Forum Moderators: martinibuster
On this particular site i'm finding:
- The highest click through rate is with just one ad unit smack in the middle of the top heat zone, nicely blended of course.
- The highest earnings per click is with the ads hidden away towards the bottom of the page, the ad up top removed.
With the ads at the bottom visitors aren't clicking off on them on the first page they visit as often. This sites bounce rate matches the earnings per click at every step. As the bounce rate goes up, the earnings per click come down and vice versa.
The site is not an MFA site by any means. Its becoming apparent that if someone clicks on an ad from the first page they visit that they increase the average bounce rate and I earn a reduced amount. Is the best way of keeping bounce rates low, and earnings high, to simply not show ads until a visitor is looking at the second page of his/her visit?
I could easily code my site to not display AdSense on a visitors first page but would that violate the terms of service? For this site at least its simply not worth having someone click on an ad from the first page they visit, it's costing me money and traffic.
Is this new? is it possibly to combat MFA sites who try to get visitors to click on ads right away? Thoughts?
I assume you'd be using PHP server-side, and not some Javascript kludge which might well violate the TOS.
I think you may be on to something.
Is the best way of keeping bounce rates low, and earnings high, to simply not show ads until a visitor is looking at the second page of his/her visit?
For sites like mine that's quite a problem [and I hope AdSense aren't taking bounce rates into account.] People mostly come to my site from SE's and direct links. The rest are bookmarks.
They go directly to pages like "Left-handed widgets with green tops". Probably print that page out and then move on.
They are NOT interested in "Right-handed widgets with red tops" which is also on my site.
"Surfers" may work their way through my site but the majority come, get the answer they want and depart. My own web search habits are almost identical, I usually find my answer on the landing page. Rarely do I go to anyone's home page and wander around the site.
[edit] made an addition
[edited by: IanCP at 1:26 am (utc) on Sep. 27, 2007]
- If a user is arriving via search and immediately clicking on an ad, the publisher isn't contributing much, if any, "added value" to the process. Most of the value is being supplied by the search engine (which delivers the prospect to the site) and the advertiser (whose ad is producing a click).
- If, on the other hand, the user is clicking on an ad after visiting a few pages of the site, or possibly after spending X amount of time on the page, then it's at least possible that the user has been "qualified" to some degree by the text on the page(s), and that the publisher has added real value to the process instead of merely flipping the arriving prospect to the advertiser.
It's reasonable to guess that, if bounce rate is figured into compensation, it's merely one of several factors that Google uses to determine what a publisher's traffic is worth (beyond the nominal bid value of the clicks).
There is a small positive correlation 0.3 between bounce rate and earnings, meaning the higher the bounce rate the higher your earnings.
These are over 600 data points.
BUT: time series data, the algo might have changed
So let's do it over the last 90 days, to at least reduce the chance of algo events screwing with the data.
There is a relatively strong negative correlation (-0.6) between bounce rate and money earned.
The higher your bounce rate the lower your earnings
Now it could of course be that lame users don't click .. that summer users bounce more or don't click, so how you interprete this is another matter. Additionally my users are science etc based so I have a rise in traffic and earning as they all return to school and uni, so a simple analysis like this has to be seen with caution.
On the long term data, I would say it's possibly trivial. If your users click away you of course make more money .. The ideal user reading a lot and then clicking will be far away if you can't say "I recommend clicking on the ads after you read all this, it might just what you need." ;)
The higher your bounce rate the lower your earnings
If that is reality then AdSense need to get a real grip on reality.
Does anyone seriously suggest that searching for "XYZ Widgets" should take you through ten pages of a site and in time of around about 20 minutes?
My expectation is about 15 seconds for search, 15 seconds to load the page, 5 seconds to confirm it is within my expectations and the print time is purely local.
[edit] Typo's
[edited by: IanCP at 11:18 am (utc) on Sep. 27, 2007]
But one could also support the hypothesis the better Google does its job the higher the bounce rates would be. If they lead you to exactly what you need, the users would spent less time on your site possibly.
But one could also support the hypothesis the better Google does its job the higher the bounce rates would be. If they lead you to exactly what you need, the users would spent less time on your site possibly.
