A couple weeks ago, I noticed that this site and Life of Brian were serving up nothing but Big Brother and Big Sister ads. This happened more or less at the same time that I experimented with putting some Adfly links around the site.

It turns out the problem wasn’t some kind of AdSense / AdFly rivalry at all. It was me, making a few changes to other advertising. I had been using ContextWeb (previously AdsDaq) to serve up some ads on these two websites.

The way the service worked, you got paid per impression. You set a PPM price (a bid, basically), and if there was an advertiser willing to pay it then the AdsDaq / ContextWeb ad showed. Otherwise, it defaulted to a “backup ad” – in this case, AdSense.

Well, these weren’t performing very well, so I turned them off. Apparently, when I paused the ads in the ContextWeb dashboard, that killed the script altogether. Instead of defaulting to the backup ads (AdSense), it just showed those public service ads. Doh…

I went back and put the AdSense ad scripts into the theme directly, and regular AdSense ads starting showing up again. It’s nice to know there’s no problem; it’s a bit annoying that it took me two weeks to figure out I was the problem.

The solution, it turns out, is very simple.

Why Randomize Ads?

There are a few reasons you might want to do this.

If you’re looking into changing the theme of your AdSense ad, you may want to test the two themes against each other. By giving each theme approximately 50% of the impressions for a given time period, you can make some direct comparisons between their performance.

Or maybe you and a friend both run a website, so you want to split the advertising time between two AdSense publisher ids. That can be done too.

How the AdSense Code Works

Before we randomize it, let’s take a look at how the AdSense code works.

Here’s the snippet of code that creates the skyscraper on the left side of this page.

<script type="text/javascript"><!--
google_ad_client = "pub-2399151883698113";
/* Web Cash Tall Skyscraper */
google_ad_slot = "8702750734";
google_ad_width = 160;
google_ad_height = 600;
//-->
</script>
<script type="text/javascript"
src="http://pagead2.googlesyndication.com/pagead/show_ads.js">
</script>

You’ll notice two major pieces to this code. The first piece – in the first <script> element – defines the ad unit’s properties. It sets google_ad_client (your publisher id), google_ad_slot (the saved ad unit’s id with color info), google_ad_width, and google_ad_height.

The second section – the script element with src “show_ads.js” – actually displays the ad based on the variables that were set previously.

In order to vary the type of ad shown, we need to come up with a conditional statement to vary the google_ad_client/slot/width/height settings.

The Javascript Code…

And here’s the moment you’ve all been waiting for – the Javascript code to randomize which ad is shown.

<script type="text/javascript">
<!--
if (Math.random() > 0.5) {
  google_ad_client = "pub-2399151883698113";
  google_ad_slot = "5396010225";
  google_ad_width = 250;
  google_ad_height = 250;
} else {
  google_ad_client = "pub-2399151883698113";
  google_ad_slot = "3900536874";
  google_ad_width = 250;
  google_ad_height = 250;
}
//-->
</script>
<script type="text/javascript"
src="http://pagead2.googlesyndication.com/pagead/show_ads.js">
</script>

You’ll notice that there’s an if/else statement, and each branch of the conditional statement defines a different ad to display. The results of that conditional statement rely on the outcome of Math.random().

Previously, we built a function to get a useful random number out of Math.random(). This is actually a situation where we can use the value from Math.random() just the way it is.

Math.random() generates a random value between 0 and 1. In other words, it gives you a probability. By checking that Math.random() is greater than 0.5, the statement should evaluate to true approximately 50% of the time.

We could also make the rotation more lop-sided. For example (Math.random() > 0.70) would be true 30% of the time. Therefore the first ad would be displayed 30% of the time, while the second ad would be displayed 70% of the time.

I hope you weren’t looking for something more complicated because, well, it isn’t. It’s just that simple.

If you operate a website (like this blog), Google Adsense is probably one of your main streams of revenue. Although you may eventually outgrow Adsense, chances are you’ll have used it at some point.

While looking through my Adsense stats, I noticed an odd trend. This trend suggested a few things about monetizing a site through Adsense – questions that I don’t have an answer for yet, but that I will be looking for.

On this and my other site (Babel), I use a few main types of ads – a horizontal image banner, a skyscraper with text and images, and square boxes with text and images.

It’s natural to wonder, “Which ad is the most efficient and most effective?”

Both of these sites are relatively new, so I don’t have a lot of data to work with. From the data that I do have, though, I’ve noticed a few key trends.

• The tall vertical skyscraper and the square blocks seem to have the best click-through rate
• The horizontal image banner (like the one in the header) seems to have a low click-through rate
• The tall vertical skyscraper and the square blocks seem to have lower pay-per-click values
• The horizontal image banner seems to have a much higher payout than the other two options

Like I said, I don’t have a whole lot of data to support these trends at the moment. Over the next month or two, I plan on watching these trends and testing them out (using random tests of text verse image ads).

It’s something to keep in mind, though. The horizontal image banner seems to be the most effective – the high payout makes up for the low click-through rate. Although I wouldn’t get rid of the other ads, it’s a good justification for making sure you have at least one horizontal image banner on your site.

Notice any trends in your own Adsense reports? I’d be interested to hear what other people have come up.