because of videos too) and pageviews per visit is somewhere between 3 and 6 mostly (4.3 according to Alexa). The data were available on July 18th 2011 (http://www.alexa.com/siteinfo/homeconstructionimprovement.com). Based on my knowledge I can judge that many people with small screens will close the site in 30 seconds although the questions they'd like to get answers on are explained on that site. Here are two resized snapshots; The first one shows random article from homeconstructionimprovement.com on a 1024x768 screen, the second shows the same article on a 1600x1200 screen. It's also possible that searchers are in buying mood and once they click an ad, the destination URL doesn't suit their expectations, so immediately use google search again in order to try same or similar search query.
| 1024x768 | 1600x1200 |
 |
 |
It is obvious that if searchers use a 1024x768 screen, many will leave the site immediately, because they will see some links and no answer on their question. However it doesn't mean that moving adsense down would fix the issue! I cannot predict user behaviour and it would be wise to test performance of different articles and pages and then make final decision. Bear in mind that even a 30% bounce rate, if all of these "bouncers" leave your site in less than 15 or 30 seconds and click on another search result, WILL be a trigger for Panda as it means 3 out of 10 searchers weren't happy with the result! 3 out of 10 is quite enough as you could see it above! So theoretically if 3 out of 10 searchers are idiots, the site can be in trouble. Bear in mind that 3/10 doesn't have to trigger a -40% penalty - it can be a -10% or -20% penalty too. I am not a google engineer.
Bounce Rate and Video Sites, Online Stores and Related Types of Websites
If you're operating a type of website that requires the bounce rate to be low, then high bounce rate must be reduced at all costs. I saw traffic statistics of one unaffected online store and the bounce rate of homepage was approximately 31% (traffic source: google text search) and item pages were anywhere from 15% to 80%, however the most popular pages had the bounce rate closer to those 15% than to 80%. If you're operating an online store, then the bounce rate must be low - also the rate of clicking other google's search result in a short time must be very low. How about video sites? Video sites are another specific category - If you're just showing YouTube videos and do nothing else then the Panda trigger has been raised. Which of these options is more likely to occur? Firstly, people stay on your site browsing and watching videos - or - Secondly, people go directly to YouTube and watch videos there? Also, are your visitors interested in watching videos or are they looking for information? This would also explain why YouTube videos rank so high in search results - often kids with no other purpose than watching videos skip other results and just watch. I've been monitoring several YouTube videos that were ranked on 2nd page and slowly made it to the top10. Perhaps some pages from top10 were demoted, perhaps kids wanted to watch. Once again, we don't know practically anything about visitors outside of our websites - and we know only a little about visitors of our own websites.
Misunderstanding of Bounce Rate
What is low bounce rate good for if people aren't satisfied with page, information, document they found and if they still need to perform other search query, eventually click another link within original search result? Let's look at the table below.
| Total time spent on site | Returned to search results? | Bounce rate | Result |
| 5 minutes | no | >90% | Search was successful, searcher found the subject of interest |
| 2 minutes | yes | <30% | Search could be successful, but searcher is also interested in other results |
| 2 minutes | yes | <5% | Search could be successful, but searcher is also interested in other results |
| 2 minutes | no | <1% | Search was successful |
| <30 seconds | yes | <20% | Search wasn't successful, the answer/service/product wasn't found |
| <30 seconds | no | >95% | Search was successful |
Bear in mind, Google can determine "time on page/site" by clicking other search result. But you don't know whether a searcher clicked other search result - with my tool you can predict whether it happened or not, because my tool allows you track real length of visit per page.
- Google CANNOT determine bounce rate.
- You don't have to use Google analytics.
- Google analytics code doesn't have to be used on every single page of your website.
- I was testing panda-effect without google analytics code - pages were demoted and then some of them (modified ones) started to recover - all without google analytics code.
- Google CAN determine length of visit in case searcher clicked another search result or performed similar - more specific - search.
- Google CAN determine whether a searcher was satisfied or not - if clicks on 4-5 results took 10 seconds, then searcher used right mouse button and opened all results in new window/tab. If clicks on 4-5 results took 10 minutes, then searcher visited page 2, page 3, page 4 and then page 5 after leaving previous page (we don't know anything about order of clicks - perhaps first click was the 2nd result, then 4th, then 1st, ... you get it). Bear in mind that searcher could have lost interest in topic (or gained enough information) after leaving 4th or 5th page - But it's possible that google will take it as the most satisfying result!
