News Source: http://seoblackhat.com
Published: Nov 25, 2008 - 05:43 am
Story Found By: UtahSEOpro 1175 Days ago
Category: SEO
Here’s are screenshots from one of my sites (it’s a “white hat” site or I wouldn’t be posting the footprint here). The first shows overall traffic:
41 Comments
Comments
If this is true and actually happening, then isn’t the answer to simply not use Google Analytics on your website?
@Jill, won’t help if these bouncing guys are coming from and going back to google serps. so would rather try to engage them :) in general though it would be a good thing, why would you care about ranking for not-relevant phrases? on the other hand that is just trusting the users, not so smart neither :)
It´s an interesting thought, but I don´t think Google uses bounce rate as a ranking factor. There´s simply too much noise in this signal: There’s no standard definition for bounce rate. Some analytics vendors define bounce-rate as the percentage of visitors that stayed less then 5 or 10 seconds on the site. Others (like GA) see the bounce rate as the percentage of visitors that only visited one page on your site.Also, if the bounce rate is high, is that a good thing or a bad thing? Are visitors leaving because they can’t find what they’re looking for, or is the information provided so complete that they don´t have to browse through the site anymore?And what to do with all these other sites that don’t use GA? Would they be excluded from this signal?
Desphunn for spreading of rumors and faulty interpretation of data.
@Jill: Quite the contrary: we should all be using Google Analytics in combination with some script that reloads the webpage every X seconds. Mwahahaa!@Tadeusz: Do you care to tell us what the correct interpretation of the data is?
@ onreact: good to know there is one other person on the planet smart enough to have caught this. desphunn.@ lawnchair: look at the graphs. the data is not consistent with the thesis.@ everyone who sphunn this: did you bother to even compare the data provided against the author’s thesis?
Google might very well use bounce rate in its algorithm, but the evidence provided in this article doesn’t come even close to "confirming" this.
I could see bounce rate being considered -- IE, watching what people are clicking on in search results, noting if they come back to the same page and click on something else. That’s a long-time signal used in the past with Direct Hit and probably with others. I wouldn’t doubt if Google considered it among many other factors.But these charts prove nothing. His Google traffic dropped the same time his bounce rates went DOWN? If it’s bounce rate that’s important, than if his bounce rate down, shouldn’t his traffic go up.I suspect what happened is that the site was ranking for a particular term, getting a lot traffic for that term and wasn’t particularly compelling for it. It loses rank one day (which could be for a variety of reasons) and a chunk of traffic goes. The remaining traffic, they’re more relevant for those terms, so the bounce rate gets less. When they start ranking again for the other terms, bounce rate goes back up.Anyway, I’d like to see a heck of a lot more data.
Danny.. DM me on twitter.. i will explain the experiments (charts/graphs), i have been compliing since july..
finally some people have some balls to desphinn something. good work.
@LawnchairLarry Do you mind reading my desphinn comment? I rarely desphinn anything but here the faulty logic is too obvious and the claim too spectacular.@arteworks Just distrustful ;-)@Chris1 Yeah. They probably will one day but bounce rate can mean a lot of things (not just the page is bad).@dannysullivan Thanks for the common sense.@UtahSEOpro Sorry to desphinn you but a worthy discussion does no good if the claims are so wrong like in this case.
You guys really missed his point.Lets go over the basics again - Bounce Rate (%) refers to the percentage of people that bounces away from your site without visiting any other page in your site. So the Lower the bounce rate the better.Imagine Google Search sends you 1000 users a day (from 10 keywords) and you have a bounce rate of 85% for Google Search ONLY (You can see that it analytics).And lets say out of that 10 keywords, 4 of the keywords was not relevant to your site (You were ranking accidentally, or your content is just not good enough) - so the share of the bounce rate for those 4 keywords will be the highest which makes the major portion of that 85%.So if google stops ranking you for those 4 terms and gives the ranking to another site (to test the performance), What happens? 1. You Traffic will Decrease -- As you are only getting traffic from 6 keywords and not 10.2. Your Bounce Rate will Decrease -- Because that 85% bounce rate included those 4 "bad keywords", now since those are gone, your bounce rate will go up as the other 6 keywords that are sending all the traffic from Google are all good.Hope this clears the confussion.
