Fill rate went down to 4% from >90%!

[LSparkle]

Hi,

the fillrate of one of our apps was at about 90% and went down to
0-10% about 1 month ago and stayed there since. I tried to contact the
admob-support but they didn´t answer to my 2 inquiries. All our other
apps are still at around 70-100% fillrate. Did anybody encounter a
similar problem? As we implement the ad-banners in all our apps the
same way I would guess that it can´t be because of the ad integration
but as nobody from admob talks to us we just don´t know.
I hope somebody in this group can help us guessing ;-)

Thanks

LSparkle

[Alexey Khripkov]

Exactly same prob, one of my app now have 1%-4% fillrate instead of
80-99%.
But look at your CTR , mine is 100%, it point to error.
And about support, same story. No answer at all.

I write 2 mails about this problem and one mail about problem with
activation my advertise company.
For last mail I receive answer in 2 days.

PS: As for temp solution - you can change your ad refresh interval to
~10 second.

[LSparkle]

Thanks.

Our CTR is at about 25-30% which is clearly too high. Now I changed
the refresh interval to 60 seconds but that changes just the CTR but
not the fill rate. Did changing the refresh interval change your
fillrate?

LSparkle

[Alexey Khripkov]

It can`t change your fill rate , but may increase your real clicks.
You have money not for fillrate but for clicks :)

[Jim]

@LSparkle - changing the refresh rate probably will not impact
fillrate, but might help your CTR which, as Alexey points out, is what
matters.

This seems similar to the discussion here:

http://groups.google.com/group/google-admob-ads-sdk/browse_thread/thread/6eeb11c58e3e96ca#

Have you checked filters? Also, in that thread I ask about "AdMob Ads"
vs. "All Ad Types" - there does seem to be a reporting problem with
those reports. I haven't heard anyone with this problem that does not
run in-house ads (people with in-house ad campaigns seem to be the
only ones that see this).

Maybe that helps?

-Jim

[David]

I reiterate my previous posts.

Have an app that is worthy of receiving ads that people want to place in a relevant app, and you'll have a HUGE fill rate.

What these facts might be saying to you is that Admob by way of the Google transition FINALLY got around to making the algorithms that spider your app description and then MATCH the right ad for your app, have finally KICKED in and what you've really been doing is getting away with MURDER by getting ads, advertisers really don't want in your app and that is now the real cause of your eCPM and subsequent CTR to drop fhru the floor.

I love Prosecutors that shoot first, and ask questions later, never examine the evidence, much more get it admitted into the Court of judgment.

This is akin to children killing their parents and then crying to the Family Court that they are orphans.

: (

As living proof that the Admob fill rate system works and works quite well, we have had 100% fill rate for 11 days straight.

Once again, perhaps it could be attributed to the fact that we have the widest demographic of relevant ad space humanly possible?

Only the Google Gods on top of Mt. Mountain View will ever know the method to the madness.

Perhaps they have the ON set as OFF there too! LOL (that was for you Jim!) wink wink!

[Jim]

David,

I've been thinking about your claim that a popular app will have a
better fillrate. I don't see how that could be the case, but if it is
the case, I would like to know how that works. I am referring to your
claim on this thread as well:

http://groups.google.com/group/google-admob-ads-sdk/browse_thread/thread/6eeb11c58e3e96ca#

Now you have re-posted it, so I'm really curious...

Your recent post suggests that AdMob has created something that will
"spider your app description" so they can better match ads to apps.
So, I guess you are referring to AdMob doing a statistical analysis of
some kind, like Google search uses algorithms to match searches with
appropriate results. Google search is text based, like the app
description, and therefore the indexing algorithms are conducive basic
database structures (that's how Oracle made all it's money, after all
- capitalizing on the development of relational databases). But
matching ads to users is a combination of user interest, ad visual
appeal, ad placement and tactical activities that cannot easily be
indexed and put into a database. Using a publisher's self-generated
description of an app as a basis for automated ad placement does not
sound like an effective approach. Also, knowing what the description
says, even if it is accurate, does not provide any information about
the ad. So it requires a similarly biased, advertiser-generated
description of an ad to use for matching purposes.

