| This 193 message thread spans 7 pages: 193 (  2 3 4 5 6 7 ) > > || |
|Is Panda all about Exit Rate?|
Commercially, what I am about to do may not be the most sensible thing, but I feel itís right and I want to share what I have discovered about Panda. It may help you understand more about quality and how to escape Panda.
Please note, I have not escaped from Panda yet Ė I came to these conclusions on June 22nd 2012 and began addressing my issues based on a new understanding of how Panda works. This theory could develop and I could end up with egg on my face massively, but it makes more sense than anything Iíve ever read anywhere before. Here goes............(apologies for the long post in advance).
Since Panda hit my ecommerce site in April 2011 Iíve been trying to improve the quality of my site using Amit Singhalís guidelines as a basis but completely without success.
I always imagined that Panda was a magical formula Google concocted using their human guinea pigs when they sat them down and asked all those questions relating to quality, and that ĎPandaí is Google crawling your site looking for the signs of low quality. Itís not.
Last week my attention was drawn to a statistic in Google Analytics that for some reason Iíd never noticed before Ė Exit Rate. Thatís when it dawned on me Ė Panda is all about user metrics, they canít Ďseeí your site, they donít crawl it using a magical formula, they collect signals given off by humans as they use your site to tell them where the bad quality is. If you have too much of it, they demote the rankings of the pages with bad content and any pages that link closely to those pages to protect Googlers from hitting your bad content (what we know as Panda).
I compared pages on my site with very high exit rates to those with very low exit rates and immediately it struck me how much better the pages with low exit rates were. It also struck me how many different reasons there were for the high exit rate pages being worse (in many cases it was just a bad product that rarely or never sold, or the price was too high, the description was poor, the image as poor, etc.). The low exit rate pages were our top sellers, good products, good descriptions, nothing bad to say about the product or the content or presentation of the page.
Then I realised this is where Google started. They wondered about Exit Rate, sat people down, asked them to compare web pages, asked them why they liked or didnít like a page, and found that Exit Rate correlated with human feedback. Itís obvious really Ė people leave your site because theyíve either done what they came there to do or something put them off. This is the ultimate test of quality.
Google doesnít need to Ďseeí your pages, it just looks at where people leave your site, maybe what they did before leaving your site (how long they were there, how many pages visited, etc.) and if your site has a high proportion of pages with a high exit rate, your users probably donít like the quality of those pages.
Of course, people have to leave your site at some point, and that may be because theyíve found what they want, so there has to be an allowance for that. And there may be a different model for different types of sites. But I found, when I looked at my high exit rate pages, in most cases it was obvious why people didnít like them. In the case of our product pages, the high exit rate pages were generally non-sellers, cluttering up the site and, as I now realise, turning off customers.
To try to disprove my theory I started reading back through Amit Singhals guidelines and it all made sense (as I knew it would one day!). I also looked back at various discussions about Panda, things that people did to recover from Panda, and it explained everything.
Itís beautifully simple and it deals with a huge range of Googleís problems in one hit. Users naturally react differently to webspam, duplicate content, scraped content (or original content if itís been scraped), brands. Sites with a high proportion of high exit rate pages tell Google all they need to know about the quality of your site, from a web userís perspective (which takes into account an unfathomable range of considerations that even Google have struggled to document - what Google needed to say is what Iím saying now, look at your exit rates!).
This theory explains so many of the things weíve all noticed about Panda, how it works, itsí effects on Googleís results and our sites. Hereís a few......
Why canít they run Panda more regularly?
They need a monthís worth of user metrics to be able to make a judgement about your site.
How did I recover from Panda without changing anything?
Scrapers can hurt your exit rate. So can competitors. If you have content people have seen elsewhere it affects their perception of your site. If Google got rid of your scrapers, your user metrics would improve without doing anything. The user metrics of your site are affected by whatís happening on other sites so even a new competitor doing something similar to you can affect your user metrics.
Why are brands dominating the results?
Itís not brands that are dominating the results, itís websites people like and trust that are dominating. Not every brand will always be loved and trusted, and their user metrics will reflect those changes. But generally people trust what they know so sites people like (letís create a new term to replace brands Ė SPL, sites people like) can have bad pages but people wonít leave their site just because of it, so their user metrics are better. You could set up an identical site with your name at the top, the Ďqualityí of the content would be identical, but the user metrics would be much worse.
Why did Google suggest merging pages?