Update: 2. Your Bounce Rate will Decrease -- Because that 85% bounce rate included those 4 "bad keywords", now since those are gone, your bounce rate will go DOWN as the other 6 keywords that are sending all the traffic from Google are all good.And Just to add on to my previous comment, I think some of you are confused because you haven’t seen his follow-up post (with some more graphs) - http://seoblackhat.com/2008/11/21/bounce-rate-questions/The dated graphs kind of clears everything.
@onreact don’t apologize. that’s what social media is for...creating healthy discussion. if this post raises a good discussion then it’s served its purpose.
This is a psychology experiment right? :)
Well said UtahSEOpro :)Desphinning an article should not be taken personally. It’s what it is there for, heh.
How is the supposition that bounce rate is a ranking factor ’confirmed’ by this data?Interesting theory, yes, but not even close to being confirmed.
@"all who believe they can design and conduct a valid - scientific :) - experiment involving page rankings and are at the same time forgetting about hidden variables"have some fun reading thishttp://www.timeshighereducation.co.uk/story.asp?sectioncode=26&storycode=404397&c=1
Hey guys - if you check the write up of teh October Webmaster Q&A here you’ll find this was raised as follows:chrisff, palo alto, ca: Is the bounce rate and speed taken into account when ranking a page? i.e. if you see a searcher click on a result then return very quickly and choose another result, is the first page ranked lower?JohnMu: Hi chrisff, assuming that users will be jumping out of a site like that, there’s a high probability that they won’t be willing to recommend it to others (or come back themselves). So yes, indirectly at least, if a site is made in a way that users run away right away, then chances are that we might not be recommending it as much as other sites that users like (and recommend to others).Read into that however you like. In the meantime I’m going to desphinn the post. Too many gaping holes and not at all a proper test!
I agree with the bounce rate theory. It holds just as much water, if not more, then some of the other nonsense being floated around, like W3C Validation, by SEOs.If you think Google is going to annouce to the world how they rank webpages, you are dreaming.After all, Google claims to be interested in the user experience. Do you guys really expect Google to hand you everything on a silver platter along with the Thanksgiving roast? It is good to see that some folk are considering other things other than Mob Hysteria.
A very misleading post title. Nothing has been confirmed. An observation has been made on a correlation between two variables: traffic and bounce rate. <div><div></div><div>The explanations offered earlier in the comments are valid - search rankings fluctuate, and websites often drop out of the SERPs for certain keywords. </div><div></div><div></div><div>If a Womenswear website was sporadically ranking for ’mens clothes’, I would expect higher traffic and higher bounce rates on these days. It could also be to do with a US website ranking in the UK, or any of the other reasons that could account for a higher bounce rate/disengaged visitors.</div><div></div><div></div><div>I am not discounting the idea presented by this post, but the notion that it acts as a ’confirmation’ of bounce rate usage in search rankings is rash and unfounded.</div></div>
Danny is on the ball here - but let’s face it, "proving" anything relating to Google’s use of data in its algo is not easy at all. However, Quadzilla is an extremely experienced SEO operating in very competitive markets, and he’s offering an insight into what he sees as an indicator of bounce rates being used. I’ve posted similar before, but as with QZ, cannot prove anything.Rather than diss the observation, some people would do better to learn from those who can still teach. You don’t have to accept what is said, but you ignore it at your peril.2c.
I believe that there is a surface level correlation between bounce rate and traffic levels as indicated by the pretty analytics graphs provided, but the causality between the two occurences is questionable at best:1) Danny Sullivan above mentioned a different hypothesis alluding to other potential factors that could be at play in affecting traffic levels independent of bounce rate. 2) There could also be a myriad of other ingredients causing the traffic fluctuations that haven’t even been considered up until this point.3) The amount of data is insufficient to make a strong causality argument. 4) Maybe the link between bounce rate and aggregate traffic levels should be considered more independent from the link between bounce rate and search engine rankings. I see no fundamental problem with Eric Schmidt using bounce rate as a algorithmic variable, my main argument is that I don’t have sufficient evidence from the article, that this is indeed the case.