As an example of using automated indexing for ad placement, an
advertiser's ad may not work inside of a certain app because of visual
elements. Those elements subsequently change and make it "pop" (or
whatever) and then it is very effective. You can't index that. And
besides using very esoteric "genetic programming" or other pseudo-
heuristic, pattern-matching algorithms or exceptionally large, multi-
variate testing with huge assumptions (I think they have enough going
on that they are not experimenting with such things), there are not
standard tools or algorithms available to correlate app popularity,
app functionality and user behavior with advertiser goals, ad
targeting and ad design for the purpose of matching ads to apps based
on app descriptions, app popularity, app design, app functionality and
ad descriptions, ad design, ad relevancy and ad appeal. It appears to
me that they are simply providing advertisers and publishers with
tools to automate ad publishing - but the system is far from simple,
and therefore the data we collect is not easy to interpret.

If the spider is not in place, but advertisers manually control it,
I'm trying to discover by what means advertisers would filter out
"unpopular" apps. I believe they have similar options to publishers
when it comes to ad placement; they have very limited control. I am
not aware of how AdMob might provide app popularity measures to
advertisers in order for them to eliminate "unpopular" apps (number of
installs? requests? devices?). And I don't think advertisers in
general would be so naive as to assume that the definition of
"popular" should be "it's popular if I've heard of it." I don't think
advertisers get information about our apps individually, just like we
don't get individual ad information.

Also, it seems to be AdMob's strategy to place generalized ads for
wide audiences based on app categorization, much like a television
station is "targeted" to a viewer. Unpopular TV stations still have
the same number of ads, but they can't charge the premium prices
because viewership is low. Most of AdMob's ad inventory is not
targeted toward "premium" apps (they are a "blind" ad network), so ads
are placed based app categories selected by advertisers and set by app
publishers. Then ad control diminishes rapidly for both parties; or
becomes excessively difficult to manage.

Also, I'm not sure I see how your "popular" app would get a better
fillrate compared to my "Kalamazoo" app with AdMob advertisers even if
it is "the widest demographic app on the planet." I can't see where
advertisers with AdMob would conclude that users of "Super Popular
App" are better than "Kalamazoo" users. A wider demographic of users
would only establish that an app has broad appeal or is well marketed.
Why would they assume a low usage app delivers ads to a user group
that is unappealing? The app may not be generally well-known or even
appealing, but that doesn't reveal anything about the users of the
app.

They could use location-based advertising to include "Kalamazoo."
Location information for AdMob would need to be enabled and requested
by the publisher. But I don't think AdMob advertisers eliminate apps
based on location information (e.g. I don't think the advertiser have
the option to avoid apps with users located in Kalamazoo). It's
inclusive, in that more ads are available to an app when user location
data is provided.

In limited cases, an advertiser may want to claim that they had ads
placed with very specific, very popular apps. But, that sounds more
like bragging rights than advertiser strategy. Why would "Super
Popular App" have users that are better for the advertiser than
"Kalamazoo?" Kalamazoo would get fewer ad requests, but not lower
quality users for an advertiser simply due to its low popularity
status.

Popular television networks can demand premium prices because of the
large number of viewers reached and, in some cases, a target
demographic. App ad placement is not broadcasting (it's tracked and
paid for one ad impression/click at a time on identifiable devices),
so ad placement with a popular app does not increase exposure - only
placing large numbers of ads without repetition to a single user
increases exposure. And a larger variety of apps can serve to increase
exposure regardless of app popularity. Limiting the number and variety
of publishers would actually have the potential to reduce exposure.

And broader demographics reduces premium priced ad placement - popular
television stations charge premiums for popular shows with broad
demographics by having high viewership. With this analogy, I could
actually see an advertiser request low usage apps because they might
feel that it has the potential to reach people that their other
advertising efforts might fail to reach.

