Forum Moderators: martinibuster
During this test, my CTR went down. Perhaps not surprising, since the new ad unit was smaller. But my EPC went up. Why?
Here are the possibilities as I see them:
a) Co-incidence.
(Unlikely - ok this isn't the most scientific test ever conducted - but I've done others before and this always seems to happen - too many coincidences).
b) My site is so HUGE that when my CTR increases, the supply of ads available in my industry is lower, and hence EPC is lower.
(Errr... yeah - RIGHT! - (I don't think so).
c) A rise in CTR means that the people clicking are less targeted, and don't convert so well. Google measures this with its conversion tool, and reduces my EPC accordingly.
(Possible - though it seems to happen in real time - not at the end of the month or something. I'd have thought they might need a delay to gather a statistically significant sample of conversions...)
d) When your CTR rises, Google takes a greater % of the money, so your EPC falls.
(I hope it's not this - but it really FEELS like it - the way this happens every time).
Summary: I reckon it's C or D, depending on how paranoid I'm feeling on that particular day.
Any other mechanisms that I might have missed?
Does everyone notice this same effect? Anyone get the opposite effect - reliably?
The answer is very simple. Higher paid ads appear in the first slot and less paid ads in the next slots. More ads mean more choices for a visitor to click, but there are more cheaper ads. The same effect can be seen with multiple ad blocks. In the end, it is eCPM that counts.
Because yes - if there were different numbers of ads, that would invalidate the experiment.
Higher paid ads appear in the first slot and less paid ads in the next slots
Not always true. Ads can be ranked higher in the adblock by virtue of better CTR.
I do agree with "coincidence"
Can nobody else reproduce this effect? Just change from one ad unit to another with the same number of ads but a better (or worse) CTR. Or do it with colour schemes - just change that. If your CTR changes significantly, I think you'll see that your EPC will too.
If you change it back and forward enough times - and see a change in EPC every time - that should convince you that coincidence is unlikely...
Maybe you should post the actual size of the ad units. Generally smaller ads mean less available slots. And yes Fclark is correct, but generally it holds true that lower paying ads are towards the bottom.
Plus there are other issues such as page layout.
In a nutshell: there are so many possibilities you should be investigating - don't waste time on Google conspiracies.
Large Rectangle (336 x 280)
Medium Rectangle (300 x 250)
They're both in exactly the same place on the page. Image ads are switched off for both.
Both get 4 ads.
You're right about A/B testing fclark - I do this on other parts of the site - but haven't done it for AdSense yet. I agree, it would make for a more scientific test.
0.1% CTR and $1.00 CPC is the same as 1% CTR at $0.10 CPC.
What am I going to serve up(assuming that I'm optimizing)? Which set of ads makes us the most per 1000 impressions?The new set with the 5% CTR because it gives the highest CPM.
What do you see when you check your stats?
Last week: CTR 0.5% and EPC $1.00
This week: CTR 5% and EPC 0.15.
So you cry why is my EPC going down when CTR is going up? There is no conspiracy here. From Google's point of view, they are choosing between different CTR and EPC and trying to give you the highest combination, but from a users point of view, you see an inverse relation when you check the stats. Of course, Google would like to serve up 5% CTR with $1.00 EPC, but they don't have that option. That is determined by the market place for ad buyers (EPC)and surfers who decide what ads tickle their fancy (CTR).
Again as a publisher, it LOOKS like there is an inverse relation between CTR and EPC but that is the result of Google optimizing sets of ads for you (which is not a conspiracy but is in fact for your benefit).
And of course this is all in a highly theoretical perfect world (which with Google's payouts of late I'm beginning to believe).
That's an interesting analysis--and it shows why publishers should be focusing on CPM instead of obsessing about CTR or EPC. :-)
why publishers should be focusing on CPM instead of obsessing about CTR or EPC. :-)
I agree EFV - we DO focus on CPM. I choose which ads to run based on their CPM performance. But that doesn't mean I ignore the other figures on the reports, or stop trying to understand why they behave the way they do. After all - if I understand them better, perhaps I can increase my CPM :-)
Very interesting theory creepychris - and I think I understand most of what you're saying.
But I still don't understand this:
If one week I'm serving 4 ads on a page in a small rectange, and getting (say) 1% CTR.
And the next week I swap to a large rectange, also with 4 ads, and get (say) 2% CTR...
Why can't Google serve the same 4 ads at the same price - so that my EPC stays the same, and my income doubles?
(Ok - on a one-off basis there are loads of answers - advertisers dropping out, seasonal factors, etc etc - but they don't explain why this can be reproduced consistantly).
The only explanation I've got is that as CTR rises, conversion rates go down for advertisers, and smart pricing adjusts your payout accordingly.
If that IS the explanation, we can conclude that smart pricing isn't a monthly process (as has been hypothesised) - but happens in real time.
So people...CPM flat (monthly averages) since smart pricing introduced or not?
No, I haven't seen that.
For what it's worth, my editorial travel site's variation in CPM (as measured by channel) has followed predictable seasonal trends, with one exception: CPM for one of my city channels has been climbing steadily over the last three months, presumably because of increased competition among advertisers.
For example, I am the search engine and I have some sets of possible ads for your website for the next three weeks:
CTR 1% CPC .11
CTR 2% CPC .05
CTR .9% CPC .12 (available from third week)
CTR .5 CPC .05
The first three sets of ads provide a roughly similar CPM, but the fourth set is clearly inferior.
First week: I give you ad set 1. Your CPM is $1.10.
However, a few advertisers drop out so the second week I give you the next highest set which is number 2 and your CPM is $1.00.
On the third week, a few new advertisers join and this new set of ads I serve up is set number 3 with a CPM of 1.08.
You check your stats and find that EPC has swung wildly with a 50% decrease followed by a 200% increase. Yet CPM has barely changed at all swinging about 10%.
Again it appears that your CPM has been capped and it is quite flat. But this is a result of optimizing over many choices of ads with varying CTR and CPC.
In conclusions, swings in CPM should be much slower than swings in EPC and CTR. But haveing said that, swings in CPM do happen as shocks occur: new advertisers, new publishers, advertisers dropping out, publisher quitting or getting kicked out.
