Forum Moderators: martinibuster
I've been reading various things about whether or not CTR impacts your earnings, with some folks saying that it makes no diference, and others explaining that Google rewards the higher performers (sites with higher CTR's) by showing higher-paying ads on their sites.
First of all, is this true?
If so, when you are doing work on your site, do you turn off the adsense, so that your own page refreshing will not lower your over-all CTR?
Thanks!
First of all, is this true?
Welcome to the WW Adsense forum, Ellen.
Yes, it is true that some people think it makes no difference (eg: me) and others think it does. It is a debate that rages on because, in truth, no one knows for sure except Google, and they aren't going to tell you.
However, this debate only applies to smartprice. If you can increase your CTR, you are very likely to earn more money because you'll be getting more clicks and you get paid per click. So it is a generally a good idea to try to get the best CTR on each page (if you have excessive CTR, though, such as more clicks than page views, there might be a different argument to be made, because of smartprice)
Personally, I don't switch off Adsense when working on my site or specific pages, because I like to monitor what ads are appearing. But it is discouraged. In fact, we don't view pages very often, and I could anticipate that viewing pages lots of times (without clicking) might make the Adsense robot think that it needs to try some different ads, so that might make the targetting worse. And you can't click on your own ads (that's fraud) so it is wise to keep views of your own site to a minimum.
Anyway, that's the view from the Red corner, I'm sure someone else will step forward from the blue corner...
I sometimes see an odd relation between CTR and earnings on one of my sites. In this niche are only few high paying advertisers and they tend to use a fixed budget. Almost every time I I have an increase in CTR or impression rate (by changing ad placement, increased traffic, etc) I see the effect that in the morning the CPM and earnigs increase, but this is compensated for in the evening. The fixed budget of these high paying advertisers causes the ads to be replaced by lower paying ads during the day. The nice thing is, that evety time after some weeks those advertisers increase their budget enough to provide ads during the whole day again.
While that does not clarify things... it does make me want to figure out how to turn off adsense while I'm working on my site and constantly refreshing my pages.
it does make me want to figure out how to turn off adsense while I'm working on my site and constantly refreshing my pages.
Use an Opera or Firefox browser and turn off javascript...problem solved:-)
How would I develop it on my local PC without uploading it?
The key thing is to create/edit the web pages, save them in a directory on your pc, but not "publish" (or whatever the equivalent is in your software). You can view the html files on your pc in IE by using file - open and navigating to the directory containing the files.
"Publish" only when you are finished and happy with the edited pages - either using the publish facility in your software, or using ftp software to upload the files.
Google rewards the higher performers (sites with higher CTR's)
I understood smart pricing has to do with how well the ads convert not how often they are clicked on.
There are definatelly some sort of rewards system involved, but it seems to be reset every day.
My CTR increased about 5 times, but EPC remained in the same range.
So this is for me a proof, that CTR does not influence EPC.
And I thought I'd find ANSWERS here, now I have more questions! LOL
You have to see what works on YOUR site. I get ideas here but I try them out before I believe them.
I thought I'd find ANSWERS here, now I have more questions! LOL
If there were only answers and no questions, most of us wouldn't be here in the first place!
I tend to agree with 21_blue's first post.
But no matter what your theory is on CTR in relation to earnings, you can never go wrong by getting your CTR up.
Lets say you have a CTR of 10% for round numbers sake. Lets say your conversion rate (unknown to you) is 5%.
So for every 20 people you send to the advertiser, one person is converting.
- Now
Lets say that you do something to push your CTR up to 20%. If your conversion rate doesn't go up along with your CTR... you're walking on thin ice.
Going further - this part more opinion than proven fact - I believe that if the average CONVERSION percentage of other webmasters, is higher than your average, you can probably expect your EPC to decline.
I believe that this is on a per advertiser basis, and that it is not an account wide conversion percentage.
I would much rather be at the very limits of smartpricing, than drive my CTR through the roof and let the .03 cent clicks come pouring in. Smartpricing can literally CRUSH your earnings if you drive that CTR too high and your conversions don't keep up.
Smartpricing can literally CRUSH your earnings if you drive that CTR too high and your conversions don't keep up.
Or if the vendor fails to report conversions for whatever reason.
Honestly, to me this just seems like they can really control the price of their ads because it is really up to them to report conversions. If they report none, or only a percentage of them, the publisher gets knocked for it.
Kinda lame..
Lets say that you do something to push your CTR up to 20%. If your conversion rate doesn't go up along with your CTR... you're walking on thin ice.