Misunderstanding the Bounce Rate - #2
Some Panda-affected websites have low bounce rate, relatively high conversions, so where the hell is the problem? Look at the tables below.
| % of visits leaving in 14 seconds (or too fast with clicking other search results) |
Total bounce rate | Good or bad? |
| 50 | 50% | Definitely bad - Panda trigger is likely to be going raised |
| 10 | 80% | Good - No negative Panda effect should be expected |
| 10 | 20% | Good - No negative Panda effect should be expected |
Bounce Rate as Useless Metrics
| Bounce rate of 50% of documents | Bounce rate of remaining 50% of documents | Total bounce rate | Good or bad? |
| 10 | 50 | 30% | Relatively good, however if 25% (0.5*0.5) searchers leave the site in short time, some documents could be demoted. |
| 0 | 100 | 50% | It is necessary to determine whether these "100% bounce rate documents" get long lasting visits. If they are short-visits with searcher clicking other search results, these documents are going to be demoted. |
| 70 | 95 | 82.5% | Alarming rate, however if the main purpose of documents is informational, no need to worry. If this is an online shop, these values mean plenty of problems. |
As you can see, bounce rate is relative. It is necessary to know "real time per visit".
The eHow Case
This part of the document was added as the last addition when everything else was 100% finished, I got the "eHow case" idea when taking a walk with my dogs. How come the first Panda rollout didn't hit eHow - eHow's traffic has improved instead(!!!). So now I tried to open a random article (http://www.ehow.com/info_8720486_african-cichlid-compatibility.html) about fish as this is my passion. I tried to act as ordinary visitor; Thus waited for the page to load, then looked around and decided not to click any ad but read the article instead. Bear in mind I wanted to act as an ordinary visitor who is genuinely interested in content. My feelings? When I reached 1/2 of the article I wanted to close it, but I decided to give it a shot... I finished reading the article in 1 minute and 21 seconds. Obviously someone interested in compatibility of African cichlids would require more specific information, not some useless info - just like the eHow article offers. What would I do if I found that page in search results? I'd hit the back button and would open another page. Bear in mind I wanted to close the page when being at 1/2 of it! Most people would do it this way, however some people would do it my way. 50% of 121 seconds is approximately 60 seconds. If only 30%-40% left the page in 60 seconds and clicked another search result, eHow's score would decrease. Obviously the first Panda was more benevolent and satisfied searcher was redefined later. I'm not stupid, I know some people would click ads, some people would click another article of eHow, but most people wouldn't do a thing except for leaving a page and clicking another search result. Their bounce rate, according to Alexa (http://www.alexa.com/siteinfo/ehow.com) is over 72%, average time on site is roughly 2 minutes and 16 seconds and pageviews per user doesn't exceed 1.7 . Data were available and valid on July 18th 2011. So this is exactly a type of site that should be demoted by Panda. Take a close look at it and do exactly the opposite thing with your website Articles that contain ~ 300 words and are almost or totally useless aren't going to be liked by searchers! Bear in mind I've used my 1600x1200 screen! When I later opened that page on a 1024x768 screen, I was confused (I had to reset my mind) because there were some images (ads), some links (ads), some unrelated links pointing at articles about dogs, and a little of text (real content) along with one picture of cichlid.
How eHow evolved with Panda
- On February 24th 2011 eHow gained greater traffic from google.
- After being criticized google launched Panda 2 after almost 2 months - on April 11th 2011. eHow lost a reasonable amount of traffic with Panda 2!
- There are speculations whether there is Panda 2.1, 2.2 or 3. eHow lost traffic again in mid-June 2011. Panda 3? Panda 2.2? Not sure and it actually doesn't matter at the moment.
How to Recover from Panda
Firstly - Some websites have reported a partial recovery while some reported a full recovery. The full recoveries occured overnight - just like Panda penalisation occurs. Thus, I don't think this is really a recovery - after knowing everything that is mentioned above; Perhaps a false positive or modification in the Panda algorithm. As we can see in the bizrate's example, it's a long process and Google must gather enough data to remove penalisation. Documents will start recovering once Google gathered enough data - it's different for each document! Panda is a top layer of rankings - don't be fooled, every affected website ranks just like before - in old google. Old google still runs, still crawls at usual rate (unless something serious happened) and still does the old tasks. This is why owner of pubcrawler.com reported plenty of "DOS attacks" from google even though the site was penalised - these requests had nothing to do with Panda. Here's a simple to-do in order to recover. It's what I would do... your approach might be different.