Bounce rate give mixed signals. If the page is highly relevant to what the searcher is specifically looking for, they can get their info and leave without going to any further pages - fully statisfied. A Big vote for relevance.Alternatively, if the page is not so relevant and the searcher cannot loacte the piece of information they want - they could end up searching many of the site pages, all in vain. A Big Vote for non-relevance. The statistics that could be more useful from G’s side is if a person clicks on a site in the SERPs and then, after visitng that site, comes back to the SERPs and clicks another site. That would give a pretty good signal that the first site was not sufficiently relevant. Also - there would be no need for G Analytics as all of the data takes place on the SERP.
Look, I can buy that bounce rate might indeed be one factor. Like I said, that’s definitely been used by Direct Hit in the past, and I’m pretty sure Google’s published some patents or research papers on it, measuring bounce off click going back and forth from search results.It’s just that the charts themselves don’t prove this to me, not even the second set that were put out there. I see a drop in traffic from Google and a drop in bounce rates. And that proves Google’s pulling traffic that generates high bounce rates? Or it proves that maybe that the site lost traffic for some terms for a variety of other reasons that it wasn’t particularly relevant for. And then the traffic goes back up a few days later? Because Google just decided to experiment for three days?What I suspect is that the charts seem so clear to Quadzilla because, as he’s written, he’s seen a pattern of things that make him convinced that this is going on. But we haven’t shared seeing that pattern, and if you put out a post saying "confirmed," then I really want it not to absolutely clear.What I’d really like to see is the actual keyword data. What terms were lost.But overall, wouldn’t surprise me at all if Google’s considering bounce rate in a variety of ways -- probably primarily by looking at clicks back-and-forth from search results, then using some toolbar data (for those using web history). That’s my speculation, by the way. I see comments from folks saying "Google does this, Google does that." They don’t actually say what they do. Usually, they just say they can’t say what they do.
I dunno; I hear this a lot, but my (admittedly anecdotal) own evidence goes to the contrary - I have a couple of sites myself that have very high bounce rates, because there’s one only effective page - it’s database driven, there’s a search box, and the search displays the results on the same page. Or I might have a highlighted event in a box on the page, people come, look for the event, find it and leave. And this sites ranks #1 in G, Y & M/L for almost every relevant search and has for years. I’ve got clients with shockingly high bounce rates (at least as interpreted by GA) who are ranking all over the place. I also have clients with extremely LOW bounce rates who can’t get arrested in the SERPS (so to speak)Over maybe 300 sites, I just don’t see it.
@dannysullivan I agree that the title may be a bit off.But its not new that google looks at over 200 variables to rank your site. GAnalytics does show a number of varriables (bounce rate, avg time on site, Pages/per visit etc.) so I feel that its ok to assume that ’bounce rate’ is indeed a ranking factor. If you were running google and you had access to all these different kind of data...wouldn’t you be using some of them? If 90% of the people comes to a particular site and bounces off imeditely that shows Google - either the site doesnt have a good user experience or its not ’relevent’ to what the user have searched for. So obviously that site doesnt deserve the ranking for that specified keyword.Similarly CTR in the SERP is also one of the ’ranking factor’ in my opinion. if you are number #1 for the term "Auto Insurance" and the searchers are clicking the No #2, 3 and other results more than your result then that means definitely something is wrong. Google might put you down in the SERP and promote the sites beneath you to test. (And nowadays they do this quite often – They will promote a brand new site way into the top for few days/hours and then suddenly Its gone again).If that new site can prove its effectiveness and cross the ‘threshold’ set by Google…Their ranking would stick or improve from their previous position. Even in PPC CTR is one major factor for your Quality Score.I know that none of these can’t be proved until we hear anything Official From Google. (Which is never going to happen). But for those who are constantly doing A-B testings and tracking their results, I think they already know many of the ’ranking’ factors out of that 200.And that 200 variable is likely to increase if they keep their searchwiki default for everyone.
Google Analytics does show a bounce rate, but it’s an exceptionally bad idea to assume that because of this, Google Web Search uses bounce rate as a factor in general or the GA rate in particular. They are entirely different products, run by entirely different teams. And since not all sites use GA, Google Web Search -- if it really were the GA bounce rate -- would be blind to some sites.instead, if Google Web Search is measuring bounce (and again, I wouldn’t doubt it), they’ll look to measures where they have universal data for web sites. The toolbar gives them this (and more, it will record time on site). Pogosticking from clicks on a search results page to a site and back to the results is another example.Ironically, there’s not much to do with the idea that bounce might be taken into account. This isn’t a new concept. Google’s been doing clickthrough tracking from results for ages. We’ve had advice for ages that content need to be compelling and on target for the terms people come to to help with for ages. It always seems to come back to the usual advice. Have good content.