We have had conversion rates from very low to very high. Statistical analysis shows, on our site, that the CTR doesn't make any difference: correlation between EPC and CTR is zero.
Having said that, we do have a small number of landing pages where CTR can be silly if we design the page right and have ads highlighted in the hot zone so that you can't miss them (CTR ends up being much higher than your CTR of 20%).
What we've found is that, if we moderate this CTR, we can get a slightly higher earnings per visitor because of greater site stickiness and possibly a higher smartprice. But when it happens it is really only slightly higher, and it is difficult to be certain about it because of the constant fluctuation of rates.
IOW, you may have a point. But for someone new to Adsense I recommend getting as high a CTR as possible at first, and worrying about smartprice later.
However, this debate only applies to smartprice. If you can increase your CTR, you are very likely to earn more money because you'll be getting more clicks and you get paid per click. So it is a generally a good idea to try to get the best CTR on each page (if you have excessive CTR, though, such as more clicks than page views, there might be a different argument to be made, because of smartprice)
The thing I'd add is that in my experience, increasing CTR will improve earnings to a point, depending on quality of ads!
At one point I had a very good CTR, but there were a lot of MFA's showing. The bottom line $$ was falling through the floor. I junked the MFA's, the ctr fell to about a third, but the bottom line cash figure started to steadily rise.
I'd suggest that increasing ctr on the basis of quality traffic to advertisers that are actually selling goods and services (as opposed to scrapers, arbitrage and junk ads) is the way to go.
However, it's not always that simple. You eventually come up against budget exhaustion in some cases, and what some feel is Google's profit-raking algo (sudden increases in profit = Google up's it's take more than yours).
Work on having a good site that is well visited, and sends quality traffic to advertisers is the best recipie for long term success with adsense - you can't do anything about the bizzarre behaviour of Random Pricing(tm) or any of the other factors that affect adsense income except concentrating on your site.
At the end of the day it all averages out and the more clicks they make the more you earn, the less clicks the less you earn.
Seems so simple you could commit it to a cross-stitch hanging on the kithcen wall doesn't it?
"God Bless This Click"
It is a debate that rages on because, in truth, no one knows for sure except Google, and they aren't going to tell you.
Actually, this is one issue one CAN figure things out with some simple statistics. By calculating correlation coefficients between the relevant variables, almost anyone can figure out the relationship between, let's say CTR, and CPC (or others).
So, all the speculation in the world is unnecessary. I've always wanted to do those simple number crunches, and maybe some day I'll find the time, but in any event, I wouldn't post the results publicly.
Actually, this is one issue one CAN figure things out with some simple statistics. By calculating correlation coefficients between the relevant variables, almost anyone can figure out the relationship between, let's say CTR, and CPC (or others).
Actually, one CAN'T figure this out conclusively using correlation co-efficients. I have done the maths and I have published the results, - eg: [webmasterworld.com...] - but even though there is a zero correlation between CTR and CPC, this is only suggestive and not conclusive.
There are various reasons that we can't claim this as "conclusive". Eg1: a correlation does not necessarily indicate a causal relationship. Eg2: there may be confounding factors that produce unusual results in this study. Eg3: philosophically, the absence of evidence does not prove the absence of the thing being studied. Eg4: Overall stats may mask/lose smaller variations, including some that cancel each other out. Etc..
Nevertheless, the evidence is sufficient for me to work on the basis that CTR does not directly impact smartprice. Having said that, our current policy is that if CTR on a page is very low we'll try and increase it, if it is very high (>30%) we'll try to reduce it.
Nevertheless, the evidence is sufficient for me to work on the basis that CTR does not directly impact smartprice. Having said that, our current policy is that if CTR on a page is very low we'll try and increase it, if it is very high (>30%) we'll try to reduce it.
I love this :)
Your head is telling you one thing, but your heart says the opposite, and is winning the argument!
Your head is telling you one thing, but your heart says the opposite, and is winning the argument!
Hmm... I think they are all fairly consistent.
Eg: the 'facts' are suggestive but not conclusive. So most of what I do assumes no relation between CTR & smartprice. I don't remove low performing pages, for example, and trying to increase CTR on high volume/low CTR pages is sensible from any perspective (I don't bother with low volume pages).
What might seem odd is my approach to very high CTR pages. In this case, I have no hard evidence but instinct tells me they might convert poorly. As I said earlier in this thread "if you have excessive CTR... there might be a different argument to be made, because of smartprice". This is something I'm still experimenting with and gathering data, so this particular instinct hasn't been through the full testing mill yet.
