How much of a scam is AdMob?
alex_slobodnik
]
We ran a campaign on AdMob to have sense of how many people would show
interest for SpeakAdvisor (see my previous post for what it does).
We only targeted Android devices as a start. Yesterday, we supposedly
got 391 clicks on AdMob but Analytics only show 124 pageviews on the
landing page we were using.
Sure, Analytics can be a bit off, but that's a huge difference (only a
third of clicks are tracked?). And I don't see many people disabling
javascript on their Android browser.
So, in your opinion: what's going wrong here ? Is AdMob just a scam ?
See the original post and screenshots here:
https://plus.google.com/101564269619147838281/posts/4DBrPsUir2D
SPA Support
ends at 5PM PDT right now - Analytics will continue to midnight PDT),
and traffic for 8/25 on AdMob would include the evening of 8/24 on
Analytics. Does that help explain the data discrepancy?
2. Also, you posted at 6 AM on 8/26. I have seen Analytics not have
the previous day's data fully updated for a day or two. Have you
checked it since then? Has the Analytics data changed since your post?
I suspect the date difference will play a large role (since evenings
are usually busy app times and the evening of 8/24 for Analytics is
included in the "day" of 8/25 for AdMob.
-Jim
--
Thanks,
SPA Support
Jim
which the events occur. Not much happens from midnight Pacific time
through the international date line, though. Also, that means a 5pm
pageview in Hong Kong appears to happen at the same time as a 5pm
pageview in New York, even though the pageviews actually occur 12
hours apart.
I think AdMob does all their data according to GMT (clicks that happen
right now are recorded right now with a GMT timestamp - regardless of
the event timezone). It's very difficult to pin down how all the
moving parts impact each other because of those differences, among
others...
-Jim
On Aug 31, 1:03 pm, alex_slobodnik <alex.slobod...@gmail.com> wrote:
> [ fromhttps://plus.google.com/101564269619147838281/posts/4DBrPsUir2D
alex_slobodnik
Thank you for the quick response.
Setting aside the question of why Google would calculate dates
differently for two related products, the main question is this:
whether or not Google internally has a mechanism to reconcile the data
from AdMob with the data from Analytics? That is, for every click
AdMob registers, does Google check that Analytics registers the
corresponding legitimate pageview triggered by that particular click
through AdMob?
Alex
On Aug 31, 11:02 pm, Jim <secondphonea...@gmail.com> wrote:
> Actually, I think Analytics posts data according to the timezone in
> which the events occur. Not much happens from midnight Pacific time
> through the international date line, though. Also, that means a 5pm
> pageview in Hong Kong appears to happen at the same time as a 5pm
> pageview in New York, even though the pageviews actually occur 12
> hours apart.
>
> I think AdMob does all their data according to GMT (clicks that happen
> right now are recorded right now with a GMT timestamp - regardless of
> the event timezone). It's very difficult to pin down how all the
> moving parts impact each other because of those differences, among
> others...
>
> -Jim
>
Benoit Curdy
I've checked the Analytics data for August 26th: there are 8 visits
(It then goes down to 0 as expected). No move for the 25th August. So
the difference can't be explained by delays in counting visit or time
zone difference.
SPA Support
US (all timezones), the evening of the 24th is after midnight GMT on
the 25th (Greenwich, England). So far, you have excluded all US
evening traffic from the comparison. Also, your average Time on Page
seems very high to me - is it to you? Just looking for potential
sources of problems...
AdMob is a recent acquisition of Google. It is a fact that AdMob and
Google have not yet completing system integration. Even so,to my
knowledge, they do not have a method of reconciliation (a direct
module or tracking mechanism) for either AdMob or the old Google
Adsense. I'm not sure it would work, though - too many moving parts.
After accounting for the timezone issue, you still have other issues
like:
What if there is an error running the javascript for Google Analytics
(GA) on the webpage - like an out of memory error? What if you don't
send the GA data packet (URL request) before the person leaves the
page (i.e. the ad caused them to click, but the very title of the
webpage is not what they expected, so they back out)? What if GA
receives the "hit" but the URL has an error? Or a GA server processing
error occurs? GA does not pay you, so errors are less important... in
fact, you might pay them for the service, in which case you expect it
to be accurate. But, 400 events is a very low number when they are
processing billions of bits of data. A undetected problem could easily
arise. Have you been able to validate the GA data gathering process
using another test scenario? In other words, you are questioning
whether they "system" works, but are you sure that your testing
methodology and implementation works?
