Forum Moderators: martinibuster
September saw an 11% reduction from August's EPC
October saw a 5% reduction from Sept.'s EPC
November saw a 23% reduction from Oct's EPC
December (1st two days) is seeing a 20% reduction from Nov's EPC
By my calculations there has been a 45% drop in EPC from August.
My site is very broad-based (directory site), so these results are likely not from a single industry's advertisers reducing their budgets, etc.
I would have expected a jump in EPC as we approach the holidays, as advertisers bid for more traffic.
When reading claims of increased EPC on this board, I get concerned -- though I'm sure results are mixed. Even with more than double the amount of traffic I had in Aug/Sept/Oct I'm still seeing less payout.
Any ideas, or is this just a sign of the times?
Mipapage,
Actually I measure the number of alternate ads. The URL emebbed in the adsense code loads a page which is logged.
PSA's cannot be measured directly.
>>by the way, sorry it's been a bad week.<<
No problem, I’m there now as well except I don’t have a bright spot yet. :-(
Your theory looks OK to me. As a matter of fact, I’m impressed, for what that’s worth.
It’s just a question of the sample size and consistency.
>>3-5 days each<<
Make a choice. Base the sample size on the ANSI/ASQ Z1.9 or at least the Mil-Spec. Lets face it, sample size is an entire other science and the smaller the sample, the larger the margin for error. Look at the total hits though, and get a sample. If we are saying that the dollars have dropped over a 3 month period, then we need to look at 3 months of data. Don’t take all the samples from the same time period or days of the week. That would be like trying to get an AOQL on one lot out of 300.
If you want worse case, look at you peek hit times, and I think we can assume that G gets no less than the same percentage. Do the same with day of the week. For example, my peek days are Tue, Wed., and Thur. And peek times are 9AM and 2PM.
I just noticed that with the ‘update’, the peek times are flat.
Automate the calculation without Excel. I’m writing an article on how Excel is not the best stat tool. Some versions have a rounding error larger than what you are getting because some brain at M$ didn’t reset the Floating Point Processor in the VBA DLL’s. You have to call the integer routine using an integer to reset the FP processor to zero.
Visi
>>96.458 % <<
That is terrible, 3.4 sigma. Standard set in the 30’s was 3 sigma and in the 70’s, by Juran was 4 sigma, today, 6 sigma by Motorola, GE, et. al.
The software industry (and if anything web logs are even more useless) is more accustomed to 1.9 sigma.
SN
PS: I must be doing something wrong... some months I have over 60% spider traffic, in the heavy 7 digits on one of my sites...
IMHO all this statistical stuff is overkill anyway. Google appears to be reporting accurately, but because of the way the Internet works it is impossible to get anywhere near 100.00% agreement.
I'm talking about firewalls and ad blockers, time synchronization of reports, odd behaviour of robots and user agents, dropped packets, timeouts and interrupted transfers, obsolete browsers, filtering of impressions or clicks which may occurs at Google's end, etc. etc. etc. etc. Not to mention the browser confusion about what server to query for what file, that I see in my logs every single day. IMO two sites using two different underlying technologies can only hope to see ~95% agreement at best.
Me ...I'm still at the stigma...err...sigma post:)
well richmond...waht would results be if google actually didn't remove robot hitting of pages from the numbers?
From what I've seen they are and if the FAQ is to be believed they should be since an impression is defined as displaying an ad block and I'm not aware of any of the major bots loading the JS. If there are any, I'd love to know which are so that can be taken into account. Spider traffic on the 2 sites I measured is currently 15-25% of total impressions (both sites have several thousand pages, mostly from Google and Inktomi) so I have to agree your findings are suspicous if you think that more than 3.x% of traffic is from bots, which sounds more than reasonable.
I think I'm with your advice earlier in this thread though - I've already spent too much time analyzing, discussing and thinking about it and it's not making me money and it's not bringing me happiness. And unless Google isn't counting all of the clicks (which I have no way of knowing) I suppose it really doesn't matter much. Of more concern to me is why my EPC which was was b/w +/-15% of X daily for the last 10 days dropped 50% after my impressions increased 30-fold over the last 24+ hours. I can think of a number of possibilities, but with no real evidence that makes one jump out I'm waiting for something to stick out.
Then why do I have M$, Intuit, Oracle and a couple of others signed up for my six sigma newsletter? I was taught how to apply six sigma in IT back in 1986 at Motorola, so I know for fact that Motorola IT people use 6 sigma, and I’m pretty darn sure GE and at least Honeywell does as well.
richmondsteve
>>but with no real evidence that makes one jump out I'm waiting for something to stick out.<<
That is why we look at stats first, so that we have some idea where to really start digging.
