The core implementation was pretty straightforward, and they worked closely with Google on it. End to end it was a 6 to 7 day effort plus 3 to 4 more days for the home page. The implementation was completed in May 2017. See a video on their experience with implementing AMP.
The reason Myntra chose product list pages is because most of the traffic lands on these pages first and the bounce rate for these pages was high. On its site, the product search listing pages/merchandising pages represents 2/3 of its SEO/SEM traffic. The home page is responsible for another 20% of traffic. The company has not converted additional pages on their site.
Instead of implementing the rest of the site in AMP, Myntra implements them as a progressive web app (PWA). PWAs are an elegant next-step solution because they have a component called a service worker, and the service worker operates in the background and preloads pages (such as individual product pages or special offer pages) before the user requests them.
As a result, the user gets a great experience on initial page load because of AMP, and then again on the next pages because they have been preloaded by the PWA.
Conversions on the site are sales. As with all of the other publishers, Myntra saw no indication or rankings changes due to AMP (and they are not part of the news carousel).
Overall page load time was reduced by 65%, and the bounce rate decreased by 40%. Mobile revenue contribution is higher than before, and it’s continuing to move up. Once the full session stitching solution is in place, the company will have a better measurement of the overall improvement to mobile revenue.
Myntra’s PWAMP implementation was launched in the last week of September. Initial stats indicated an improvement in engagement metrics, and an increase in the average session length and the number of pages viewed per session. There has also been a significant increase in the number of product pages viewed per user since launch.
It’s hard to attribute any increase in revenue/revenue contribution directly to PWAMP, but the engagement metrics mentioned above can be attributed to our PWA and AMP investments.
One place where Myntra had challenges was with analytics, because when they started, there was no guidance on how to handle the session stitching issue. However, it participated in the beta project to resolve that issue with Google.
The AMP-bind component is a big deal, because it provides support for dynamic page elements, including sorting and filtering, which is critical for eCommerce. There are still some limitations in AMP-bind, such as fetching content real time from different systems. This is still being worked on. Myntra wants to do more dynamic pages as AMP-bind.
Myntra noted that AMP is evolving rapidly, and that it’s no longer just for static pages. It was used initially for category pages and entry pages to improve the top of the funnel, and a PWA implementation for the rest of the site, but the company is now considering converting to PWAMP (PWA plus AMP = PWAMP) for the listing pages. It believes that this will take a couple of weeks at most to implement.
Myntra is happy with the ROI on its AMP implementation, as it was not that much effort. It works well with all kinds of networks, and results in a huge difference for the page load time which has favorable impact on bounce rates, consumer experience, and conversions.