The Good Approach
- Setup a testing subdomain or a completely new domain (subdomain is better).
- Tell which documents are most important for you - Select 10 most important ones.
- Place these documents on the testing subdomain and monitor their performance with the tool that you can download at the bottom of this page. The documents should be redirected with 301 redirect - I haven't tried 302. Bear in mind that immediately after google crawler discovers new URL's, they'll show in SERP without being affected by Panda. If you decide to remove them from sub/domain later, use the 301 redirect again - and they will rank just like before redirecting.
- Documents with high "short-visits" rate must be redone unless they're entrance pages (call to action), however in such a case it's necessary to monitor performance of target pages - for this purpose it's wise to set-up goals in your analytics software.
- Panda won't catch you in a day or two as long as your pages are receiving several hundreds or thousands impressions/day, it needs more data - After you recognised poor content work on updates.
- If 5 out of 10 documents perform very well, then don't remove them from the subdomain. Instead, work on the remaining 5 and improve their performance. If your subdomain gets penalised again, your 5 good documents shouldn't sink to the bottom. The bad documents will lose, but should still receive some traffic. Continue working on updates; Change font size, colours, even layout or add something interesting onto the pages - images or anything that will attract attention of visitors. You can test these changes on the "good content" pages too. Perhaps you'll improve them further. Watch performance of good documents closely! The task is clear: To give visitors a REASON to stay on your site longer and NOT TO return to search results!
- The bad documents should recover - if there's no way how to improve their performance, consider rewriting or "trashing" them - or merging with other related documents that perform better.
- Select new 10 documents and move them to the subdomain - if the first subdomain hasn't recovered yet, set-up a new subdomain. Repeat the process until all important documents recover - or partially recover.
- It's also possible to setup a "guaranteed" subdomain which will contain pages that are performing well. So once you determined which documents perform well on your testing subdomain, move them to the "guaranteed" subdomain - all using 301 redirect.
It's Complicated a Little - Any Other Approach?
Yes, the above-mentioned technique is complicated and in case you're dealing with thousands documents, you could spend a year doing this. Or even more. Another approach is making all changes "on-fly" on your website without any subdomain. Modify fonts, layout, move ads or "call to action" buttons, change header image, ... whatever you want. Don't forget to monitor the performance and compare data! Bear in mind that normally 20% of documents are responsible for 80% of traffic. So it doesn't need to be a hard task; If your website contains 1000 pages, you should start focusing on the top200 visited ones only - and work with those top200 which are visited from google text search only. Ignore google images search, this usually sends "less than 15 seconds visits". Hopefully google doesn't consider images traffic in the Panda algorithm - I haven't tested it!
Hold on a Second - Just Another Approach
You must have been wondering why is wikipedia so liked by Google(?). Are they whitelisted too? I don't think so. I kept watching their Aquarium filter article falling down to the 10th position from 4th spot for this query: http://www.google.com/search?hl=en&q=aquarium+filter&btnG=Search. The reason is simple, the article contains information only - thus must be left in less than 15 seconds too often. However why does it still rank 10th? The answer is: links and positive Panda score. This search query shows how much was the article copied - sometimes with links pointing at it, additionally there must be tons of links pointing at the article from other sources, and Wikipedia must be carrying a positive Panda score. Up until now I've been talking about penalisation only, but surely there must be a positive score too (this is my speculation only, I haven't proved it to exist - but I believe it exists). Wikipedia must be carrying a positive score as it's commonly used by students for "copy&paste" homeworks and "projects". Sometimes rewritten, sometimes simply copied A to Z. You visit wikipedia and you don't need to go anywhere else - simply because wikipedia is the largest content farm. Have time? Try following links...
