Their initial implementation was kind of "hacky.” They rebuilt it so it came direct from their back-end. They still want some things like video tracking, and click tracking (their dev team is struggling with this).
The team now looks for the AMP sign, because it’s so much better as a reading experience. It gives an assurance that whatever network you’re using the content will load fast.
They initially prioritized their article content, but have added their live blogs. They currently have no plans to implement non-article pages, and no plan to do category pages. However this may come down the line as the internal engineering roadmap evolves.
They implemented AMP in time for Google’s public launch date in February 2016 (they started developing in October 2015). We had a team of three developers working for roughly four months to build the first iteration.
They focus on reach and engagement as their key success metrics, including pageviews per visitor over 30 days, unique users and pageview volumes. AMP performs well on all metrics – in terms of engagement it doesn’t quite hit the heights of Instant Articles – however, that reflects wider trends between search and social. In terms of reach, AMP has shown positive results; however, there is a bias towards national news sites here, which isn’t seen on other platforms.
As with all the other participants, they saw no change in rankings (other than that resulting from being in the news carousel (which they are in). Now that most mobile news carousels only show AMP results, there would be a big hit to search traffic if they were not included here.
They have open questions about monetization. Are they making more money with AMP? That’s currently not clear.
They can’t share specific numbers on impact but can share that overall mobile traffic has grown significantly.
They did see improvements in metrics overall, but can’t share specifics.
They found the open source approach refreshing and relatively easy to work with. The public development roadmap was a huge help in prioritizing features and development work, and keeping the wider business informed of what is going to be possible in the near future within AMP.
One key area they struggled with was analytics. They use Adobe analytics, with Google Analytics as a backup. Adobe’s native AMP analytics tag does not currently support data on session stitching, new vs. return visitors, or integration with core report suite, so the tracking implementation had to be pixel-based, which also has a limited feature set.
They initially had some issues with the overall ad fill rate, due to the requirement that AMP be implemented on https (the Google CDN is on https). Evidently, there are advertisers that can’t serve ads through https, particularly from a programmatic perspective. However, DFP (Doubleclick for Publishers) and Ad-X are their main advertising platforms – and they say they have AMP support.
A fair proportion of articles have various types of embedded dynamic content. They initially used placeholders for types of content that was not supported, and have worked to integrate the most widely used types over time, and adopt any new formats alongside standard web. For example, for the UK general election, they had an interactive AMP widget for finding your own MP. They did not try to convert all types of dynamic element on legacy pages, as that was not worth the effort.
We had a slow take-up on AMP ads, because all creative needed to be served over HTTPS. As AMP has grown and we were able to give more details to our suppliers we saw an increase in uptake; Ad-x is the main filler for this inventory so all now should deliver over HTTPS. Overall in terms of ROI it is good and getting better with time.
They are happy they did it, but as yet benefits over the standard responsive mobile website are hard to quantify. Implementation was a bit of a struggle, but the good documentation and support available has made it a lot easier than it could have been, and new developers to the project have been pleased with how easy it is to use now it’s more mature as a product.