Google AdSense experiment

Google AdSense experiment

Updated: 09/12/2019.

In this post I’ll write about my experimenting with Google AdSense, with a very brief explanation of what AdSense is. Note: I’m not some AdSense (or website monetization) expert, but here I’ve provided metrics (information) and explanation of how I’ve measured. Convinced that the info as such (objective) can help both experts and amateurs – it is certainly valuable to me. Because I’ve found too many guess work and hear-say on the Internet (not just on this topic).

Contents:

  1. What is Google AdSense?
  2. How to apply
  3. My initial setup of AdSense ads
  4. AdSense experiment
  5. First impressions – September
  6. Second month of testing – October
  7. Preliminary conclusion
  8. November stats – back to discrete ad placement


1. What is Google AdSense?

Shortly: it is a program for showing adverts on a website. Companies pay Google for advertising, Google chooses which ads are shown on which website and gives a part of that commission to the website owners. Commission depends on the type and number of shown ads, as well as on whether the ads are clicked on by website visitors.


2. How to apply

Go to Google AdSense page and register. You will need a Gmail account and a website with enough good quality, unique content.

After providing all the needed data, you will get a snail mail letter, with a code – which you will enter to AdSense account once you receive it, in order to confirm you have provided a real address.

In this post I won’t be explaining the technique of placing ads on the website – perhaps some other time (though I have explained how to ad them to AMP pages (what are AMP pages) in the post about implementing AMP with WordPress websites).


3. My initial setup of AdSense ads

I originally placed ads on the right hand side column, below all the menu options.

These ads were invisible for visitors using mobile phones (only after reading an entire article would they get to the ads section).

At the start of 2019 I added a small banner on top, that is also shown on top to visitors with mobile phones. At about the same time, I implemented AMP page versions, with ads included.

An interesting thing about AMP pages is they are “immune” to ad blockers, so ads are always shown on them. Downside is that Google chooses where to place those ads on AMP pages (which is good for earnings, but you can’t design the layout).

Here is how the ad layout looks like:

Page ad placement: Top banner, Side menu ads and Matched content, showing a mix of website links and adverts
Page ad placement: Top banner, Side menu ads and Matched content, showing a mix of website links and adverts
Picture 1

This drastically increased earnings from ads, much more than the rise in the number of visitors. Graph of monthly ad earnings change from month to month:

Earnings graph after introducing AMP and top banner ads
Earnings graph after introducing AMP and top banner ads
Picture 2

Overview of the ad revenue that each ad position has made in a month, relative to other ad positions:

Earnings by ad position / type
Earnings by ad position / type
Picture 3

It is interesting to note that the cycling website in English, with most visitors from USA and UK, earns a lot more money per visitor (RPM) than the cycling website in Serbocroatian, where most visitors come from former Yugoslavia. While this website, with computer related topics, makes even more revenue per one visitor:

Page RPM per website: io.bikegremlin.com (computers) bike.bikegremlin.com (cycling in English) www.bikegremlin.com (cycling in Serbocroatian)
Page RPM per website:
io.bikegremlin.com (computers)
bike.bikegremlin.com (cycling in English)
www.bikegremlin.com (cycling in Serbocroatian)
Picture 4


4. AdSense experiment

The plan is to, starting from September 2019, let Google automatically place ads where it thinks it is best to place them, then see if there are any significant revenue changes. The experiment will be let to run for two months. I will be following the changes in both revenue and the number of monthly visitors.

Google automatic add experiments and placement optimization
Google automatic add experiments and placement optimization
Picture 5

I’ll also be visually checking what it looks like (for me) on the pages: is it too crowded with ads, or are there even fewer of them.


5. First impressions – September

Subjective impression is there are more adverts on websites. What bothers me the most (what I hate) are the ads inserted between paragraphs of a page. As for the metrics, I haven’t noticed a drop in the number of visits, nor shorter duration of visits. Will be monitoring this, as well as whether Google, after the experimenting is done, reduces the number of adverts, keeping only a few that pay best, in places where they give the best results.

Normalized stats for RPM and total website income compared for August and September (the experiment was started on September 1st):

AdSense RPM per website for August and September
AdSense RPM per website for August and September
Picture 6
AdSense earinings per website for August and September
AdSense earinings per website for August and September
Picture 7

RPM of io.bikegremlin.com leads to a conclusion that adverts on computer websites are paid a lot more than ads on cycling websites.

The fact that the English version of the cycling website makes over 10 times more revenue than the Serbocroatian version, while it has only 5 times more visitors, leads to a conclusion that the ads served to visitors from the US and EU pay a lot more than those served to the visitors from Serbia, Croatia, Bosnia and the other countries from the region.

In those terms, which is not shown on the graphs, practically all the income generated by this website (io.bikegremlin.com) comes from the US visitors. If the structure of visitors changes in the future, so that majority come from Serbia, it should, based on the above stated stats and conclusions, drastically decrease website’s RPM.

Will update this article after the October is through with new stats.


6. Second month of testing – October

In the second month of testing I added a variation to the experiment. I.e. added another experiment. In Google AdSense options, I set so that every second visitor sees only 65% of the available ads – those that are most relevant and most highly paid. As the graphs below show, this resulted in a symbolic drop in total earnings. Only the computer website (this one) saw a drastic drop in both earnings and RPM.

Google AdSense RPM stats for September and October of 2019
Google AdSense RPM stats for September and October of 2019
Picture 8
Google AdSense total earnings stats for September and October of 2019
Google AdSense total earnings stats for September and October of 2019
Picture 9
Google AdSense ad balance experiment
Google AdSense ad balance experiment
Picture 10


7. Preliminary conclusion

Allowing Google to place ads where it “thinks” it’s best resulted in the following:

  • Increased total earnings by about 20%. As promised (“let Google handle it and see the income maxed out” is what they practically claim, to paraphrase it).
  • A lot more ads displayed on pages (and across the screen on mobile phones). Even when it’s been set to show only 65% of the most relevant ads.
  • Google didn’t (noticeably) decrease the number and frequency of shown ads during the two months of testing, no matter what setup I chose.

While I think it’s fair to get a few dollars for the good quality content I provide (for free), I don’t like when there are too many adverts on the site. Especially over the screen and in between paragraphs on a page. I find it a bit annoying and distracting at the same time.

So I returned the ads setup to the previous settings – in the corners, discretely (see picture 1). In November and December, I’ll keep tracking the stats for RPM and total earnings. Also, in November, I’ll be experimenting with showing only 75% of the most relevant ads – to see what difference in earnings it makes, with this more discreet ad setup, compared to showing 90% of the available ads.


8. November stats – back to discrete ad placement

With the start of November, I reverted the ad placement to the old way – so that ads fill the unused corners of pages and don’t get between article paragraphs, or “jump” over the entire screen on mobile.

As expected, this resulted in about 25 % decrease in revenue. A difference that is noticeable, probably important to those who make a living off the ads, or to whom the ad revenue presents an important source of income. Since my website is a non-profit one, the ads serve just to cover the expenses, I put more value into keeping the site clean and informative, not crowded with ads. That is why, after the two-month experiment’s conclusion, I decided to switch the ads back to the old setting.

Share...

Leave a Comment