we recently started to use Google DoubleClick for Publishers (Google DFP), a cloud based ad server offered by the ad network giant himself. As you might expect from our recent articles on serving mobile ads, we didn’t just import all our locally hosted ads and waited what is going to happen. I personally designed a small test to check, how the ad performance running through an extra server is going to be. As the most basic test I could think of, I compared ads from the AdSense network served directly and through Google DFP.
This article is not just naming the results, but is also going to describe the setup in detail, so you will be able to design your own tests. Anyway, you can also jump directly to the numbers at the end of this article.
Why test the Google DFP performance?
Wait, why bother testing Google’s ad server and not just use it? Didn’t the company already earn our trust with AdSense? It did, still there are several reasons why I wanted to test different angles.
1. Bad experience with ad servers and networks
I recently had a bad experience with a passback ad code I gave to another ad network. The passback was just an ad code the network should send, when they were not able to deliver ads themselves. I was even able to set a minimum CPM they had to reach with their ads. Else, the passback code is used.
Pretty great, I thought at the beginning, but after a few days I noticed, the passback was performing just about 10 percent of the income the same ad would generate if served directly. Of course, I know this because of a split test I ran at first. To solve this, I had to limit the frequency cap to 3 ad impressions a day so the fill rate would increase high enough that the passback didn’t really matter.
At the end, this experience made me very cautious about trusting ad servers.
2. Unexpected differences with AdSense banners
A few month ago, when I installed my first local ad serving solution and just started split testing banners, I created different AdSense banners for the same position. I just served these 3 banners randomly on the same placement and surprisingly had different results. I even came up with a theory that might generate you some more income, but I am still testing. It is not easy to “generate” old banners
3. Ask the Google DFP forum
Before starting something new, you should at least search for others opinions. I did so while reading through the great step-by-step introduction of Google DFP and found a few posts where publishers complaint about the bad performance or decreased income from AdSense ads served with Google DFP. The linked post mentioned a lot of possible settings where the same settings don’t seem to apply for everyone.
4. Saving time
As split tests are kind of my thing, I really wanted to test this. But there is also a more practical explanation. Due to tests, different ad codes depending on browser size and optimization we really have a lot of AdSense codes for our word game platforms. Transferring them to an ad server is a lot of work. So there is an economical risk for doing this just on a hunch. Testing might not only save us time in case of a negative result, but also help us to determine if transfering just a few of the ads wouldn’t be a better solution.
Google DFP or AdSense?
Ad server vs. ad network
Let’s start by shortly explaining the difference between Google DFP and AdSense.
Google DoubleClick for Publishers is an ad server in the cloud. Under the name Small Business this service is free for up to 90 million monthly ad impressions. Google DFP does not come with its own ads, but rather provides the infrastructure for you as a publisher to serve ads based on your settings.
Google AdSense on the other hand is an ad network. I would even call it ad auction. So as a publisher, AdSense helps you to find advertisers who want to present their product on your website and sell it to the highest bidder. Almost, because Google uses the information it has about the visitors surfing behaviour to display the ads most likely to be clicked on. The settings regarding the content or number of servings of these ads are rather limited. You are also not able to combine AdSense ads with your own ads or even target a specific advertiser like you would on an ad market place.
Comparing apples and peaches?
Saying that Google DoubleClick for Publishers is an ad server and Google AdSense is an ad network might raise the question of them being too different to compare. Indeen, as a service, they aim into different directions. But remember, that Google DFP is also serving AdSense banners as a fallback. These ads even come from your AdSense account (I suppose, because there is no Google DFP without having an AdSense account).
Serving just AdSense seems like a perfect base line that makes this test reliable and it makes sense to compare banners from Google DFP AdSense passback and native AdSense. As you will see below, you can also combine both services.
Testing with 4 banners
Setting up the test seemed to be easy. Take an AdSense banner code and a Google DFP inventory with the same size and AdSense as a fallback and run each of them with the same number of impressions on the same sites. Unfortunately, the results you get might be misleading.
control ad
As you remember my experience with different AdSense banners I mentioned above, I needed to make sure this is not going to happen again or that I can at least measure the impact of the newly created AdSense unit. In social sciences this is called a control group. This shall draw a base line to check, if other than the expected parameters might apply to a test and influence the results when comparing just two test groups.
In our case, I created a control ad. This is, by all means, a complete identical AdSense banner as the one I already used before. Same size, same channels, same placements, same number of impressions. It has just another name and is younger.
