Forum Moderators: martinibuster
Over the course of a month my page views show a standard deviation of 19% about the mean.
My clicks show a standard deviation of 17% about the mean.
My CTR shows a standard deviation of 15% about the mean.
However my CPM shows a variation of 52% about the mean.
So my pageviews, clicks and CTR hold pretty constant and show maybe an extreme range of a factor of 2 (i.e. the highest numbers are twice the lowest) and a standard deviation of around 17%. But my CPM shows an extreme range of a factor of 10 (highest value 10x the lowest) and a standard deviation of over 50%.
Any idea why CPM swings so much? I know it depends on the individual value of the ads people click on, but I think I have enough traffic to average out clicking on different ads. The ads are fairly consistant, i.e. by the same people for the same products. Does the value of a click on a particular ad vary widely on a day to day basis and if so why?
I'm just trying to make sense of my data. It's difficult to gauge the effect of changes when the CPM seems to vary randomly over such a wide range. I can see the effect of changes on pageviews, clicks and CTR, but if the CPM is fluctuating so much, the effect of changes on the bottom line are hard to analyze.
Since my earnings correlate most closely with CPM and CPM doesn't seem to correlate well with anything, it makes analysis of the effects on my bottom line due to any changes I make quite difficult!
Big jumps on Impressions & CTR are due to an introduction the second set of ads on some pages.
CPM = (earnings divided by page impressions) x 1,000
Multiplying by 1,000 might exaggerate fluctuations in the two factors (earnings and impressions) that determine CPM.
Maybe it's just me, but my CPMs are all over the place. Rocky's CPMs seem to be a bit more stable.
The effect of variation in CPM to to make it really hard to see how earnings relate to traffic or CTR, at least over short time scales (weeks to months).
The problem is that my best earnings come from days with a high CPM, not days with high pageviews or high CTR. To some extent I can control (or at least attempt to maximize) pageviews and CTR via search engine optimization and suitable ad strategy, but there's nothing I can do about CPM, and it's CPM that's by far the largest factor in determining earnings!
If pageviews and CTR have an indirect (and possibly inverse) effect on CPM (and I've no real evidence if they do or don't), than changes might not end up having the desired effect!
... CPM seems to vary randomly over such a wide range
This may be just one of the cases. Mine is much lower range (21% vs. 52%).
The problem is that my best earnings come from days with a high CPM, not days with high pageviews or high CTR.
Conclusions for my case are just opposite to yours. Since my CTR and CPM are relatively stable, earnings can be improved by increasing the page impressions and no. of clicks. To do so I have to optimize the second set of ads more.
it's CPM that's by far the largest factor in determining earnings!
CPM doesn't determine earnings ... earnings determine CPM. CPM is calculated FROM earnings and page impressions.
... the term standard deviation in statistics is NOT expressed in percentages ...
You're right.
It has a unit.
For example, the standard deviation of earnings/day will be $..../day. Since we do not want to show the real values due to AdSense's TOS, we expessed them in relative terms. That is a reason why BTAS2 used standard deviation about the mean which can be estimated by using [stdev/mean in percentage].
CPM doesn't determine earnings ... earnings determine CPM. CPM is calculated FROM earnings and page impressions
That's very much a "chicken and egg" argument. In fact the data are equivalent. In your terms what I'm saying is that my earnings do not correlate well with either impressions or CTR. They depend much more on the (arbitrary?) decision by Google as to what each click is worth. So while my impressions and CTR stay fairly constant, my earnings fluctuate wildly at times.
As above commented, standard deviations aren't normally given in percentages (though they could be). They have to be here because the Google TOS prohibit the discussion of actual numbers ($earnings).
I guess maybe it's my subject material that gives rise to wildly fluctuating "effective CPM" values. Not everyone seems to see the same thing. Just another Google AsSense mystery!
That's very much a "chicken and egg" argument.
It's not. CPM doesn't determine earnings any more than how much income tax you pay determines your income. There's a correlation between the two, but the causality is strictly one-way. CPM is the cost to advertisers for 1,000 impressions ... which is based on your earnings and number of impressions. CPM is a *reflection* or indicator of how you're doing but not a "factor" in anything. It's a result, not a cause.
Let me rephrase my question in light of this:
Why do my total earnings divided by my total number of clicks (otherwise known as CPM) show a 5x greater daily fluctuation than my total number of clicks do?
I know each click is individually evaluated wrt to earnings, but you'd think that statistically it would even out - and for me it doesn't.
If one day I get X impressions, Y clicks and make $Z and a few days later (same site, no changes)I get X impressions, Y clicks and $Z/10 (and that's what I see), it makes site optimization difficult since I've no idea (and no control over) what's causing the changes in $Z
The cause of this high fluctuation in CPM may be due to CPC as Camper suggested above. To validate this, the stantard deviation around mean for CPC (or Earnings) is also high.
If this is a case.
1. You should set up channels to monitor and analyze the CPC for each ad unit if you're using a multi-ad units.
2. If you have a single ad unit, you may have a mixture of high and low paid ads in the ads shown.
A reduction in no. of ads shown can reduce the CPM fluctuation. However, the bottom line is the total earnings, not CPM.
BTAS: You said "Why do my total earnings divided by my total number of clicks (otherwise known as CPM) show a 5x greater daily fluctuation than my total number of clicks do? "
Total earnings divided by total clicks is NOT CPM.
To FromRocky: You said:
"The cause of this high fluctuation in CPM may be due to CPC as Camper suggested above. To validate this, the stantard deviation around mean for CPC (or Earnings) is also high."
That'd be my bet, but I don't see how you can calculate standard deviations for this, because you don't get (at least thru google) the data to do so. You can't calculate the variation because you don't have the individual "scores". I suppose that you could calculate the standard deviations having each day as a single data point, but you wouldn't be calculating around CPC.
Total earnings divided by total impressions multipled by 1000 is CPM.
The variation in earnings is clearly being driven by CPC, which manifests itself as CPM since my number of impressions remains pretty constant.
I guess my "question"(?) is why, given that I presume I'm getting enough clicks to average out clicks on different ads, does CPC vary so much.
Does the same ad yield (cost) the same CPC thoughout the month or can the CPC on any particular ad - which appears often on the same page - vary by a factor of 10 over a few days. I guess this is Google proprietary information. Whatever it is, it's the major factor in determing earnings on my site.
Does anyone know (guess?) if CPC is a function of total impressions and CTR for a site? Do bigger/smaller sites or sites with high/low CTR see a different CPC, hence a different CPM?