In such cases, isn't it Google (not the publisher) who's adding the most vaue to the process, and who ultimately deserves the most credit for any advertising clicks that result?
It seems to me that the publisher deserves more credit when he or she has presold or prequalified the user (i.e., the lead) instead of merely acting as a second chance for Google searchers to click on AdWords.
ADDENDUM:
Let's look at a couple of "could be real-life" examples that illustrate my point. Both involve a user's clicking on an ad for a Widgetco WC-1 digital camera:
- User A visits one of the major camera-review sites, reads an eight-page review of the WC-1, and decides "I'm interested in buying that camera" as a result of the review.
- User B visits an "AdSense entrepreneur's" Web page that consists of 300 words about the camera (cribbed from the manufacturer's product sheet) and three AdSense ad units. He sees nothing of interest on the page and clicks on ad ad whose headline is more compelling than the page itself.
In the first instance, the camera-review site is providing value and referring its user to the advertiser; in the second case, the AdSense entrepreneur's site is merely flipping Google's user to the advertiser. If it offers a greater payout to the review site (the one with the lower "bounce time") than to the AdSense's entrepreneur's flip-'em-as-fast-as-you-can site, Google accomplishes several things:
- It earns more revenue on clicks that are driven primarily by Google Search and AdSense ads (i.e., flipped Google traffic);
- It encourages real Web content while discouraging junk made-for-AdSense sites;
- It encourages the kind of user experience that, over the long haul, is necessary to keep the AdSense network viable.
DISCLAIMER: Again, if "bounce rate" (or time on site, for that matter) is a factor in publisher earnings, it's likely to be only one of several or even many.
[edited by: europeforvisitors at 2:53 pm (utc) on Sep. 27, 2007]
The Google help pages define Bounce rate as the percentage of single page visits. Bounce rate is a measure of visit quality and a high bounce rate indicates that the site entrance pages aren't of high quality.
Now this is in my opinion oversimplified. A normal Wikipedia page I visit, I use this one page and then I am off. Additionally if there is a lot on info on one page I just visit one page.
Only if the time on site is VERY short then I would say it fits. I expect Google to do that anyway.
The values marked by them primarily as influencing quality is Average Page view, time on site, bounce rate and new vs returning.
In a Nutshell, bounce rate isn't everything and one would really need to do a thorough analysis of all factors to draw a valuable conclusion on what's going on.
Also, some advertisers which use popup windows on their landing page probably don't mind high click-through rates.
p/g
In a Nutshell, bounce rate isn't everything and one would really need to do a thorough analysis of all factors to draw a valuable conclusion on what's going on.
Of course. That's pretty much what I said earlier But there's no reason why bounce rate or time on site couldn't be one factor that Google takes into account, along with other factors, when determining publisher compensation. (To oversimplify, Google could be asking, "Is this user primarily the publisher's user or ours?")
[edited by: jatar_k at 3:52 pm (utc) on Sep. 28, 2007]
[edit reason] remove notify [/edit]
Very ad hoc as it's a times series, I am 3 years away from the last stats in my thesis, so not 100% but hey there is something there, the following linear model explains 82% of my money :)
R - squared = 0.829
May I quote stat soft:
(e.g., an R-square close to 1.0 indicates that we have accounted for almost all of the variability with the variables specified in the model).
I kicked "time on site" (p value was over 0.05) out and the data is on last 3 months assuming no algo changes, no time influences (obviously they will be there start of school etc), but I think Googles algorithms should be independent of school start (although they could be ;))
*** Linear Model ***
Coefficients:
Value Std. Error t value Pr(>¦t¦)
(Intercept) 700 160 4.4 0.0000
bounce 880 190 4.7 0.0000
ppv -60 24 -2.5 0.0148
percnew -1000 120 -8.5 0.0000
Multiple R-Squared: 0.829
meaning
money = 700 + 880 * bounce -60 Pages per visit - -1000 * percentage [in decimals] of new visitors. I rounded the values so its easier to read.
So my money is in this model influenced by bounce rate, pages per visit and percentage of new visitors.