Iím guessing user metrics show that users donít like seeing several similar pages on your site, in the same way they donít like seeing similar content on numerous sites.
Do images or photos help improve quality?
Not necessarily. Every page on the web can produce a different response from users. The only way to know is to experiment, check your exit rates, repeat until exit rate is low.
Should I add more content to my site?
See ďDo images or photos help improve qualityĒ.
Can I escape Panda by improving my brand signals?
If you mean getting backlinks with your website name in them, no. That does not make you a brand. Google doesnít actually care if youíre a brand or not, it just knows that people react better to SPLís (sites people like). User metrics prove it.
Can I escape Panda by getting better quality links?
No. Users canít see your links, links make no difference to how users perceive your site alongside the rest of the web. In my experience, while demoted by Panda, links wonít get you anywhere. Once you get out of Panda though......well, hold onto your hat.
Why does moving content to subdomains work for some people?
If you correctly identify, fix or remove bad content from your site (using Exit Rate as a guide) you will be left with only good content. It doesnít matter how you do it, what matters is that you get your exit rates down.
Will no-indexing or blocking robots from pages of bad content help?
No. If a user can see it, Google has user metrics on it. This is not about googlebot, itís about your users.
I could go on, but you get the point. User metrics tell Google everything they need to know about how real human beings perceive your site in context with other sites they may have come across. It really is as simple and as complex as that.
What I donít know is where the threshold is and Iím guessing it may be different for different types of sites (information versus ecommerce for example) and there will be other factors combined with it but, put simply, I think Exit Rate is the place to start looking if you want to find and fix your bad content issues. It really opened my eyes. (Note: I think bounce rate impacts exit rate, I might be wrong on that Ė removing bounce rate from the Exit Rate calculation may give you a truer reflection on how people react with your content as they move around it).
I suspect that some sites may not able to get below it if they are basically set up with spammy intent (you see how effective this simple method is) but for many of us, understanding that exit rate tells you where you bad content may be, could be the answer to your prayers.
I hope Iím right, or at least on the right tracks. If I am, itís time to end the Panda woe and improve our website KNOWING what quality content really is. Maybe a discussion here will help test the theory and perhaps help us gain an even greater understanding.
I hope this helps you and me escape Panda, I truly do.
There is a difference between "exit rate" and "bounce rate" that is important for discussing this idea. Key points:
Exit Rate is the percentage of page views that were the last page in the session, no matter how many total pages were viewed in that particular session.
Bounce rate is the percentage of page views where it's the start page and the visitor then leaves without viewing any other page.
Here are some examples from Google that should help clarify:
|Now let's extend this example to explore the Exit rate and Bounce rate |
metrics for a series of single-session days on your site.
Monday: Page B > Page A > Page C
Tuesday: Page B > Exit
Wednesday: Page A > Page C > Page B
Thursday: Page C > Exit
Friday: Page B > Page C > Page A
The % Exit and Bounce rate calculations are:
Page A: 33% (only 3 of 5 sessions included Page A)
Page B: 50% (only 4 of 5 sessions included Page B)
Page C: 50% (only 4 of 5 sessions included Page C)
Page A: 0% (no sessions began with Page A, so it has no bounce rate)
Page B: 33% (bounce rate is higher than exit rate, because 3 sessions started with Page B, with one leading to a bounce)
Page C: 100% (one session started with Page C, and it lead to a bounce)
My own feeling is that Exit Rate (measured from user data, not Google Analytics) is very likely one of the metrics in the Panda formula - and in some cases it might be a big one. It's certainly worthwhile to understand pages with a high exit rate, as well as the most common Exit Page on your site, whether their "rate" is high or not.
[edited by: tedster at 12:54 pm (utc) on Jun 25, 2012]
It cant explain everything ( not that I think you re trying to say it can :) because it falls over when related to single page EMDs with a short content page ( under 200 words ) that have no "exit" available other than a mail link, and still rank in the top 3 on very competitive terms ( 500 million, 1 or 2 billion + results ) and that do not have analytics on..and very few inbounds
No adsense nor other ads..
Didn't get hit by any panda iterations ..nor penguin..
So how would they measure "exit rates" ? ..less than 20% chrome users, so insufficient data from that..and even if they could measure them in those cases ..they would appear shocking and would tend thus to penalise the page instead they leave them alone..
Google has a lot of user metrics..but not , IMO , enough to be using them to rate small sites that have no adsense and whose visitors are not being tracked..