Advertisers not seeking broad exposure but rather niche markets would
want to match the target audience of the ad to the app's user profile.
AdMob does not survey our users and we can't do anything more than
categorize our apps and provide a basic description of our app (we
don't do any app profiling or user profiling for AdMob). So, ads
placed with AdMob are not like ads placed with Mobile Theory or
NAVTEQ. High-cost, premium ads are not relevant for most apps; the
cost to establish and maintain a targeted ad campaign with highly
focused apps and user profiling is not feasible. And this type of
advertising excludes all popular games/apps because popularity
requires general audience appeal, not a niche market user group.

I just don't see how fillrate is related to app popularity with
respect to ads provided by an ad network like AdMob. And most of us
should not spend the time and energy for more targeted premium
advertising, that might pay better in theory, because the reality of
the high risk and expense involved with developing time-consuming
relationships with advertisers that may never spend money on our apps.
We are unlikely to succeed in making more money with that approach.
(Similarly, most advertisers don't have the time or energy to seek out
and maintain relationships with large numbers of publishers, and then
build systems to deliver mobile app ad campaigns. So they sign-up with
AdMob and get all that in one package.)

I can see where AdMob may give preference to popular apps in order to
"sell" their services to advertisers (i.e. when talking to an
advertiser, it's easier for AdMob to make a sale using the line, "see
your ad in Super Popular App" compared to "see your ad on Kalamazoo
Montana Bear Patrol"). But that seems more like an easy AdMob sales
pitch than anything to do with fillrate.

Also as evidence, networks seem to have a lack of app ad placement
awareness (within AdMob and across the industry). I have seen several
recent posts here and elsewhere that complain that users see an ad
that they have already clicked on, users are over-exposed to specific
ads or age-inappropriate ads are displayed (which information would
probably be in the publisher's app description). That sounds like a
very fundamental filter set. But if that is hard to do/is not in place/
is not working properly, it is hard to imagine that a far more complex
"super-system" like what you describe, that reduces fillrate, matches
popular ads to popular apps, etc. and simultaneously over-exposes an
ad or serves inappropriate ads to a child's game, is in place at this
time.

Or maybe the system is too complex - and the "Ad Saturation" option is
"Off" because it looks like it prevents ad over-exposure, but setting
it to "off" actually disables the Ad Saturation Limiter feature
resulting in over-exposed ads... so the super-system is in place, but
AdMob messed up the settings!

hmmm... I'm trying, but I don't think I can convince myself of that. I
don't think the fillrate issue is explained by that line of thinking.
I think we have to keep working on this one to figure it out. Of
course, if you're right about the super-system, I hope someone from
AdMob chimes in to let all of us know what factors impact fillrate for
the system, at least in general terms. It's possible we might have a
few more questions, too.

-Jim

On Oct 17, 3:41 am, David <casinogaminggr...@gmail.com> wrote:
> I reiterate my previous posts.
>
> Have an app that is worthy of receiving ads that people want to place in a relevant app, and you'll have a HUGE fill rate.  
>
> What these facts might be saying to you is that Admob by way of the Google transition FINALLY got around to making the algorithms that spider your app description and then MATCH the right ad for your app, have finally KICKED in and what you've really been doing is getting away with MURDER by getting ads, advertisers really don't want in your app and that is now the real cause of your eCPM and subsequent CTR to drop fhru the floor.
>
> I love Prosecutors that shoot first, and ask questions later, never examine the evidence, much more get it admitted into the Court of judgment.
>
> This is akin to children killing their parents and then crying to the Family Court that they are orphans.
>
> : (
>
> As living proof that the Admob fill rate system works and works quite well, we have had 100% fill rate for 11 days straight.
>
> Once again, perhaps it could be attributed to the fact that we have the widest demographic of relevant ad space humanly possible?
>
> Only the Google Gods on top of Mt. Mountain View will ever know the method to the madness.
>
> Perhaps they have the ON set as OFF there too!  LOL (that was for you Jim!)  wink wink!
>
> On Oct 17, 2011, at 1:25 AM, Jim wrote:
>
>
>
>
>
>
>
> > @LSparkle - changing the refresh rate probably will not impact
> > fillrate, but might help your CTR which, as Alexey points out, is what
> > matters.
>
> > This seems similar to the discussion here:
>

> >http://groups.google.com/group/google-admob-ads-sdk/browse_thread/thr...