Our Android apps use GA and we track over two million pageviews and
events daily. Only about 60% of our "visits" have "events" - but they
should all have events. So, we are getting a pageview, but no GA
events on pages where GA events are coded? Strange... maybe that's
similar to your problem. Anyway, we also get app errors stating that
Google pageviews/events are not sent due to processing errors. We have
not yet been able to determine the extent to which it happens, but it
appears to be significant. In other words, the device does not
transmit GA data even though it is supposed to.
The biggest problem with that level of tracking, though, would be that
you, as an advertiser, would probably get hammered by us publishers if
you had a bad ad. That is, if you wanted to "experiment" with an ad
and it turned out to be a dud, we would tear you apart. We would
notice that it's bad and cut you from our apps but we wouldn't know
if/when you changed it (not until we allow it again - but why would we
do that? if your ad was bad, would we really want to just randomly let
it back on our apps?). It would be like having a store that only sells
the "best selling products" - a store like that would have no
inventory most of the time, but when it did, it would make money!
Why would we want to run low-paying, unappealing ads? The level of pay
is not subject to opinion, but ads have a very subjective "appeal"
aspect. Ad departments/companies spend a lot of time and energy
"experimenting" with ads to get the best ones. Do you really want
large companies with deep pockets pushing out the small guy because
they can afford to with better paying, professionally developed ads?
I'm sure publishers would get more creative and figure out a system
that might not be as bad as what I'm saying, but it would still cost a
lot more for advertising creativity. It would certainly be a lot more
expensive for an advertiser to try out an ad if the information were
that transparent and low-cost. This cost differential would impact
small advertisers the most.
Regardless of my opinion - maybe that is the best way to do this, and
the tools simply aren't there yet.
-Jim
--
Thanks,
SPA Support
alex_slobodnik
That's a very thorough response and I would look into each of your
possible causes in detail and then respond, I will need the time to do
that. And thank you for the explicit acknowledgment that AdMob is not
integrated with GA in any way.
So far, after reading your text several times, unfortunately, I can
not say that I see how all the reasons you mentioned can account for
the difference that bad: 391 on AdMob vs 124 on GA.
Most importantly, Google has the numbers for AdMob/GA comparisions on
a very large scale. Why are not these public? If the same x3+
difference exists for millions of clicks, Google has to come out and
acknowledge that.
Alex
SPA Support
the differences. I expect a large portion of pageviews to have
occurred on the 24th for GA and were included in AdMob on the 25th,
though.
As for Google "acknowledging" a discrepancy - they would first need it
measured, which requires the integration that you are looking for.
Without that, the discrepancy could be elsewhere in the system (in
your code or mine, for example, as third parties). The biggest one,
for example, is how "dispatch" is handled in GA. That's where we have
experienced a lot of unexpected results. The dispatch doesn't trigger
as we expect, so the pageview is not counted (not sending the data
packet).
Good luck - hope you can nail this down...
-Jim
--
Thanks,
SPA Support
alex_slobodnik
check for discrepancy?
GA encodes referrer data, right? So Google knows which GA pageviews
came through AdMob. Total number of such pageviews should be roughly
equal to total number of AdMob clicks (no x3 times difference).
There shouldn't be any difficulty in running map-reduce computation to
check this is true for all or some subset of domains. Thus, no
integration required in order to ensure GA/AdMob coherency.
SPA Support
Having been asked many times, and having attempted to capture such
data, it is unreliable. Firewalls, network software, etc. can strip
it, modify it or remove it. And mobile devices have issues on this
matter overall.
(http://searchengineland.com/rip-referrer-data-how-mobile-apps-can-kill-your-mobile-metrics-79982)
Furthermore, if that were a known mechanism for monitoring this type
of activity, spoofing referral data is easy and GA would suddenly have
even less reliable data. (I don't know why an advertiser would want to
make it look like more clicks occurred than actually did, but I could
see a competitor having a field day with that one; it would suck dry
advertiser budgets with virtually no results, making them look like
they can't deliver or, as you are worried about, they are "stealing"
revenue.) Stripping it is more common, and I would suspect that AdMob
strips it for security reasons, such as the above (more likely,
though, spoofing URL referrals could invalidate AdMob tracking
mechanisms based on invalid/excessive requests that could occur
through error, spyware or web bots). If they don't strip it, then what
purpose does it serve them? Does the client need to know the referral
source? Perhaps in some cases...