I only recall one bot reading js before, and I don’t remember now which one, but it snagged one of my URL’s in a .js file.
Clicks or epc's....when clicks are doubling...epc is...ummm...half...when revenue is flat. Looking at the click rates tell me ads accepted by visitors...epc means advertisers paying less...or google is taking more.
You can analyse it to death....but it doesn't help me improve the declining epc that started this thread.
We can only analyse the data we have, and sometimes it is as obvious as the above:) This is especially true if the trend continues on revenue/visit lines that we actually track by, to account for varying page views.
Isolate, analyze, report, and do it all over again for each variable.
Try this scenario on for size...I just started using Adwords. I pay $10x per click. With adsense I receive $60X per click. One in every five of my visitors clicks on an adsense ad (I have very few return visitors due to my niche). This means that I make a profit using Adwords and the traffic Google sends me. It takes a bit of time twisting your head around it, but Google pays ME to send ME traffic. Can't get much sweeter than that!
Dave.
SN
My Clickthrough rate is stable, my Page impressions/day is stable, but the medium pay per click is droping like a stone for about a week.
From an average of 14 cents to 7 cents per click in 7 days :(
Something is wrong.
We must though remember that this is the first December we have all had with AdSense, and there may well be seasonal swings. Although from what I have read this does not look like a seasonal swing more a realignment of EPC.
Google will imho be looking to find the balance whereby they keep the quality sites happy by offering sufficient remuneration and from which they keep the advertisers happy and interested in more. Plus they also need to keep the average sites happy for visuality and numbers.
We have AdSense on many 0000's of pages and add over 40 new pages a day but I do watch the stats, and if revenue goes too low then it would be removed and sales people brought back into that area.
Google gives webmasters say 90% of the money from the clicks to start and get everyone on board.
Google then slowly starts decreasing webmaster percentage to more internet average industry levels. Common marketing tactic, give a taste, get users on board, then do what you wanted to do all along. This spans all business sectors and it is not really very surprising.
Webmasters complain, google says its in the tos and most webmasters do not have enough clout to change the outcome. The only people that could really rock this is microsoft and overture and yahoo. So the Little guy takes what he can get and waits for the next level of competition to take hold e.g. overture.
The only thing that keeps this whole wacky internet world in check is competition, which was nicely eliminated by the google tos for the most part. I know they can do it and its their right to do it, but coming from a small upstart company they should know better than this just from that perspective.
I like google, I think they have a good product, and are givng webmasters with small niches some bucks. I have no problem with that, but terms of service that don't specify a payout, elimination without explanation at a whim (alleged), and lack of useful statistics so that even if we are locked into google to at least let us maximize our income would be fair.
I'm not asking for exact payment amounts, but you could at least let us know what size ad is working with what ctr and on what pages. You don't have to tell us dollar amounts but at least give us a shot at optimizing what we have on our pages. This is really not an unreasonable request.
Tell us the stats so we can at least know where to put the ads.
All that tech power and we get a simple sum of clicks? Impressions that don't by default tell us how many are psa?
Google has been the darling of the industry and we have all rewarded it for its behavior and it in return has rewarded us.
I know you guys have to fight the clickbot crap and unethical people, but the large majority of us are ethical building our sites and doing what we love, please don't treat us like the trash. I really don't want to be afraid when I get an email from google!
Reputation has gotten you in the door, but now action is what's going to speak to the majority of us here.
/rant off
BUT, nobody has any proof that they are doing it now. I have just been doing some research for somebody who has seen a sharp drop in EPC. I have examined his site in some detail, including the Advertisers who appear on his Adsense ads, and identified the keywords that are most likely making him money on Adsense. I have then checked Adwords and Overture and find that those keywords now cost less than 20% of what they did two months ago (One has dropped from $14 to $1.29 on Overture)
Despite this being Christmas month etc there ARE many keywords for which the price has dropped sharply. Whether they were over-priced in the first place is a moot point. But if the advertisers are bidding less then the reasonable assumption is that Adsense payouts will drop even if Google has NOT reduced the % they "share".
I would sometimes get like 2 or 3 clicks and make 50 cents. Now if i get 3 clicks i get 15 cents.
Maybe it has nothing to do with the tax stuff, but it is going down.
Also, on the issue of google impressions not equaling actual stats....one thing I recently noticed is that Google adsense is not displaying on Netscape 4.7 mac version....Netscape 4.x accounted for about 4% of my traffic last month...has anyone noticed adsense not displaying in other browser versions?
Marco:
Respectfully, that would not cause a drop this radical in this short a time span, you don't go from 1 day XX EPC and then .66XX EPC for the same site with the same content and the same advertisers.