Actually, one CAN'T figure this out conclusively using correlation co-efficients. I have done the maths and I have published the results, - eg: [webmasterworld.com...] - but even though there is a zero correlation between CTR and CPC, this is only suggestive and not conclusive.There are various reasons that we can't claim this as "conclusive". Eg1: a correlation does not necessarily indicate a causal relationship. Eg2: there may be confounding factors that produce unusual results in this study. Eg3: philosophically, the absence of evidence does not prove the absence of the thing being studied. Eg4: Overall stats may mask/lose smaller variations, including some that cancel each other out. Etc..
Nevertheless, the evidence is sufficient for me to work on the basis that CTR does not directly impact smartprice.
You are right that since we don't know other variables in play we can't statistically control for them, but then again, we aren't doing hard science here. Analysis is never perfect, and using data is far better than nothing. Besides, the situation you describe is almost universal in any stats. analysis of real life data, including that in human and many hard sciences.
That said, you can only use YOUR data to examine the relationship on YOUR sites, and not draw overall conclusions.
The basic question here is: If I know the relationships between CTR and earnings for MY sites, will it guide me in terms of my actions? The answer should still be yes.
As for 21-blue's results, without knowing how you computed them, sample size, etc, it's hard to have any idea what they might mean, or if they mean anything at all.
Here's the bottom line on data analysis. You CAN learn a huge amount that is useful, BUT the more you want to learn, the more time and energy you need to put into the data collection and analysis. For example, you can do various multiple regressions to control for variables that you know about that can confound data.
Since we don't know the variables involved in the adsense algo, the time needed is extensive. There's an infinite number of analyses one can do even with the basic numbers we have. That's because we can't assume anything about the algo. Does it operate differently on one site versus another? Does it change depending on the day? the time? And dozens/hundreds of other questions that you CAN answer statistically for your sites, or overall if you want to collect data from a wide variety of other sites/people.
It's just that uncovering how the algo's work so that you can benefit, at this level, is probably a full time job in itself.
re: your correlation attempts -- Did you have the data plotted, and/or check to see that the (or any existing) relationship between CTR and CPC was LINEAR? If not, your zero correlation is possibly invalid.
Smartpricing can literally CRUSH your earnings if you drive that CTR too high and your conversions don't keep up.Or if the vendor fails to report conversions for whatever reason.
Honestly, to me this just seems like they can really control the price of their ads because it is really up to them to report conversions. If they report none, or only a percentage of them, the publisher gets knocked for it.
I don't believe that Google has ever said smart pricing is based on an individual publisher's or advertiser's conversions. That may be a factor (at least when data is available), but then again, it may not be. And it doesn't have to be. Google can simply make assumptions (based on analysis of data that it does have) about the likelihood of conversions for a given type of content at different clickthrough rates.
Did you have the data plotted, and/or check to see that the (or any existing) relationship between CTR and CPC was LINEAR? If not, your zero correlation is possibly invalid.
I'm not sure what graph you are asking about, here, when you refer to 'plotting the data'. I did create a plot, but don't have access to my data at the mo to look back at what I did - it was either a polynomial regression in my stats package or a trend curve in Excel (which would likely have been polyomial trend, 6 factxrs). But I do recall the result, visually, was a virtually flat/straight line, with perhaps a very slight tailing off near the two ends - 0% CTR and above 50% CTR. But it was so slight as to be hardly perceptible.
With regards further data behind the correlations, I'm a bit reluctant to publish too much more because of the Adsense TOS restrictions. But as long as we don't stray in to that area I could dig out extra stats (eg significance) next week.
With regards further data behind the correlations, I'm a bit reluctant to publish too much more because of the Adsense TOS restrictions. But as long as we don't stray in to that area I could dig out extra stats (eg significance) next week.
No, that's ok. Like I said I wouldn't publish any statistically derived info either. I was asking because we can't assume the relationship (or algo) is linear (in terms of CTR and CPC). So, if you had an exceedingly non-linear relationship (ie. a u shape type thing) and applied a formula for linear regression, you get 0.00.
Same obviously applies to the relationship of impressions and CPC, which is something that we all want to know.
I was more interested in your techniques than your results. Plotting the data can help visually to check for linear/non-linear relationships (along with best fit lines, etc), or statistical tests for linearity.
More impressions, with no corresponding increase in clicks (as when we are viewing our own pages), lowers CTR, of course.
Don't ASA's comments suggest a bad consequence of low CTR? If not a negative impact on our earnings (not necessarily by way of the Smart Pricing mechanism, but by some other unspecified, mysterious, surmised discounting mechanism), what then?