Probably there is also a relationship between all the variables in themselves (for example higher bounce rate leads obviously to less pages per visit, unless many bounce but those that stay read a lot.)
You could write a thesis about that so I am not gonna ... as I doubt anyone would want to listen ... Let's continue guessing ...
Disclaimers: I do not claim this is perfect, does reflect your server, or in any way Googles algorithms. 1 + 1 = 2 but 4 - 2 is also 2 and 2*2-2 = 2 ... :)
That's the "bounce rate" that Google is talking about in the quote listed above.
If a surfer lands on a publisher site, then goes forward by clicking on an ad, that's not really a bounce at all. The number of pages viewed before clicking an ad could conceivably be used by Google in some way, but IMO it doesn't say anything whatsoever about the quality of the traffic, so Google would be foolish to rely on it for anything. I could imagine this information feeding into their click fraud algorithm, but that's it.
In other words, if people are seeing a statistical effect, it's probably because they are sending more, but less motivated clicks, to advertisers. The clickthrough rate is related to this, but is unlikely to be the actual mechanism Google uses to determine the quality of the traffic.
The Google help pages define Bounce rate as the percentage of single page visits. Bounce rate is a measure of visit quality and a high bounce rate indicates that the site entrance pages aren't of high quality.
Well there is only actual data on the analytics bounce rate. There might be loads of bounce rates, but this seems to the one Google uses and at least quotes as a measure of quality.
And why not make an attempt to take one of the claims now seriously and least try to analyse it with some real data?
All I'm saying is that bounce rate is very important from an advertiser's perspective. That's because most advertisers are in a situation where surfers must visit multiple pages before a "conversion" can occur. Arbitrage sites, I suppose, are a notable exception. But bounce rate is NOT necessarily an informative statistic for websites that get organic traffic to deep pages from organic links and search engines.
The bottom line in these "what Google decides to pay" threads is that if Google pays out less, it's because they're charging the advertiser less (and thus Google themselves earn less as well). I don't see any logical reason why Google would give an advertiser a discount based on how long the user spent browsing the previous website before deciding to click an ad.
The bottom line in these "what Google decides to pay" threads is that if Google pays out less, it's because they're charging the advertiser less (and thus Google themselves earn less as well).
Not necessarily. Smart pricing and the compensation formula (a.k.a. the payout percentage) are two different things.
[edited by: jatar_k at 3:52 pm (utc) on Sep. 28, 2007]
[edit reason] remove notify [/edit]
Again, if "bounce rate" (or time on site, for that matter) is a factor in publisher earnings, it's likely to be only one of several or even many.
You got me thinking, EFV. My best performing sites have very high bounce rates, but VERY long time on site (page): a 900-1100 second average for a month isn't unusual. The eCPM's are far above what I expected when I started the project.
Could G be using the time-on-site as a factor (one of several, as you suggest) in setting the pay-back?
- bounce rate can be defined as a single page visit, I've confirmed that a click of an ad is the same as a click of a backpage button.
- The ad units were identical and text only. The content on the pages didn't change in any way besides the moving of the ad location. I left it for 14 days in both locations, repeated three times.
- Time on site is a wildcard but its effect can be surmised to be minimal. The reason being that it would take longer to get to the ad at the bottom than to click off up top. Tracking showed no real time difference for either location.
The data being what it is i've optomized as follows.
- top entry points receive ads at the bottom.
- top exit points receive ads up top.
- added an image banner at the very bottom of every article and offered it up as an ad placement. ecpm doesn't care about bounce rate and the image ad won't compete against the text ad unit.
If anyone does any further testing on earnings vs location vs bounce rate i'd love to hear about it. I might even be down for a thesis.
You can study that if you want, although every single site has a unique profile so that makes it hard to extrapolate data.
Each animal/company/server has unique behaviours and common behaviours.
If you want to study anything you start with a model organism/company/server and a model and try to extract as much knowledge from it as you can.
If someone else would do something similar or better you have a useful discussion on actual data and not just the usual chaos, which isn't that helpful either, is it?