[David]

Jim

First and foremost, I wonder how many ad networks you work with other than Admob?

I am engaged with approximately 28.  They include text ads, banner ads, call to action ads (phone dial numbers), rich media, immersion ads and video ads.

i also come from the world of music, television and film where arbitron, nielson and box office sales determine everything from a radio ad, a tv ad and a big screen pre-movie ads are sold.

You will find that although what you have written below is incredibly well thought out, it is missing some massive controlling factors.

You can have 10 million children tune into SpongeBob orange and the sync fee for an hour is 1x because it is at Saturday morning at 8am and the buyer is a child who has no buying power

You can have 10 million femaie adults at a demo of 25-49 watching Seinfeld for an hour and the sync fee is 10x because it is Thursday night primetime at 9pm and the buyer is the most coveted Procter & Gamble or BMW buyer.

Identical quantity of users, with identical hour time slot, DAY AND NIGHT ad targets, ad demos and ad prices.

The advertisers pay based on so many more factors that what you have listed below.

Now neither one of us run Admob.  Neither one of us run Google, but one thing can be assured, they have more patented algorithms that control it all than Carter's got peanuts.

One thing that is undeniable in virtually every ad network is that they analyze your app description and run it through algorithms that determine what it is and then apply more algorithms to decide what ads are going to get delivered and why.

Read this article below to see just how perfect this science is becoming.  Everything comes down to math at the beginning and the end of every day, each day, predictable human behavior is getting closer and closer to being mapped into a proof theorem, despite that fact that free will defies the morality of reason on that concept. It is a reality.

http://www.nytimes.com/interactive/2011/08/20/business/20110820-is-that-review-a-fake.html

You see, the ad relevance by networks are analyzed right down the color of your app backgrounds.  The more information you provide to the ad network in their SDK that you pass through, the better ads you can get, and the higher fill rates.

The "popularity" that I speak of is a bird spotting app vs. a word game app.  That universal level of popularity that is spidered through the algorithms that simply look into the iTunes connect store (if you're on Apple) and the read of your category (which has been verified and trusted) and the content of words in your app description.

This is a science that is a science of sciences that makes this stuff work.

Although I highly doubt that Admob will chime in on this, as they would be forced to publicly reveal trade secrets that would violate their own protocols to answer and reveal PRECISELY how they do things and how their algorithms work.  Not a snowball's chance in hell of getting the truthful answer.  It's like Eric Schmidt testifying to Congress a few weeks ago.  Despite talking for hours, he said nothing of any consequence, didn't admit, didn't deny anything.  He simply proffered, "uh huh" yeah, we employ algorithms that find the best ad for the best publisher.  Have a nice day.

Run a website that is about business enterprise apps and I can promise you that Google (adChoices) is NOT delivering a FLIRT APP.  

Deliver an ad request to a mobile device that is a game that 25-49 age females play, and you'll get match.com ads, cupid.com ads, flirt.com, skout.com ads, you name it.  That's their demo.  MIddle aged women that are divorced with children.  You won't be getting an ad selling SaaS products.

The algorithms are so smart it's almost inconceivable.

In most ad networks we deal with, you even have a ratings system as to the completedness of your app description and you are required to post screenshots of your app before you can get ads.  Millennial Media is one of them for example.

If you are only 60% complete, you're going to get a completely different fill rate than you would if you're 100% complete and then further factored into the algorithm content of the contextual information you are forced to give.

So, with all due respect, and you already know, I do respect your opinion, it is not entirely completely accurate because it appears that you haven't factored a myriad of other variables that go into the master algorithm (that is a formula of algorithms) that = a result.

I promise you the ads you get and I get are different if you have the Kalamazoo Bear Spotting app and I have Words with Friends and all of my ads are Admob ads, just as much as I do on a website that is about technical software programming vs. TMZ.com

: )