Forum Moderators: martinibuster
Just a few days ago, I thought I would easily beat March earnings but now I am not sure. I have very high search CTR (around 65-70%) but the last 5 searches haven't accrued a single click.
I am just hoping for a big click dump. I am not sure if it will come though. I just get the impression that clicks have happened but they have not been recorded.
What statistical value are you referencing Jean. Please explain?
These statements are indicators that the poster is probably trying to infer too much from too little data:
I know lots of folks that talk about samples being statistically significant, but very few that actually know what that means. Jean is saying that this poster wouldn't have enough clicks in a month to draw any conclusions. I think Jean is wrong. That's why I asked.
>>I know this may not sound much traffic, but if you go 2/3 days without clicks then you have had 100 page impressions without a click.
Based on this statement, I would think crick is getting around 150 impressions a day or roughly 4,500 impressions a month. I'm pretty sure that's enough to get a pretty accurate picture of crick's expected monthly CTR. I'm sure I'll get flamed for this, but an approximation could be 1/N^(0.5) for margin of error. For this particular sample, the accuracy (or margin of error) would be plus or minus 1.5%.
No doubt that 100 is too small a sample, but 4,500 should be a good sample. For all of you with advanced degrees in statistics, I'm sure this is not 100% correct, but I don't have crick's data on my desktop to analyze.
I know this may not sound much traffic, but if you go 2/3 days without clicks then you have had 100 page impressions without a click.
Based on this statement, I would think crick is getting around 150 impressions a day
LOL, priceless, I never thought of that. But I think that the poster was probably trying to convey the idea of 2 OR 3 days elapsing, not 66% of a day.
I do agree that a month's data would give you a reasonable picture. And from a practical point of view, who's going to wait more than a month to see how they're doing anyway? But the point is that day-to-day variations are going to be quite large.
Google only records an impression if the ad block loads. If the visitor clicked out before the ads loaded, the visitor had javascript disabled or was running an ad blocker, then those impressions won't count hence the discrepancy.
I think Jean is wrong.
I might very well be wrong. I am no mathematician so it was only a gut feeling.
30 days doesn't "feel" like too little to get the sense of a trend....but 4500 impressions still "feel" like this is not enough.
I have search just to offer my visitors a choice but I get a VERY low number of clicks.
The ads are set up on a dynamic master page [i.e. only one Adsense code placed on the master page ].
I might very well be wrong. I am no mathematician so it was only a gut feeling.
I am a mathematician (though not a statistician) and your gut instinct serves you well.
BillyS, I'm not 'flaming' you, but I think your argument is much more wrong than Jean's gut instinct that you accused of being wrong. The 'margin of error' is virtually irrelevant in trying to interpret real day-to-day changes in Adsense statistics - because it says more about fluctuations in the means of measurement than variations in the actual data.
In fact, my gut instinct is the OP's variations in data are probably to be expected. Crick can put this to the test using the process outlined in my Golden Gate post [webmasterworld.com]
Taking one page as an example, AdSense shows 80 impressions for every 1000 in my stats. Is this normal, or could there be an error?
This type of question comes up frequently in this forum. In no case, as far as I can recall, have we ever ascertained that Google are reporting incorrectly. This is a large discrepancy and there probably is an error somewhere - but it is most likely in your channel/stats setup rather than Google reporting.
My immediate question is whether, with this size of discrepancy, the 1,000 refers to hits rather than pageviews? There are other sources of discrepancy (eg: cached impressions don't appear in URL channels, PST vs local time differences, javascript being disabled, robots calling pages but not ads, etc..) but these discrepancies aren't usually of this size.
If the 1,000 refers to pageviews, then you'll need to do some investigation to identify the source of the discrepancy. One trick is to put a small (blank) gif on the page under the ads and see how often it is called - is that number nearer the Adsense or the stats report?
I wasn't talking about day-to-day changes, I had already mentioned that Jean was correct about that statement. I was addressing the statement concering "not even a month's worth of your data is enough' (paraphrasing).
I'd argue that a month's worth of data (4,500 impressions)is enough to find some meaningful data.
Firstly, thanks for your interest :-)
The counts I refer to are page views. This was for an internal page of sub categories.
I also note that AdSense stats for the homepage are 25% of those showing on the sites stats for page views.
I am using figures for the month, so that a mean average can be sought.
I'll try the blank gif, and see what figures come up for that :-)
I assume that PSA's don't count towards AdSense stats, but I don't seem to be receiving any of those.
