Publishers hate low RPM
Ask a thousand publishers, what are their top priorities? You’ll only hear quality content and high RPMs (Revenue per thousand impression). Had display advertising not been invented, sites we love like Search Engine LandSmashing Magazine, Search Engine Journal and even Wired might not have even existed. But over the last fifteen years, display advertising has witnessed a saddening trend – The constant decline of display advertising CTR. The average Click-through Rate has recently hit .10% and even worse, in countries like Finland, it is 0.05%.

Publishers work hard and in most of the cases, their only source of income is the advertising revenues.

Our Offer
Last month, we offered free ad revenue optimization services to AdSense publishers on Google’s Official – Learn with Google for Publishers Community. The primary aim was to use AdPushup, which is currently in private beta mode, to optimize these websites and help them earn more revenue from their existing traffic. We had originally planned to enroll only 5 websites, but seeing the large number of applications, we ended up enrolling about twice that number.

So, we have Market Validation
The astonishing response to a simple post in the community lounge clearly underlined the fact that there exists a ripe market for our product – a lot of publishers who want to optimise their ad revenues. Seeing the exchange of emails, it was clearly visible that everyone was interested in knowing more about what we planned on doing for their website and how it could help them improve their ad revenues. Surprisingly, a large majority, had never used A/B Testing (many had not even heard of it) or any similar optimization model in the past, which is why, we have now decided to make Customer Education one of the primary objectives for this blog (more on this at end of this post).

The Optimization
So we started off with our Optimization process. Our product is still in development, so stability (specifically experiment delivery) was a concern. We did not want to make any changes that could lead to errors and ultimately, hurt the revenue of the accounts. This meant that we had to quickly hack up a small parallel program, which would do the revenue measuring, dynamic placement of ads in the test locations, and exploration/exploitation of ad variations.

For this we followed a two stage approach. First, for every account, we identified the key positions and ad variations which would potentially lead to CTR & RPM rise (for some accounts the variations went as high as 50!). After creating these positions and ad variations, we installed the program on the website, and let the program run and collect data for some period. The length of this period depended on how much traffic the site was getting. Once we received enough data (statistically significant) about the ad variations & location performance, we analyzed the data, and fed it to our continuous optimization algorithm. The framework and algorithm isn’t perfect, but still it led to immediate hikes in CTR and RPMs.

A golden rule in optimization says don’t trust your gut, and we completely second that! Sometimes, the variations & positions we thought would perform best, failed miserably. On the other hand, the ones we thought would be bummers had excelled! It was great that we had a lot of variations to test, as you can never simply have a hunch about what works well for your site.

Voila! The Results
Now, to the Interesting part, we feel pleased to announce that every single account which we have worked with has reported a significant rise in overall income and RPM (Revenue per 1000 impressions). What do we mean by significant? At least 30%, yea – and thats on the lower side, while most of the accounts have easily doubled their ad revenues!

We’re pulling up a couple of accounts to show you how publishers have been benefited using AdPushup:

Account 1
For about a week our script collected data on this website, after which you can clearly see how the CTR skyrocketed. The CTR rose by 107% as compared to the average CTR of the account for the last 3 months. The RPM for this account rose by 48%.

Account Screenshot

Account 2
The account witnessed a 96% rise in RPM as compared to the average RPM of last 3 months. Similarly, the CTR rose by 98%, in comparison to the average CTR of last 3 months.

Account Screenshot

Additional Notes
In both the cases, the number of ad units per page (before and during optimization) were kept constant. Similarly, no short term techniques (such as placing all the units above the fold) were used during the tests, so we’re confident that improvements are sustainable. The charts which you see above have been edited to remove important account stats to meet the AdSense program guidelines about now disclosing key account statistics.

Case studies
In order to educate more people on the subject of optimization and A/B testing, starting this month, we would start publishing case studies on this blog. We will feature websites which have optimized their ad revenues and show you how they did it! I’m sure a lot of other publishers would benefit from this. You can subscribe to our RSS feed or join our Beta list to get notified, when we publish the case studies.