- http://www.conservapedia.com/Examples_of_Bias_in_Wikipedia
- http://www.catholicplanet.org/articles/wikipedia-errors.htm
- http://voices.washingtonpost.com/capitol-briefing/2009/01/kennedy_the_latest_victim_of_w.html
- http://www.timesonline.co.uk/tol/news/uk/article730025.ece
- http://ronconte.wordpress.com/2011/06/19/wikipedia-errors-on-the-immaculate-conception/
- http://www.britannica.com/blogs/2009/05/wikipedia-playing-the-game/
Back to the point! Links, links, links. It's one of the the answers on Panda. But you will really need thousands quality links depending on size of your website. It is quite difficult to gain links naturally when your site isn't very visible, isn't it? So you will have to find really a good link building strategy (directory submissions won't work - you need quality links) and thus some of you will be forced to use XRumer or other similar software. Actually it's possible to make link building completely white-hat, but you must bring something cool, something fresh to the world of internet. It's often difficult to find the right product or application - preferably free and preferably totally unique - in the end you'll gain millions of links.
The Black-hat approach
OK and this paragraph is why I started writing this article . You don't need a link, you don't need to update your site. You can build an application that will send queries to google - search terms that you're aiming at and that are occupied by competitors of your website. The application would have to run on thousands of zombie computers and the queries must look natural. Just like you search. Click result 1, click result 2, click result 3 - and do it relatively fast. Never leave a single page open for more than 30 seconds - and in 1-2 seconds after leaving the page click next result. Continue this way and end up with 6-7 pages opened this way. Last page is what you've been looking for - it suits your requirements. It isn't that hard to simulate a browser with proper knowledge - no matter what programming language you're using (Java, C++ and Delphi should do it - but I am not going to encourage you on this - it is black-hat). However you don't need to be a top-notch coder to do this manipulation! You can do it yourself even it's going to be time consuming. Contact your friends and ask them to do a favour; Use different queries (but related) and tell them how to "search". As soon as your competitors start to sink, you can also click your own website in search result and stay there - you found a good website, so why click further? Occasionally click another result, because it must look natural. Use proxy servers. Always clear cookies. Perform other unrelated searches too, so it looks natural. Don't ask me to develop such an application... I won't do it.
Why am I Telling you how to Fool Google?
Because someone had to do it. What's described on this page would be discovered by someone else sooner or later, actually some people realised it much earlier than I did. What if it was discovered by spammers only? They would manipulate rankings and would benefit from it. Rankings cannot be left to average visitor without basic knowledge how to use websites! The Google company has enough money to pay human "rankers" - why is google aiming at removing some article directories with an algorithm that can be fooled easily? They removed co.cc sites, so why aren't they able to demote poor-quality sites - manually - too? Why doesn't Panda rely on human decision as the final step? Why doesn't Google publish a "guide for searchers" and link to it on their homepage?
Conclusion
Panda is a great idea and in a ideal world it would perfectly filter low-quality websites! I have several low-quality websites and I know they don't deserve to be ranked. However this world isn't ideal and searchers are far from perfect. So I hate Panda as much as I like it. Thanks to this update I made many visitors of aqua-fish.net happier; By increasing font size, by making text really black, by updating layout of articles and fish/plants profiles and aquarium biotopes pages, by moving advertisement lower, by understanding visitors' needs better! I really noticed more questions being posted after doing these simple changes - even more comments and questions than prior to Panda with 10000 visits/day! However Panda still needs some serious upgrades. And it would be nice to see google educating searchers - at least if they tried to educate them.
P.S. I took this snapshot today - July 21th 2011 - (had to stop using Piwik because Piwik and PHP eAccelerator don't get along well on my server) - note that it was 11:40AM when I took it, so the 12th hour doesn't show full traffic! The language in snapshot is Slovak, but you're getting the point - it's a comparison of July 21th 2011 vs. April 28th 2011 (both days are Thursday).
Bear in mind
- US, UK, Canada, France, Australia and every other country - Visitors aren't same. They don't like same things!
- If your site is moving down in rankings, it isn't necessarily because people don't like your site - What if they skip your site because the title in SERP's doesn't look attractive? It's a known fact that CTR matters!
- Why do ebay, amazon and other big companies show high for many search terms? Positive panda score is the likely answer. If 95% of your site performs well and rest 5% is shit, count on that these 5% will perform well in rankings!
- Average is useless information. If 3 people left your website in 15 seconds and if 1 searcher stayed there 20 minutes and did 40 clicks on other internal pages, it's 5 minutes per visit and 10 pageviews per visit in average. But 75% of searchers (3/4) didn't like the result!