I can understand market fluctuations but adwords just are not that liquid. I can see a trend if words start going lower in value but right across a category? I hardly think so.
My site which is publishing not retailing, should be booming with ad money right now, so the CPC should be going up not down from a buyers point of view.
Let me restate this, I don't think that google should not try to maximize revenue, hell its what we are all doing. I am just pointing out from my numbers that a radical shift in epc downward coupled with the fact that this is a SPEND month for advertisers is not something that can be just attributed to competition forces.
In fact those forces should be pushing prices upward rather than downward. What I do see happening is that during this time of year market is doing much better, people are shopping, people are clicking a lot more, and google has determined (my guess not fact) that since this is happening reducing epc is fine because overall income will still go up and will not cause them to lose publishers.
I've had a LOT of years in the business field and I've seen a lot of stuff that most people don't even think about in the biz. With google at the size and market penetration that it has, a shift of 5% is HUGE across the board in profits. Now I'm not saying this is a bad thing, but understanding what is going on is a good thing.
Just imagine if you had a site that generated say for arguments sake 1 cent per click. If you had a million clicks a day at 1 cent that's 10,000 dollars a day. You can extrapolate what a BILLION clicks would be, and what changing a small percentage of that 1 cent would cause.
Respectfully, that would not cause a drop this radical in this short a time span, you don't go from 1 day XX EPC and then .66XX EPC for the same site with the same content and the same advertisers.
Marcus, I hate to break it to you, but that's exactly what my site has done. Google started having a serious problem with ad relevancy on my site in mid-October which I have previously documented on this message board at length. I do not get any PSAs, but I now get a lot of mistargeted ads which have a lower EPC. These ads are targeted to the theme of my site but not to the page on which they appear.
>>this is all off topic, but if you take a cpu produced to six sigma specifications, <<
Your right, your post has nothing to do with percentage of hits, but also, no one was talking about hardware. However, you are apperently failing to figure into your calculation the millions of lines of instruction code in the processor, coprocessor, video processor, and the list goes on, plus the number of times the lines are executed, not to mention the millions of transistors in the CPU. What are you measuring to come up with such statements?
>>CANNOT possibly be still six sigma<<
Of course you have millions of data points from every motherboard made so that you can certify your point, right? But no one was talking about hardware, so what is your point? I don’t know you or what you have done in your life, but, I do know that I qualified tens of thousands of IC chips for Motorola when I worked in Component Engineering, most with at least a million transistors, and have seen them perform using test computers to include at the chips temperature extremes.
>>And you can'T tell my your average PC box assembler has six sigme quality standards<<
Define what an average PC box is. I don’t know what that is. Can you produce the data that you have collected to certify you statement and exacly how one would define ‘average PC box’, or is this all hearsay? Can you define and under what conditions the samples were taken?
>>and you CERTAINLy cannot tell ANYBODY that any kind os software has failuireates<<
OK, just a guess here, you have never had formal six sigma training? Ergo, you really do not know how accurate what you are saying is. Did you know that six sigma is about 90% culture?
From the link…
‘This corresponds to a 1.9 sigma quality level according to how most six sigma people calculate quality. Intranets are at 2.2 sigma.’
‘We'd be wise to adapt some of the six sigma methodologies to aid our quest for improved Web quality.’
It’s because I have yet to meet a webmaster that has had formal six sigma training, not saying that there aren’t any, so I don’t know what this means or how it is relevant. That people without the training cannot meet six sigma. I would agree with that. A lot of webmasters are independent, and could not, or would not, put up the $50,000 to $100,000 into formal six sigma training.
You stated...
'The software industry (and if anything web logs are even more useless) is more accustomed to 1.9 sigma'
The author was referring to web development, not the 'software industry' as you are saying. It is a quantum leap to assume that most development can be related to web development on a one to one ratio, isn't it?
Then, did the people that wrote the NOS have six sigma training? How about the peope that wrote Apache, or whatever? Well, you get the idea. If any one place in any process has not been designed to six sigma, you cannot possible have or get to six sigma, but, that doesn’t change what G had vs what the got now.
In my 1st post the only thing I said was that I noticed a discrepancy in two sets of numbers, but that I DID NOT have enough data to have any confidence that it was a problem, so what is your point?
If you want to know some of the types of software that meet or exceed six sigma, auto-pilots, Patriot missile system, CPUs’ in your car, think about how many millions of times your pistons fire correctly without having points on the engine any longer.
If Adsense is paying way more then you could make selling the same products to your surfer then how is the guy paying for you AND Google going to do better?