I'd argue that a month's worth of data (4,500 impressions)is enough to find some meaningful data
In that statement you might be correct - but statistical significance is not dependent on volume alone so one can't say just from the volume of impressions. More importantly, it depends on what data you are comparing - if you are comparing data from one day to the next (as crick did in part of the opening post) then there are only 31 days in a month (at most), which isn't a lot (though one can still sometimes get 'meaningful data' even from only 31 results).
But whether we are discussing a day, a month or a year, the "margin of error" is still irrelevant, and quoting a figure of 1.5% is misleading. Margin of error is used when you take samples of data, to assess how accurately the sample results reflect the total population results. But Google don't take samples, they count every single click. So, apart from reporting delays, the margin of error for Adsense statistics is, in theory at least, always zero.
One has to look at completely different statistics to decide whether crick's data has meaning or not. With regards the 4,500 impressions, that's less than half a day's data for me, and my experience is that Adsense data is so volatile that for most decisions I need far more data than that. But it does depend on the performance of individual sites, which is why one needs to do the stats in the post I linked to previously to decide whether a lack of clicks is abnormal or not.
The counts I refer to are page views. This was for an internal page of sub categories.I also note that AdSense stats for the homepage are 25% of those showing on the sites stats for page views.
The 25% discrepancy is much closer to what one would expect due to the reasons I stated above.
If the other comparison is page views - ie: 1000 page views in your stats vs 80 impressions in adsense - then it sounds to me as if there is something wrong somewhere. Are you using URL channels? Custom channels are more accurate, so in addition to the gif I'd suggest putting a custom channel on this ad to track it (as well as the URL channel, so you can then compare).
Rest assured that Google will be counting all your clicks in their total - the problem is probably just one of channel and/or stats package reporting discrepancies.
I guess my concern was that the ads are being shown every time the page opens, rather than any concern regarding clicks being accounted for.
As i mentioned b4 in the first example, the code is only placed on a master page, and when someone opens a sub cat page, the master page acts as the 'template', from which other info is added to, and then provided to the viewer.
when someone opens a sub cat page, the master page acts as the 'template', from which other info is added to, and then provided to the viewer.
Does this mean, therefore, that when someone visits your site and opens, say, 10 subcat pages, the adsense ads are only called once and stay on the screen, whilst the user navigates the 10 subcat pages?
If so, that would explain the size of discrepancy, but I'm not sure it is good from your revenue point of view. Eg: the static ads may result in greater ad blindness, and they won't be fine-targeted to the specific page. It may be better to put the ads on the subcat pages and try varying size/position/colour (the latter being dead easy to achieve in the adcode itself) to get them noticed.
Does this mean, therefore, that when someone visits your site and opens, say, 10 subcat pages, the adsense ads are only called once and stay on the screen, whilst the user navigates the 10 subcat pages?
No, sorry, but I probably explained badly. Each page is individual, and each time a new set of ads displays. The code is on the initial subcat master page, but when a user chooses a subcat, the data is transferred as a completely new page relating to that subcat.
However, I have two problems in that area. Firstly, there is extremely limited space to fit in ads, and so I have to stick to skyscrapers to the right. Secondly, becuse there is a menu to the left of every page, the Mediabot often picks up that rather than the list in the subcat central area.
I am thinking of section targeting to ensure that the bot picks up more relevant content:
[ <!-- google_ad_section_start -->
<!-- google_ad_section_end --> ]
without the brackets of course :-)
I wrote to AdSense regarding this on Friday morning. They viewed my site, and sent a very nice and detailed email back on Friday afternoon. I must say that they are extremely helpful on the few occasions I have written to them :-)
I worked out the reason for the difference in stats. What happens is on the subcats pages,people can change the order of listings, such as date, price and make, and although these register on the site logs, AdSense does not show them as a new page, which is rather good as it affects the eCPM and CTR less.
Thanks for your help anyway :-)
Perhaps it's worth noting, for the wider audience, that Google stats are reliable (apart from reporting delays, of which 'click dump' is the most extreme form). If there are discrepancies and we (mistakenly) doubt Google's reporting then we may not find the real cause of discrepancy because our attention is focused in the wrong place. It is more productive, as Eazygoin has done here, to assume Google report correctly and to look elsewhere for the source of discrepancy.
To answer the OP's original question, Google do not miss recording clicks. However, there can be delays in reporting them, and invalid clicks are discounted.
