Thanks for replying.
In March, we were in a very good situation in terms of audience (7 to 10% growth MoM in unique visitors). It seemed a great idea to offer our content via AMP to grow faster in mobile searches.
We placed 3 steps to understand the behavior (and they did not work to prevent the risks). First, a week of testing applying HubSpot AMP in 2 pieces of content. Second, 2 weeks of testing applying HubSpot AMP in 12 pieces of content. Because they were blog posts with relatively low mobile visits, there were low impacts (because there was almost 0 organic traffic going to AMP).
The 3rd wave was applying to all blog posts.
Our mistake was the following: we use a customized template for our blog (coded, not using HubSpot template builder). In this situation, we had to place, on the template <head>, the meta tag to show search bots the URL of the AMP version.
We should have rolled back from the AMP implementation for all the blog posts, and started again the tests with a few pieces. But because we did not know that was indeed the reason we had no traffic until then, we did not roll back.
The impact is that we lost the possibility of tracking users who visited our blog in mobile devices. We also lost impressions for our call to actions -- each CTA is the path to an offer, in a standard inbound strategy.
All we could see was a drop in traffic in the exact same share of mobile visits we had before that. And we the number of new leads coming from those offers also dropped, but in a smaller intensity.
We discover some issues (errors) on the AMP implementation (i.e.: our blog is set up in “pt-br”, but the AMP language meta tag was “en”).
After 2 weeks with no data about the AMP traffic and behavior, we finally concluded, with no doubts, via Search Analytics report (from Google Search Console), that our AMP pages were losing positions and traffic and impressions on the SERP.
Then, we decide to roll back.
As far as I understand, the issue was related to the fact that we use a fully customized template on the blog. So, the “conversion” of some of the HTML tags we use into AMP standards did not work well. We use that customized template for other reasons (including some SEO white hat hacks, such as good open graph and schema usages) and it did not make sense to change it to template builder.
To be clear: I am neither a detractor of HubSpot nor a detractor of the AMP feature. But it was very frustrating because we had high expectations on its potential.