Admob Network has low fill rate as Mediation Placement

[Adrian Sulumberchean]

Hi,

I'm using Mediation with Admob  eCPM floor, Mobfox and Admob  Network. I am using country targeting for eCPM floor and Mobfox and I've set the Admob network to an eCPM $0.00 without country targeting. By doing this, I've expected to see a fill rate close to 100% but I'm experiencing a fill rate of 50-60 % for Admob Network. Why the fill rate close to 100%?

Regards, 

Adrian

[Adrian Sulumberchean]

Sorry for my typo, my question was:  Why the fill rate isn't close to 100%?

[Eric Leichtenschlag]

Do you happen to have a lot of traffic in countries with lower internet performance, such as India?

AdMob normally gives itself 60 seconds to fetch an ad. But under mediation constraints, AdMob Mediation times out a network after a period of about 5 seconds or so, even if that mediated network is requesting from AdMob. That gives other networks in your mediation flow a shot at the request and we can maintain a 60s request time. So in areas with poor internet connectivity and high latency, AdMob may come back with an ad after 10 seconds, but AdMob mediation already timed out the AdMob network for taking too long, and tried the next network.

[Nizam Mohideen]

Will mediation time out a network after 5 seconds even if there was only one network in the mediation? Does this mean that if I just wrap AdMob network with a dummy mediation, there is a possibility of drop in fill rate if there are network delays?

Regards,

Nizam

[Eric Leichtenschlag]

Hi Nizam,

Yes, has historically been the case. However, AdMob recently started returning an ad in response to a request for a mediation config if it is first in your network configuration. So you shouldn't see drop anymore if you just have AdMob in your mediation flow.

Thanks,
Eric

[Nizam Mohideen]

Hi Eric,

Thanks for the input. I'm on SDK 6.3.1 and don't see this behavior. My mediation contains only AdMob, but in the logcat I can see two sets of "adRequestUrlHtml" / "Received ad url" entries and also an entry about caching gWhirl config. (Is there any documentation on how to debug mediations?)

My ad performance has become worse after switching to mediation. No significant changes in app functionality, except for switching to mediation. I now have close to half my users on the old version and the other half on the new version. My eCPM has dropped by 60% on the mediated version, along with a similar drop in CTR.

I'm also seeing inconsistencies in the reported data on AdMob:

Mediation report says 40% fill rate.

I'm using a dedicated publisher id for the mediated AdMob network. The site report says 100% fill rate.

There is also a small change in the overall request total in both the reports

How can these two reports be different?

Regards,

Nizam

[Eric Leichtenschlag]

Hmm, we may only be returning the ad directly on the first request then for those who have upgraded to the new AdMob.

When using mediation, we time out an ad network after 5 seconds. We treat our own network the same as any other network. So what you're seeing is mediation timing out and reporting no fill, and AdMob's adapter coming back after 8 or 10 seconds with an ad, but it's too late. Mediation considers it a no fill (and no ad is shown), but AdMob returned the ad and recorded an impression. You must have lots of traffic coming from an area with poor connectivity.

I'll bring up the conversation again internally about what we can do to raise that 5s timeout.

Thanks,
Eric

[Bhola Ram Meena]

Hi Eric,
 We are also facing this issue as most of our traffic come from India. Is there any workaround to increase the 5s timeout. Fill rate for Admob and House Ads is around 10% in Mediation but without Mediation it is about 100%.
  
  We have to turn off admob in Mediation despite it being better paying. Please escalate as it is important issue for us.


Thanks
Bhola Meena
HinKhoj apps

[Eric Leichtenschlag]

I'm not sure when exactly it was changed, but this timeout should now be 10s per network, which should help mitigate this problem.

Thanks,
Eric