AdSense banner through DFP
Due to my bad experience with banners served as a fallback, I also wanted to know, how reliable Google DFP is serving banner codes. This definitely needs some additional tests with third party codes, but I wanted to cover the basics here. So I added a fourth banner in AdSense and serve the code via DFP. One would expect Google to serve their own ad codes very well, even if AdSense and DoubleClick for Publishers are different services. In this case, the Google AdSense code is handled like any other Creative from third party ad networks.
Setting up the test
The inventory
So there are four banners now running the same amount of impressions on the same pages. This is the inventory:
- The AdSense banner I already ran for a couple of month now providing reliable numbers as a basis.
- The Google DFP banner filled with AdSense as a fallback.
- The control ad to test the impact of the ads age.
- A native AdSense banner served through Google DFP
Banner 1 and 3 are directly served from our website without any external ad server. Banner 2 and 4 run through Google DFP.
Serving the ads by page impression
I decided to run this test based on page impressions. Meaning, that one of the four ads is served randomly on every page view instead of on a user basis. There wasn’t any evidence for me in past tests that would lead me to the assumption that running a user based split test would result in significant different results.
Google DFP better than AdSense?
Before I get to the results, I wanted to sum up the questions I had regarding Google DFP or AdSense:
- How does it feel to use Google DFP over AdSense?
- Is there a significant difference in the way how to serve AdSense banners?
- Is serving native AdSense tags through Google DFP an option?
For now, I didn’t want to go into the details of optimization. I know, there is a chance a lot might say that if I would have fine tuned setting xy and move A to B the results for one of the banners could have gone straight through the roof. Might be, but I wanted to see if Google DFP is worth a shot at all. So I didn’t get into details like speed for now. This will be a topic of one of the future articles.
The bare figures
I ran the test for three weeks. This doesn’t say a lot about the long term, but that is why I created the control ads to see if age matters.
Your eyes probably already took a peek below to the number, but let me say the obvious: the CTR, CPC and RPM of your site will be different. However, this test is not to compare our ad performance with yours, but the different methods of serving AdSense. This might still work for you.
| CTR | CPC | RPM | |
|---|---|---|---|
| conventional and old Google AdSense | 0,19% | 0,22 € | 0,41 € |
| Google DFP with AdSense Fallback | 0,10% | 0,35 € | 0,35 € |
| Google AdSense control ad | 0,20% | 0,21 € | 0,43 € |
| AdSense via Google DFP | 0,22% | 0,20 € | 0,45 € |
Google AdSense significantly better
The numbers clearly state to use any method of serving code from the Google AdSense ad network than to rely on the Google DFP AdSense fallback. The younger Google AdSense ad performing slightly better than the one I already used a long time, but still, the difference is not really significant to me. The same goes with the difference between the two new AdSense banners. Even though we are talking about a couple of hundred thousand ad impressions for each ad, the difference of 2 euro cents in RPM might still be random.
As a result on these numbers, I already deactivated the Google DFP banner with AdSense as a fallback. The other three ads are still running for other tests.
The most interesting on these numbers is that die Google DFP AdSense in fact got the worst click through rate and the best cost per click at the same time. There was no difference in serving this ad and I can only guess the reason for it. Maybe, the AdSense served from Google DFP is not well integrated with the mother ad network and doesn’t benefit from its self optimization. So it randomly shows ads with a high CPC, but since it is not optimized for the site or the visitor, he clicks less. If you have another explanation, please leave a comment.
Use Google DFP and AdSense
So, to go back to my initial questions to this test:
How does it feel to use Google DFP over AdSense?
Is there a significant difference in the way how to serve AdSense banners?
Is serving native AdSense tags through Google DFP an option?
I might answer them all together. The test showed that both services do their core business very well. Google DFP serves ads but is not performing well as an ad network. Google AdSense is a great ad network, but lacks some additional settings.
Since even combining both services results in a great performance, publishers get the benefits from both programs: Combining different programs and serving ads based on a hierachy or frequency cap and having custom channels in AdSense to track and fine tune their banners.
Fine tuning Google DFP
From now on we are going to use Google DFP when there is a more complex setup. When displaying only AdSense banners we will stick with our local setup.
Anyway, this test only answers the basic question whether to use Google DFP or not. There are a lot of settings we still need to go through and see, if we can optimize the results a bit by playing with options like synchroniously/asynchroniously ad serving. Keep in touch for more information.