All they could measure is "bounce" ( return to serp ) and unless the visitors are all reading a few hundred words following the sentences on the screen with their fingers and mouthing the words slowly, the "bounce" would be quite rapid..or they click to "mail" or click out to another page, but G are not looking over all their shoulders, they can't track a non chrome user moving page to page who doesn't arrive on successive pages with G's "beacons", be they adsense, double click , youtube , whatever..
In other words automating panda via exit or bounce rates may explain some cases..but not all..G has blind spots, where the automation gets no data to use..
Leosghost, you're right, statistics are meaningless in low volumes. Google can't Panda slap any site unless it has enough statistical data.
That's how small sites avoid Panda - not enough traffic to be certain how people are reacting to them.
[edited by: claaarky at 12:58 pm (utc) on Jun 25, 2012]
Kind of doesn't make sense on this end. I would not have a site left if I took all the high exit rate pages off.
400 page site plus or minus. Information, been round since 2002. Hit with panda I.
14 pages with an exit rate over 80%. Ten average over three minutes viewing time. The other four average about one minute ten seconds.
Something to think about as I've never really looked at things from this angle. I wonder why they include pages that have been 404'd? I show tons of those at 100% exit. Same with "translate". Maybe something is messed up internally here?
I don't think it is "not enough traffic to be certain how people are reacting to them" ( in fact I know traffic wise that isn't the case because some are high traffic ) thus "Google can't Panda slap any site unless it has enough statistical data" doesn't apply ..
But I think that smaller sites are harder for them to get either their physical or algorithmical heads around from the point of view of assigning "average scores" and "weightings", I think smaller sites may well have to be appalling ( not that any of mine are ;-) to score badly, whereas with bigger sites, 10% of "meh" pages may be enough to ding them..
IME , smaller seems to be considered by G ( and MS etc ) to be more "focused" and "closer to the query" or "more relevant" than medium or big spreading sites..of course Ehow is an exception, but I really do think they apply a different set of rules to it than to others..
I am also convinced and always have been that EMDs are considered more "relevant" if all other factors are equal ..
Well this does not seem to fit my Panda-hit site. I checked several keywords which i am still on first page. And most of them have nearly 100% exit rate.
I think if you provide the right info(what the user was really looking for) Then after he/she finds it leaves immidiately, so that means the page is awesome, and there isn't any need for checking other links on the site. I do understand claaarky's point but i think every site is a different story. So while at some cases a high exit rate means that a page is bad, in other cases(for example in my case) a high exit rate is the best sign, that the page is 100% optimized for the google visitors.
I'm not sure the idea of analysing user metrics is a new theory. This is clearly a part of the algo that has been around for a long time.
This is, though, only one part of the puzzle.
But I do agree, go through these pages and if something is clearly making visitors press the back button - fix or improve!
Google Analytics is not used, im almost 100% sure. Bounce rate is such a thing.
1.google search comes with results of sites, user clicks but it was not what he was looking for, thats not the fault of the site, but googles ranking.
2. I know a lot of sites where the user find that page what he is looking for google search and leaves again, after a few sec, he then had a look at the picture/downloaded something/found the text/ or what ever.
3. I can not imagine that google want us to make it HARDER for the user to find what he is looking for on a site, be cause thats what I would do if bounce rate was/is a factor, what I do is giving them the info at once with as little text as possible so he dont waist anytime, thats also how I would like it.
Thanks for your post.
To my best understanding, low quality page is always the combination of high bounce rate + high exit rate WITH low time on page (seconds).
High BR and High exit rate don't necessarily mean that the user hasn't gotten what he was looking for. If he stayed minutes on that page, he probably received an answer.
But if that page has these 3 metrics attached to it, it screwed and must be improved or nuked.
Correct me if I am wrong.
I agree with you that it is not about quantity anymore. Google is looking for tight quality, not mountains of crap.
I'm struggling to get my head round what a few of you are saying here about the idea of 100% exit rate being normal or okay.
Do you have no visitors navigating around your site? Do they all come via external sources, see one page and leave again?
I think that a high average time on page can flatten the affects of high exit or bounce rates.
I have a Pandalized site. From Panda 1.
I had a partial recovery two months ago.
But I have 4 KW that ranked #1 all the time (circa 2005). They both have 90%+ exit and bounce rates, but 4.5 minutes time on page averages.
My most popular pages have very high exit rates BE CAUSE they got what they looked for, time on page 5-10sec be cause its image related, another site it could be a download,.... not all sites have or need a lot of text.
