|Why doesn't AdSense display the BEST ad type?|
| 8:40 pm on Jun 25, 2013 (gmt 0)|
More and more, the AdSense algorithm that determines which ad type to show is favoring the worst performing ad type. How does this make sense? It's the exact OPPOSITE of Adwords and logic in general. Google knows that text ads far exceed the image, rich media but gives 75% of the impressions to image and rich media and only 25% to text. On some ad locations, and I'm always talk about 728x90 ad sizes, AdSense will favor text ads. I'm trying to figure out why in some cases their algorithm favors the best performing ad type, but more and more they display the worst ad type. They are not out of text inventory! This is mind boggling.
| 9:10 pm on Jun 25, 2013 (gmt 0)|
Where are you getting your stats? My numbers look pretty good. I'm aggressive with my use of channels, so I can tell pretty well which placements perform best, and which ad types.
|Google knows that text ads far exceed the image, rich media but gives 75% of the impressions to image and rich media and only 25% to text |
I'm not sure the numbers overall back that up (although they might for your site)
At any rate, there are whole lot of things that come together at the time of a click that influence how much you get paid. I don't think it's possible to know that an ad wasn't the highest paying available ad at the exact moment of a click.
| 9:34 pm on Jun 25, 2013 (gmt 0)|
These are my stats - numbers NOT actuals, but accurate representation
Ad Unit #6
Date Ad Type Ad Requests Ad request CTR Ad request RPM
6/3/13 rich media 5,000 0.25% $0.50
6/3/13 text 1,500 0.95% $1.75
6/3/13 image 4,000 0.50% $0.55
6/3/13 flash 750 0.15% $0.30
Text is clearly the winner and while this example data is 1 day it's always like this. I have 20 different ad units where the best performing ad type are getting low impressions. I do have a few, a small minority, where the text is the best CTR and RPM AND getting almost all the ad type impressions. And there's nothing unique about the ad location, ad size, or type of pages, when AdSense does favor the best performing.
| 2:00 am on Jun 26, 2013 (gmt 0)|
my stats look similar to yours, except that rpm for image/rich media is even worse. if you ask around, there are a few publishers who have good experience with image/rich media, but for most, everything else than text ads just sucks. i have experimented with image/rich media from time to time, but it was underwhelming each time and i have turned them off.
i asked myself the same question as you: why does google give room for image/rich media ads at all in my ad blocks, when they perform so badly?
my interpretation: first of all, the goog algorithm generally has to test out each ad across the network and on your website. in order to check their perfomance they also let image/rich media slip in to a certain degree if you have turned them on. why the degree is so high sometimes (and in your case seems much higher than needed), who knows. this doesn't necessarily noticably affect your bottom line, because.. well, look at your ctr: very few people click on ads anyway, so it doesn't matter much if bad performing ads are mixed in every so and so time.
vice versa: if image/rich media obviously and generally doesn't contribute adequately to your earnings, you could as well disable them. after all, if you exclusively show an ad type which outperforms everything else, this has to be the better solution.
considering your stats and as you say, they are reliable and constant that way, try it out, block image/rich media, show text only and i almost guarantee you that it doesn't make a difference. if anything, your earnings will grow as you only show the best performing ad type every time. seems logical enough for me..
google knows that their display ad network sucks big time for most of the publishers. but they have to promote it for their advertisers and keep it going that way. that's why you will be persuaded at every possible turn to use them (latest example: publisher score card).
text ads and display ads serve a completely different purpose (information vs branding). they need two different accounting schemes (click vs impression). the thing they don't dig since years is that you simply can't let compete two completely different advertising forms in one auction.
| 12:54 pm on Jun 26, 2013 (gmt 0)|
Thanks moTi. When I make ad type text only, the RPM falls off immediately. For example, if the RPM with all ad types on is $1, but text is really $1.75, when I make that ad unit text only the RPM will be about $.70. So less than when all ad types are active and much less than what the text ads do when part of the mix. It didn't used to be this way - this started right after the new year. The site does millions of pageviews of month so this issue is a major concern and opportunity....but I've tested everything and pretty much have to run all ad types to get the best RPM.
| 1:58 pm on Jun 26, 2013 (gmt 0)|
Then you're probably stuck with it. Image ads and rich media ads do pretty well for me *at certain times of the year* - mostly when the vacation and travel places in the areas I cover ratchet up their ads. The rest of the year I've tested, and I just run text (and I don't see a huge drop in RPM) Google may be applying system-wide stats that just don't fit with your site. They're not really in it NOT to make money, so their models must predict that's the best choice.
If you're doing millions of pageviews a month, don't you have a rep you can ask? Or do they not do that anymore? I do millions of pageviews for only two months out of the year, so I never expected a rep.
| 2:20 pm on Jun 26, 2013 (gmt 0)|
Yes I talk with my rep frequently and actually have met with them in person 2 times in the last 6 months, but none of the recommendations they've suggested has increased revenue. We were hit by Nessie, so I feel like what SHOULD work just doesn't for some reason. Actually doing 100 million+ pageviews a month so I think in some ways the best practices kind of go out the window at the volume of the site. And the site is very poorly categorized - means we have a very broad range of products on all pages versus tightly themed pages. That mistake is on us - but would be a huge undertaking to fix. Appreciate everyone's input.
| 4:17 pm on Jun 26, 2013 (gmt 0)|
|For example, if the RPM with all ad types on is $1, but text is really $1.75, when I make that ad unit text only the RPM will be about $.70. |
that's very strange indeed. i can tell you that disabling image/rich media made no difference in earnings for me at all. the reason why i turned it off (again recently this year) was the consideration, that showing flashy banners would distract users from actually reading my content, giving a worse general impression and reduce the amount of repeat visitors long-term. so i don't show image/rich media if i don't have to, that is to say if it doesn't noticably contribute to my earnings.
| 6:06 pm on Jun 26, 2013 (gmt 0)|
This doesn't seem odd to me at all when you consider there is an auction going on.
Think of this example. There is a large inventory of image ads available at $1.00 CPM. There is a low inventory of text ads available at $2.00 CPM. The auction will deliver the $2.00 CPM text ad only when it's available, then when it's not available the image ad @ $1.00 is the best available. This could result in your reporting show more image ads delivered then text ads.
| 8:49 pm on Jun 26, 2013 (gmt 0)|
makes some sense - considering the high amount of pageviews, it could be that the inventory for high performance text ads is dried out and the proportion of image ads explodes. whereas if you have a relatively low amount of pageviews, you never reach the threshold and can safely disable image ads.