Integration is the only way to know that valid clicks result in
corresponding pageviews. Perhaps they could encode referrer data to
support this use, but that takes time.
-Jim
On Thu, Sep 1, 2011 at 1:20 PM, alex_slobodnik <alex.sl...@gmail.com> wrote:
> Just to make sure I understand, why would Google need integration to
> check for discrepancy? GA encodes referrer data, right? So Google
> knows which GA pageviews came through AdMob. Total number of such
> pageviews should be roughly equal to total number of AdMob clicks (no
> x3 times difference). How difficult is it to run map-reduce
> computation to check this is true for some subset of domains?
>
>
>
> of GA hits through AdMob for fairly large subset of domains and then
> compare that number with the corresponding AdMob numbers.
--
Thanks,
SPA Support
alex_slobodnik
in addition to referrer?
On Sep 1, 11:35 pm, SPA Support <secondphonea...@gmail.com> wrote:
> Integration is necessary because referrer data is spotty, at best.
> Having been asked many times, and having attempted to capture such
> data, it is unreliable. Firewalls, network software, etc. can strip
> it, modify it or remove it. And mobile devices have issues on this
> matter overall.
>
> read more »
SPA Support
exposed IP is not consistent across even closely spaced requests, one
of the frustrations of stateless mobile networking (see here:
http://research.microsoft.com/en-us/um/people/maheshba/papers/ephemera-imc09.pdf).
TCP/IP is stateful, so the carrier keeps track of the client IP, but
HTTP is stateless, so web servers don't need consistent IP addresses
(or stateful connections, which is a huge benefit in many other ways),
and carriers take advantage of that.
So, determining a request source (e.g. an AdMob click vs. an embedded
web link vs. a bookmark vs. an app embedded link, etc.) even from a
single client is not reliable based on referrer, IP address data does
not help because the server does not get the true client IP and the
exposed IP changes rapidly. So, trying to collect and decipher the
data for small groups of clients/domains would be difficult at best,
and then to use that data to test the validity of other data sets
(e.g. AdMob clicks) for large groups becomes even less reliable.
You've got some good thoughts on how to test this stuff. I know we
would all like some third-party verification that the "system" works
and is valid. Keep digging and maybe we'll figure something out...
Also, any progress on your original concern? It seemed like a pretty
good test if the webpage was specifically for the AdMob ads and the GA
implementation could be relied on.
-Jim
--
Thanks,
SPA Support
jfieres
I'd prepare a new landing page solely for each campaign ( no other
links in the web point there).
Then, inspecting the server logs will give you an upper bound of the
clicks ( if there are singnificantly more reported clicks than page
requests something is wrong).
If you want to get a more reliable estimation include looking at the
user agent to filter out all robots and non-targeted devices.
After the previous discussions, using anything more sophisticated
than this (cookies, referrer, jscript, ...) seems to make the results
difficult if not impossible to interpret.
Best regards
On 2 Sep., 18:59, SPA Support <secondphonea...@gmail.com> wrote:
> Cell phone client IP address data is not exposed to servers and the
> exposed IP is not consistent across even closely spaced requests, one
>
> Erfahren Sie mehr »
alex_slobodnik
check for discrepancies. Google doesn't have access to server logs,
only to the client side data.
alex_slobodnik
SPA Support
has a great solution. did you look at your server logs? Also, did GA
from 8/24 help?
The fundamental issue is that even a GA solution would still leave
some people skeptical, regardless of the implementation hurdles. The
only ways to feel completely confident in AdMob are self verification
and/or third-party verification.
Additionally, excessive transparency leaves AdMob vulnerable to
competitors and overreaction (both advertisers and publishers seem
quick to move when errors occur). The industry is so new that there
are not established benchmarks or clear expectations.
We will all keep searching for a while, I'm sure...
-Jim
--
Thanks,
SPA Support