Benchmarking Banner Ads

A lot of folks talk about eCPMs for Airpush push notification ads and others.

However as I have posted before - my sense is that the eCPMs across formats are meaningless - and a better measure is perhaps comparing to a “benchmark” of the app’s “potential revenue capability”.

Because of the stability of Admob I have suggested using Admob revenue within the same app as a measure of “desirability” of any new ad network you are adding to the app.
Here the variable factor is that Admob while having a full 100% fill rate - does vary in it’s eCPM somewhat - and as demonstrated by their revenue per day slump some days ago there can be seasonal factors (as some suggested it was due to end-of-quarter ad inventory issues perhaps).

The other measure is to compare per-DAU (Daily Active User). Here too the results will have variations - as it depends on the type of app (and it’s user demographic) - and can also be affected by the mix of new/old users for that app (many new users will generally give better revenue vs. jaded returning users).

Now coming to benchmarking between banner ad networks - there has been talk of eCPMs here also - however what is ignored is the fillrate.

A eCPM of $0.5 at 100% fill rate is equivalent in revenue to $1.00 at 50% fill rate or $2.00 at 25% fill rate and $4.00 eCPM at 12% fill rate (in fact using Admob eCPM floor beta does something very similar - the fill rate goes down a lot and thus eCPM better rise accordingly to still give you the same overall revenue).

In some ways a $4.00 eCPM at 12% fill rate is BETTER - in that you are getting “more bang for the buck” - and you have 88% fill rate free to advertise some other banner ad network (like AppBrain etc.).

HOWEVER, in reality this is not entirely linear - sure there maybe some benefit to having some fillrate left over - but the reality is that the effect is non-linear. That is, if you fill that 88% fill rate with AppBrain or another ad network, the revenue from that may increase, but the revenue from the 12% Admob may decrease (because of distracted user attention).

Think of it this way - say you were showing hugely paying $4.00 eCPM ads at 12% fillrate - now you add 88% fillrate with ads that are $0.3 eCPM. You have just SWAMPED the high paying ads with the low paying crap.

In some ways cutting out all ads and just showing a high-paying but low fillrate ad network - is very similar to having a VERY LONG banner refresh rate i.e. more than 120 seconds and more like 1 minute.

In any case, it makes sense that if the user is seeing a more or less static banner image, then they WILL be more surprised when an interstitial ad pops up (i.e. will be more “novel”) - as compared to if they have flashing banner ads for 10 minutes and then an intersitial also appears.

So I have seen (though it was just 1-2 days and may have been a fluke) - that interstitial ad revenue increased when I had setup a banner ad format that was giving low fill rate.

So eCPM for banner ad networks are also not relevant UNLESS you can provide a fillrate figure also (thus effectively saying “how much revenue” can this ad network deliver).

“So eCPM for banner ad networks are also not relevant UNLESS you can provide a fillrate figure also (thus effectively saying “how much revenue” can this ad network deliver).”

Two comments:

  1. The above statement doesn’t make sense – eCPM means the CPM adjusted for fill rate. The CPM (without the “e”) means the gross CPM without adjusting for fill rate.

  2. The most important metric to compare ad networks is ARPU. The ARPU will always be gigantically higher for push and icon ads, no matter what the eCPM, simply because you are monetizing inactive users as well as active users. As such, comparing in-app eCPM’s to “out of app” eCPM’s is definitely a dumb comparison, since the “out of app” ads are monetizing typically 10x -20x more users than the in-app ads due to inactive users. As such, ARPU is the ultimate metric but for some reason people never use it and ad networks never report it :wink:

Yes, I was assuming the eCPM is calculated based on the actual impressions of the ads (and not the requests sent for retreiving the ads). That is, the earning per actual impression of the ad (divided by 1000 to get per thousand or “per mille”).