Yeah, i agree with zeus. For example a question answer site:
If the site is well optimized for google, then the googler will find your page with the question he/she was looking for, and after reading the answer he/she will leave. If that page does not give good info then he/she will start searching for a better question on that page:
So if he/she leaves means: good page
if he /she start to search for other similar questions: bad page...
There are other profile sites, like zeus mentioned, which can be looked good besides a high exit rate.
Okay, this seems to suggest the metrics used to judge pages of an information site are different. As Pjman says, time on site may be a more significant factor. Perhaps information sites hit by Panda need to be looking for pages where you'd expect people to stay a lot longer if they liked the page.
What I've found with my site (ecommerce) is that high exit rate and poor content seem to go hand in hand. I know that Panda hits sites with a high proportion of low quality content and my site was hit hardest in the sections where most of my high exit rate pages are.
[edited by: claaarky at 3:12 pm (utc) on Jun 25, 2012]
After spending twenty minutes reading your long-@$& post... And all the comments I'm going to exit this page... I'd say that is an indication of a satisfied user.
Clark, the post has brought out the importance of looking at my user metrics closer and "fixing" the pages that appear to have poor metrics... Be it exit/bounce/time on page whatever. Thank you for the insights.
|I'm not sure the idea of analysing user metrics is a new theory. |
It's not - there's no question that both Google and Bing are doing this. What those metrics are is the big question.
Whether Exit Rate (taken from browser data) is actually in use or not, it is an informative reflection of how your users see your various pages - especially if you bring real intelligence to your analysis and don't try to do it mechanically.
It's probably a factor, but it's not all that Panda's about. I can say that with 100% certainty.
I have a site that I setup a few years back. It was just a test site that I probably put 10 total hours into, but it basically just pulled in a feed through a third party API. There is no original content on the site, and there are currently over 1 million indexed pages.
The uniqueness of your content clearly matters.
|I think that a high average time on page can flatten the affects of high exit or bounce rates. |
So leave a bunch of browsers on an open page overnight to push up the numbers? Or make each link open into a new window?
I can see the theory being beguiling for e-commerce sites when the visitor is a shopper, but that is not the whole sum of internet activity.
It's still good to look at this statistic to weed out failing pages, but (as already mentioned above) many types of page are a success for the same analytics that make other types of page a failure.
|I'm struggling to get my head round what a few of you are saying here about the idea of 100% exit rate being normal or okay. |
Example: I want to find out 14-day weather forecast for Timbuktu. I search "timbuktu weather 14 day". I click the first result, I see the graph showing me the temperature and rain (if any). It takes me no more than 10 seconds to digest the info. I exit.
|but (as already mentioned above) many types of page are a success for the same analytics that make other types of page a failure. |
Isn't this the key that doesn't come up enough in these analysis threads? Aren't these SE's of the future supposed to be learning how to classify documents? Maybe I'm wrong but I understand that to mean, at least partially, that an SE will learn to find out what category a particular document fits into and apply the rules of that category to how it is ranked, when and how it is displayed.
I think general rules for ranking apply across types of documents, but I believe the application of rules is more granular. E-commerce sites are treated differently than professional services, large forums from small government sites, etc, etc.
Thanks for all the thoughts so far.
The consensus at this stage seems to be that:-
a) Google is collecting very sophisticated user metrics probably via the browser and feeds that into Panda to determine whether your site has a big enough problem with bad content that your site should be demoted to protect users from coming into contact with it.
b) Exit Rate may be one of the many metrics Panda looks at to identify low quality content, but the importance of it may depend on the type of site you have.
c) Time on page may be important, but again it depends on the type of site.
d) Any stats you use to help identify pages with bad content on your site should be viewed keeping in mind the type of page (if your page is an article that takes 3 minutes to read you would expect the average time on page to be around that - if there's a second page you would expect the exit rate to be low, and so on). If your stats show the average user spent 5 seconds on the page and then left, there could be a quality issue with that page. Look at it, work out why, fix it or remove it, repeat until all content is great and you escape Panda.
So is anyone with a healthy site willing to share some statistical information about their site so we can see what a healthy site of that type looks like in statistical terms?
These are my stats for May 2012 to give you the idea, but my site is affected by Panda so please don't use this as a model for your site!