Danny I totally agree. Prolly the data source can be the toolbar (provided that the user has it installed), or the Analytics (provided that its installed in the site). And if the user is using the browsers back button and clicking another result -then I guess that data can be used as well.Well the whole point of this discussion is even though the graphs and whatever facts that the author has put forward may not be substantial enough to create a hypothesis that "bounce rate is a ranking factor...", i think it makes a ’little’ sense that there MIGHT be a connection between these two. Obviously there are many other factors that can cause this and if the author discloses more info and shows which keywords he was ranking for etc. it would be clearer.But like you said, this is nothing new and has been there for ages. But I feel that sometimes even though google collects a lot of data, they don’t use it all up at the same rate. They are tweaking and improving their algorithm every day, so a variable that has 2% importance today may be increased to 10% next month. We Just don’t know...and can only guess. Btw, I really hate to type comments in sphinn as its very hard to read (as I type in the box). How do I increase the site of this? Any help would be welcome...my eye hurts :(
@SeekGeek said "The statistics that could be more useful from G’s side is if a person clicks on a site in the SERPs and then, after visitng that site, comes back to the SERPs and clicks another site. That would give a pretty good signal that the first site was not sufficiently relevant."Um..no. Like most things, it depends on the query. Scanning the web to collect data involves clicking back and forth to the top N sites...click, mental note, return, click next, mental note, return, etc. Think about a restaurant name search... you search to learn about the restaurant. Click to read an opinion/review, click back for next opinion... eventually an impression has been formed and the user is satisfied. A perfectly valid pattern of behavior for even the most relevant web pages (concise, on-topic pages are easily scanned to make a note and move along for the next opinion). That Google quote above is typical of today’s evasive Google. Don’t answer the question, but instead hint at how logic might deduce that everything that is good is recognized as good by Google eventually, via the secret Google sauce. No accountability. I fear it will stay that way until we see a competitor.
It takes a bit of thinking, but you can see the argument being made (though it seems a bit backwards at first glance).Of course, someone could take the same statistics and argue the opposite.Benjamin Disraeli would just love the modern era.
No one will ever proove these sorts of factors conclusively so... recognize it’s a possibility and act accordingly. Improving bouncerates is not only possibly good for rankings but it certainly is a big plus for users and conversion. Lots of times it has nothing to do with relevancy, its GUI or presentation or trust in the site.Do I believe that SERP clicks are tracked and more likely a factor in ranking? Most certainly! Measuring Bouncerates is one of the truest tests of relevancy for the user and quality of the site for the query. So to think that it isn’t possibly a factor is to not get that SEO is as much about what could be, as what is for sure. Using both as a criteria for usage means you stay ahead of the horde who is looking for only provable info to build a SEO strategy on.
yeah - just create good content and users will bounce less and stay longer and... :) It’s all true, though - eventually, SEs will figure an algo/equation that’ll discount all your manipulations and count quality-per-user somehow.
Sorry, this is crap. Spreading rumors. Moreover the data would prove that the lower the bounce rate the lower the traffic which does not makes sense.
The data provided by the author does not support the thesis and has been incorrectly interpreted.
I’m with onreact. The conclusion is a little incomplete and to be validated, some significant testing would need to be set up! I can’t wait to find out the truth!
Sensationalist and simply untrue. Nothing has been confirmed!
The poster is saying that the higher the bounce rate the lower the traffic, but his stats seems to show the opposite: the lower the bounce rate, the lower the traffic!
Nonsense. Bouncing is a sign of high quality (think Wikipedia), a sign of low quality (think MFA), and no signal at all (think Google Mail). Any evidence missing, those charts might tell anything.
Nice work. Got everyone wondering whether or not it’s true and probably tons of free links. There’s no proof in that article whatsoever.
this article is junk.gdfgsdfgsdfgsfdgsdfgsdfgsdfgsdfgsdfg