There is no absolute way to know for sure, thats a given.- bounce rate can be defined as a single page visit, I've confirmed that a click of an ad is the same as a click of a backpage button.
There is also no sure way to know when you die, still insurances use statistics to estimate when you pop your clogs. In the absence of all possible data, you use what you have to make some sense of the noise (variation).
Thinking about all possible variations is useless. Make a hypothesis, test that hypothesis. If the hypothesis is cr@p make a new one. Test that. You gonna get a whole lot farther, than thinking about all the possible variations.
Since Google, the one providing the ads, the stats, and the traffic made a definition, I would think it's better to start with their definition than to define your own variation.
The task is to find out what Google does and not how any of us see bounce rate. Their bounce rate and other variables are the ones that make money or not.
And yes maybe they use something else somewhere else, but the analytics data we have is from them and probably more likely to be close to what they actually do than what we dream up.
Now if they claim on analytics 4 variables are a measure of quality they must be related in some form to income and traffic.
So use your data, analyse it for your server and even if it only fits your server you A) know more and B) can possibly make more money.
A model that explains your data, does not have to be the one Gogle uses, if it still brings the right results it fits your server.
1+1 =2 and 4-2 = 2
For example, if you change your CTR from 2% to 4% by making your ads more prominent, you'll only have a tiny impact on single-page visits but you'll double the number of clicks. That will probably have a profound effect on the traffic you send - in terms of how they are browsing, what they are trying to accomplish, how motivated they are, etc.
This really seems like a more promising line of investigation. Focusing on peripheral considerations such as "bounce rate" might lead you in nonsensical directions, such as forcing users to perform extra clicks in order to get to the information they're looking for.
This really seems like a more promising line of investigation. Focusing on peripheral considerations such as "bounce rate" might lead you in nonsensical directions, such as forcing users to perform extra clicks in order to get to the information they're looking for.
Theoretically I agree,
If you want to think for your visitors and not for Google do what you think is right, best strategy there is.
On the other hand Google assesses by algorithm, if that algorithm decides your page is unprofitable they likely react to that .. so if income is your interest, their self defined quality values are what counts.
The thread title says:
Bounce rate vs earnings - they seem infinitely tied.
And from what I see that's correct.
If manipulating these values actually leads to anything is a totally different matter. I personally wouldn't do it.
Please consider that if G is tracking bounce rates, that they are not relying on the Urchin/Analytics script to determine this for the purpose of pricing - rather, the presence of the AdSense scripts would be the primary tracking method. (Otherwise, they would only be able to determine bounce rates based for those using Analytics, which would be silly).
In other words, if you choose to not place AdSense on a page in order to alter Google's idea of the bounce rate, you are altering G's statistics gathering, and perhaps not in the way you'd hoped.
I also advertise on adwords and blocked advertising on those sites that provide visitors who don't click through to other pages. ie, those visitors with high bounce rates.
In the end I'm not sure if all this helps income but I sleep better and feel at least I'm trying to do something.
Please consider that if G is tracking bounce rates, that they are not relying on the Urchin/Analytics script to determine this for the purpose of pricing - rather, the presence of the AdSense scripts would be the primary tracking method.
Sure but the data you have is obviously not always the data you need. :)
This also corresponds to the value added analysis that EFV states. IF the publisher's product is so thin - as evidenced by the click latency - then is the publisher actually bringing filtered, pre-qualified, pre-sold and higher converting clicks?
Or is the publisher driving "fast clicks" that are otherwise poorly pre-sold?
Seems logical to consider click latency a factor in smart pricing.
Of course, not being a skilled programmer, I don't know jack about whether it is possible to capture this data accurately. Ho-hum . . (Ignore the man with only rational analysis to offer . . at your peril? ;0/ )
Seems logical to consider click latency a factor in smart pricing.
Or payout share, for that matter. Why should Google pay the same share for a "passthrough lead" as for a publisher-originated lead?
As for pre-qualifying visitors, that's a possible connection, but I'm very skeptical. The concept has much more relevance to affiliate websites. With AdSense you can't control or even know what ads are being shown, and ads are often only loosely targeted to the theme of the site or the page.