1) Type of Site: Ecommerce (hit by Panda)
2) Home page: exit rate 35%, average time on page 1.08, bounce rate 35%
3) Category pages: exit rate (15-35%, average 30%), average time on page (30-60 seconds, average 50 seconds)
4) Product pages: exit rate (2-75%, generally around 15%), average time on page (max 2 minutes, average 50 seconds)
5) Proportion of pages with exit rate higher than 30%: 25%
6) Proportion of pages with average time on page of less than 20 seconds: 20%
I'm working on the basis that for my type of site, I don't want more than 30% of my visitors leaving the site from any product page and that it would take a customer at least 20 seconds to take in the product description. My stats show that a third of my product pages are performing worse than this.
I know Google stats will be a tad more sophisticated than this, but there has a to be a statistical shape sites need to be to fit the profile Google have for each type of site. If we know that, anyone affected by Panda has something to aim for when sorting out their content.
Maybe it would be helpful to classify a couple of different types of searchers, come up with behavioral models for satisfied visitors to those sites and then look at what metrics might reflect their happiness. I mean, there's a big difference between how I behave on sites when I'm looking up a historical fact ("who won the battle of whatever") or comparison shopping for a pricey item where I want to make sure I've made the best choice.
I would like to believe that just a single metric would be the answer (and I haven't paid much attention to exit rate until now), but I've always believed Panda is a combination of many many factors.
I don't think you take metrics and compare them between completely different sites. My thoughts were that Panda was a niche-based thing: your metrics are compared against those sites in a similar niche.
I have two such sites. One was smashed by Panda 1, while the other has never been touched. Both are in the blog/content space.
Site hit by Panda (SITE A):
Time on Page: 2:29
Site ignored by Panda (SITE B):
Time on Page: 2:18
Site A has far more broader and varied content. Site B has far more focused content within the niche. This explains some of the difference in bounce - users find similar things of interest in the narrow-focus site. Site A is far more difficult to integrate related content together.
Many of the search keywords for Site A tend to be asking a question. The user gets the answer (or not!) and leaves.
Site A also had lots of scrapers in it's pre-panda heyday. From the day Panda hit, the scrapers immediately outranked the original article (no matter what we tried - delaying RSS etc, DMCAs). This doesn't happen any more (probably due the low profile of the site, no-one is copying it).
I am still unsure whether the site was hit by Panda or the ATF update, as traffic was lost on January 20th, which is extremely close to both ATF and Panda refresh. But here are the stats of 2 large websites, 1 hit by Panda, one not.
Panda hit site (25,000 - 30,000 uniques a day before, 15,000 after)
Bounce rate: 32%
Time on page: 0.45
Time on site: 7:45
Pages per visit: 11.36
Site not hit by Panda (80,000 uniques a day)
Bounce rate: 63.5%
Time on page: 1.52
Time on site: 3:35
Pages per visit: 2.92
I'm not convinced that user metrics are being used as much as some claim. But that is based off my own experiences. Improving user metrics had no effect on traffic for me.
@realmaverick - might not be Panda.
My site that was hit had a dramatic drop off in traffic:
Pre-panda 25000-35000 uniques per day
Panda 1: 8000-10000 uniques p/day
Panda 2: 3000-5000 uniques p/day
I honestly can't see a correlation with Exit Rate in terms of pages that still get some Google referrals. Remember panda was originally called the "Content Farm" update...
I'll bet that Google's Panda is not just about average anything. For example, take the Exit Rate idea. At the very least, If I were using this metric it would need to be modified by time-on-page for each visitor, not just used as an average.
I think that time spent searching after one leaves the site is also an important factor and which we don't have any metrics for except for Goog.
If visitors repeated continue searching after leaving your page, is a clear indication that user experience was poor. So there must be some comparison done between sites with similar niches to see which sites have a low search rate after visitors exit.
|If visitors repeated continue searching after leaving your page, is a clear indication that user experience was poor. |
Even that metric isn't so cut and dried as I see it. First, there are two categories of people who return to Google: those who make another choice from the same search results and those who revise their query. Sometimes the metric would show that the search results were poor and sometimes that the query was poorly formulated. And sometimes that the search user is doing resarch and comparing several differnt results - all of which may have been just fine.
|If I were using this metric it would need to be modified by time-on-page for each visitor |
Each visitor? I'd love to but I'm afraid it is not practical when a certain page gets hundreds (sometimes thousands) of visitors?
| This 193 message thread spans 7 pages: 193 (  2 3 4 5 6 7 ) > > |