I meant that there will be a difference in eCPM based on the fillrate because of the AD INVENTORY. That is, any ad network holding ads will have ads willing to pay high revenue per impression (or click) - so there will be a difference from that. In addition, some ads will just be designed better to be more attractive to users, or just by-fluke may match some user demographic sense.

So in the ad inventory for an ad network they will have only so many ads that are likely to give high revenue per impression. So with Admob eCPM floor beta - you can see that when you restrict the eCPM to be higher than $1.50 - the fillrate immediately drops to 8% to 12%.

Now you could just fetch these ads faster (so despite low fill rate you are fetching them more) - but the effective revenue per ad WILL also go up if you show less ads.

So in general if you somehow reduce fill rate (i.e. show same ads for longer and not rotating as much) - you will be getting more likelihood of clicks for the same ad (as opposed to when that ad is shown with 60 other ads).

Regarding your comments on push ads and icon ads - that is the understanding that developers are given (that out-of-app users are monetized - and in the case of push advertising all the active users i.e. ones with the app installed - get push ads). However, as reported here, folks suggest that the Airpush push notification revenue seems to be correlated more with the new downloads - one explanation for that maybe that it merely LOOKS like it is correlated with that, when it is actually scaling with the total active user base (as reported by Google Developer Console).

If so, we should start to see figures from developers quoting NOT the eCPM or the ARPDAU (average revenue per DAU or Daily Active User), but the ARPAU (average revenue per Active User as reported on Google Developer Console). Of course that estimate for Active User should probably measure just the active users for the app version that includes the push notification SDK (and not the earlier versions).

So the question becomes:

  • what is the ARPAU for Airpush push notifications ?

With Icon ads - it is clearly linearly correlated with daily downloads - since StartApp monetizes (or pays the developer) just once - though they may switch to a lower up-front payment and continuing payments from continuing use of user of StartApp icon and browser search etc. However even this will face attrition as subsequent app installations will overwrite your claim to the SDK - and now revenues will flow to the latest SDK installer (in any case there will be contention eventually about who should get a royalty on that use).

For developers who haven’t used StartApp/AppWiz startup-revenue ad networks - they will want to know how much these will improve their earnings - i.e. how Airpush revenue scales compared to Admob banner ads (as a benchmark).

While I’ve seen some income reports which show much higher revenue from Airpush - there are others also which show low revenue (perhaps it is related to the installed base - where Airpush would shine presumably).

The clearest explanation was the one posted by someone who showed the graph of Airpush revenue vs. Banner or other such revenue. And one could see there that there WAS some impact of “Active User” base on the Airpush revenue - as the daily revenue curve was different from the admob etc. revenue curve.

Is that what Airpush users are seeing ? That is, their OLDEST apps (or more precisely the ones with the most Active User base as reported by Google Developer Console) are the ones where Airpush revenue SWAMPS the Admob revenue - while the new apps are the ones where Airpush revenue is anemic compared to the conventional admob etc. revenue ?

The key difference (in terms of scaling with downloads) is that with:

  • Airpush push notifications - the revenue per day SHOULD scale with active user base
  • StartApp/AppWiz icon and browser search ads - the revenue is exactly correlated with new downloads per day

In an ideal world where revenue scales with Active User base, and active user base is increasing - the revenue per day will be ever-increasing (non-linear).

In reality, the Active User base for most apps (that are non-viral) will STAGNATE (as eventually the natural downloads-per day they get from random search etc. - starts to get balanced by the attrition from their existing user base).

So if you draw a graph of Active User base you will see that for most apps, eventually the rise in total Active Users will slack off and will flatten.

If the download per day increases above current levels then the stagnation point for Active Users will rise slightly.

So in a way for a particular app, the Active User number will be proportional to the new downloads number (over 1-month window let’s say).

Even if Airpush revenues were scaling with Active User base - it would look like the Airpush revenu new users but all users with apps still installed i.e. Active Users).

I maybe mistaken in the analysis - so corrections would be appreciated.