Tuning AdSense ads

It isn't a science, but is scientific. And it makes money for you!

Earning money from ads is not a "smack the ads on the blog" proposition. It takes time. It takes research. Hopefully, after reading this article, you'll have a bit more information to help you milk the most out of your own Web sites.

If you're familiar with AdSense, I suggest you skip down to the next section. For everyone else, let's start with an intro.

An intro to AdSense

As you can probably tell, I'm using Google AdSense for the ads in this blog. AdSense is great. Google mostly pays you by the click (and, on some image ads, by impression). Since the gross of your income is generated through clicks, AdSense ads should be tuned to generate the maximum amount of clicks with the impressions your site gets (this is called clickthrough rate).

The recipe for a perfect ad

How can you tell which ads fare the best? You cannot, because you don't get to pick which ads are displayed. However, what you can control is the ad format, color palette and placement in your site. Thus, key to generating the most clicks is to use the most appropriate palette, ad format and placement.

Notice how I didn't say best, but most appropriate. Which colors, format and location depend exclusively upon the circumstances of your site. Can you fit a 200x180 ad in a certain spot? What palette should I use: contrasting (stands out) or blending (blends with your site)? Where on the page should I place the ads?

Fortunately, AdSense does publish some information about it. They tell you right away that:

• The integrated rectangle, 336x280 in size, is the ad that fares best.
• The best position to place the ad is smack mixed with the content.
• Palettes heavily depend on the palette of your site.
• Ads above the fold (on the first screenful of the page) fare dramatically better.

Analyzing which ad fared best

Another thing you get with AdSense is personalized criteria. They are a bit like tags -- you apply them on your ads, and AdSense performance reports can be grouped by them. In practice, this means that you can create a blue on white and a red on yellow ad, tag the first with blueonwhite, the second with redonyellow, and after a couple of days, see which one fared best.

The entire tuning process can be summarized in this image:

Now that you're familiar with AdSense, let's explore how to optimize ads for results.

My criteria system for the past 2 months: fair, but not excellent

Two months ago, I decided I would deploy ads on this site using a set of criteria I thought would be optimal. I'd tag my ads using criteria based on these four ad attributes:

1. ad format
2. color scheme / palette
3. type of page
4. position in page

Dumb move on my part. As it turns out, this criteria system just didn't give me enough information. It just gave me aggregates.

For example, I could tell which color scheme or ad format gave me great clickthrough, but I couldn't consistently tell if this clickthrough was due to the position of the ad, the type of page where the ads were displayed, or the format. In other words, each criteria aggregated a number of unrelated ad impressions.

Alas, it did yield me valuable information. The ad that gave me the most income (82%) was definitely the one embedded in article content -- the first ad you see when you visit an article. I actually expected this, because this ad is entirely above the fold.

Knowing that this ad was a red/black on white 336x280 large rectangle, it was a good starting point. With so many variables to track, it was rather fortunate to find that a single one determined my income so strongly. Having pinned down the most performant ad position, I used this knowledge as a starting point for my next experiment.

Experiment X: make six variants, keep the best

To perform an accurate experiment, you need to keep constant as many variables as you can. Once I knew for a fact what the best-performing ad position was, it was time to start paring down three other variables: alignment, format and palette.

Guaranteeing unbiased data

I didn't want results to be skewed, so, in the name of science, I decided to forsake a few bucks. So, using the AdSense Deluxe plugin bundled in my Turbocharged blogging system, I turned ads off for the entire blog. Since my theme embeds ads through the AdSense Deluxe plugin, I did not have to touch a single line of code.

Generating and tagging the ads

Then, using the AdSense ad creation tool and a bit of CSS (stylesheets) magic, I generated six versions of the base ad. This graph summarizes how I generated the new versions:

The criteria system I used was derived from my old one:

1. Type: always article above content. Combines the fact that these ads appear only on articles (not on static pages, or archive sections), and they are located above the content.
2. Format:
   1. 336x280
   2. 300x250
3. Palette:
   1. R3.0 standard red/black on white
   2. R3.0 with green titles
   3. R3.0 with green borders
4. Code: each ad would have a tag named after the same AdSense Deluxe code, so I can track performance per-ad, on each variant.

After generating the six versions, I saved each one of them in a new AdSense Deluxe slot each.

Rotating the ads

The whole point of the experiment is to give each ad equal exposure. So I wanted each ad to be rotated fairly, in a equally random fashion. Accomplishing rotation wasn't difficult, but it did require some coding. My template includes the ad by calling a function:

I just changed it to:

to have PHP rotate between the different blocks

Scheduling the next doctor's appointment

You know how doctors say take two of these, come back and see me in the morning? Well, after doing all of this legwork I scheduled a checkup on my Palm T|X in 8 days. I'm fully on the wait stage now. Why 8 days? Well:

• I want to have 7 full days of data.
• The amount of clicks I get daily (25~50 clicks) lends itself to fair statistical accuracy after 7 days of data collection.
• Okay, strike the last reason -- just say I'm impatient.

What I'll be looking for in the analysis is simple. I want to identify the winning ad, using the code criteria I defined. That way, I can restart the testing process with variants of that same ad.

Conclusions

Tuning AdSense is a laborious and repetitive process, but it has its rewards. While laborious, it can be done, as long as you follow a methodical, scientific process.

The right approach to doing AdSense tuning is that you're investing in a steady source of income; it's not like earning two glasses of water instead of one -- it's like changing an 1-inch water pipe for a 2-incher. If you have a popular blog, it certainly pays off.

Come back in 8 days and read me again -- I'll be writing about the results of this experiment.