27 October 2014

Responsive re-design damages ad revenue

openmicfinder.co.uk started as a desktop-only site, then it gained a mobile version hosted on the "m." subdomain, recently I replaced both with 1 responsive version. Most interesting is the unexpected negative affect this has had on advertising revenue compared to the positive affect it has had on the Google ranking. Obviously the new responsive site scores massively higher when run through Google's PageSpeed Insight tool. I've minified all the right bits, leveraged browser caching and of course it is legible and delivers a great user experience across all screen sizes, including the tabletty ones in the middle which traditionally got left out by the mobile/desktop version split. As Google are clamping down on desktop-only sites by adding mobile-friendliness to the pagerank algorithm I had no choice but to move to a responsive redesign which, I am sad to announce, has cut my earnings by a very painful 50%!

So to summarise the key differences before and after the change:

  • Before: Most users saw a site with 3 MPU adverts. Apart from those on really small screens who got a pop-up suggestion to view the mobile site on the m. subdomain which had 1 mobile banner.
  • After the redesign: Only the biggest screens get the 3 MPUs, notebooks get a leaderboard and a banner, tablets get 2 banners, mobiles get 2 banners. On all screen sizes the ad units are now beautifully in-lined with the content and proportionately sized for the user's screen.

In essence I have followed all Google's "advice" (read pressure) and coincidentally happened to slightly increase the number of adverts that each user will see. Naturally I expected revenue to increase with usability but in fact it is much worse.

I have not discovered a definite reason why, so all my current thinking is theoretical, but it basically looks like MPUs are measured as performing better, even if the user's screen is too small to display the site and pinch-zooming is required to navigate (as would have been the case for about 50% of users before the redesign). This could be because Google Adsense charge more for showing an MPU (I have no way of looking up their prices), it could be because media agencies for some reason focus on producing the best performing artwork only for  MPUs and leaderboards and neglect the new sizes which I have introduced such as banners and mobile banners. It cannot be coincidence, the graph shows a distinct and sustaining drop immediately after the launch.

There is one other potential cause that is harder to measure. The redesign also involved me applying the latest standard of moving the Javascript to the bottom of the document, a method not supported in the Google Adsense integration documentation.  I felt this was appropriate because it enabled me to query the user's screen width to calculate which ad-units are being shown by the CSS media queries and do the responsible thing of only loading the ad units which will actually be seen by a human. A lazy developer would have stuck with Adsense's recommended method, which would have resulted in ad server requests hits for all ad units, regardless of visibility, a behavior which directly contravenes Google Adsense's terms of service.

It is sad to think that the only way I will be able to test this theory is to build a branch in my code repo that temporarily - but deliberately - breaks Adsense's rules and it's even more sad to think that the outcome might incentivise me to keep it that way permanently. Adding to my woe is the thought that an alternative outcome might be that I have to move the Adsense chunk of JS back up to the document head and block rendering until the ad server responds, a process which would feel very much like moving backwards, especially with the great adsense blackout of 2014 that brought down millions of websites.

The necessary evil of adverts just got uglier.