- Stupidity and laziness are now one of the most important factors in rankings! The more stupid and lazy people visit your website, the greater chance it'll be demoted.
Download the Panda-Recognition Application
Unfortunately I have injured my arm on July 22th 2011 and CANNOT write a more detailed user and installation guide with snapshots! I'll do it as soon as my arm gets better! This is just a beta version! The WordPress version will be available very soon too! The tracking application is NOT freeware - you cannot redistribute it and you cannot sell it! But you can use it at free of charge!
Installation Guide - User Manual
You have to know what you're doing - Knowledge of PHP is almost crucial for installation. If you're not sure, use your coder or I too can setup tracking on your website for some small fee. Contact me in such a case.
Installing the tracking
- Download the tar.gz package and unpack it. You will have 5 core files there (+ calendar files - do NOT remove them!!!): analyzer.php, finish.php, sql_config_t.php, tracking.php, tracking_end.php . Other files inclduing license.txt and readme.txt are part of calendar that is used in the admin panel of tracking!
- Open sql_config_t.php in your favourite text editor and specify server, user, password and database - it's access to MySQL database where data will be stored. Also configure following variables: $cookie_server (if you owned my aqua-fish.net, then you'd type aqua-fish.net there). No www prefix!
- Also fill the $allowed_ref[] array. Simply follow the example and type yourdomain.tld, www.yourdomain.tld and perhaps subdomains in case your site contains them. yourdomain.tld is an actual domain name where tld can be com, net, org, info, and so on.
- Upload the files (all unpacked files) to your server; If your website's files are stored under /var/www/websitename/public_html/, then simply copy the 5 files along with all calendar files to that directory.
- Open analyzer.php in your browser; You CAN rename the analyzer.php file and you CAN password-protect it - it's explained in the file. So open something like this URL in your browser: http://www.yourdomain.tld/analyzer.php
- Now see the show/hide settings console link - If you can see red text saying something like Path for setting $TOP_path: /home/content/website/public_html/website then the path is now very important. If you cannot see the red text, simply click the link called "show/hide settings console" - you will get the path.
- Open finish.php in your favourite text editor and specify the path. It is 4th line and looks like $TOP_path = "/home/content/original_path/public_html"; //this is the place for path!!! Now simply replace it with /home/content/website/public_html/website that you can see when viewing analyzer.php.
- Now open tracking.php and do the same with path - it's 4th line too.
- tracking.php must be called by every single page that you're monitoring. If your website executes some common file when sending headers or cookies, it's wise to call tracking.php from that file or piece of code. If you don't know PHP, you should leave this to someone who can handle it. If you know PHP, you know what I'm talking about and simply add the following line anywhere before outputs are sent to a visitor (because tracking.php handles setting up cookies) include "tracking.php"; will simply do it.
- Just before the </body:> the actual tracking code has to be inserted. The line will look like this: include $TOP_path."/tracking_end.php";
- Disallow finish.php in robots.txt!
- All done!
Using the tracking application
Firstly, how does it work? - Tracking refreshes every 15 seconds on a visitor's computer. A small 1x1 px image (finish.php) is refreshed (thus loaded from your server) and this is actually the core that monitors length of visit. If a visitor leaves your website in 9, 12 or 14 seconds after opening it the image won't refresh and the visit is recorded as a 0 second visit - then considered as a 0-14s visit. If a visitor stays on your website 10 minutes browsing various URL's, total time of visit is recorded too and single pages are recorded too. At the time you installed this tracking there are no data for further analysis. Try to visit your site via search term from google. When you open analyzer.php in your browser click "show/hide single URL analysis" - you should see at least one URL in the list of URL's. Now select "All (including direct traffic and traffic from other sources)" and also "Show visitor data?" and submit the form. You'll get results at the bottom of the page once it reloads. Also use "show/hide average length of visit analysis" to find out more about length of traffic depending on source. This is just first BETA version! If something doesn't work, contact me and I'll take a look! I am also working on WordPress plugin and will launch it by August 31th 2011 (this has been postponed due to several reasons).
22 July 2011
©Jan Hvizdak 2009 - 2012, all rights reserved
Document last modified on Sat Aug 20 2:38:15 PDT 2011.
