The implementation was done with their internal engineering team. NoBroker implemented the AMP framework in less than three weeks with just two engineers. They converted all their listings pages which are indexed, with the approximate number being around 250,000 pages. That should be all the pages that are being indexed by Google. They did not convert their user profile pages, as they did not see a need for that.
NoBroker’s goal is to drive “connections” between renters and flat owners to drive more revenue. Their model converts to a paid subscription after nine connections. On average, it takes about 10-15 connections before a renter signs a lease. For this to work, they need to make renters aware of attractive properties, and faster pages helps them do that more quickly.
The NoBroker site is dynamic, with available inventory changing all the time (new properties being added, and rented properties getting removed). For this to work in AMP, NoBroker had to use the AMP-live-list component to fetch new content and modify the last-cached version of the page.
They haven’t tried to use the new AMP-bind component yet, but they may use it in the future. They believe that it will be pretty useful.
They were not able to implement pagination in AMP. As a result, users are not able to view more properties on the AMP pages, and they show only 10. They could have made the number larger, but with 10, that was good for AMP. However, they implemented a Progressive Web App (PWA) approach to solve that, so a user can click on a View-All button to get the rest of the listings.
PWAs have a component called a service worker, and it pre-loads content so by the time the user clicks on the View-All button the page load is already done, so the user’s access to the additional listings is almost instantaneous.
They are not considering a PWAMP (PWA plus AMP) implementation, as things are working quite well as they are. Users get introduced to NoBroker via AMP pages, and then they get the Single Page App (PWA), so they are still quite fast at that point. They are using an AMP for AdWords implementation.
For NoBroker, a conversion is when a user registers on the site when they like a property and want to get the owner details, or when an owner posts a property with them.
As with all the other publishers that are not in the news carousel, they saw no changes in organic ranking as a result of their implementation of AMP. The percentage of organic search that goes to AMP pages is around 20%.
Metrics from the Ampproject.org Case Study: “After our AMP pages went live,” Gupta says, “we saw an 18% decrease in bounce rate, 10% more page views, a 25% increase in registrations, and most importantly, a 77% increase in the number of connections between renters and tenants.” Remember that the connections between renters and tenants is the key to their business. Gupta also shared that they also saw a 10% increase in sessions and session duration.
Because NoBroker’s business model relies so much on the mobile experience, AMP is critical to the continued success of the business. “If you need to provide a super-fast experience to a mobile customer, then AMP is the easiest way to do it,” notes Gupta. “The best part is that it doesn’t require any changes to your existing site architecture, and it’s super-quick to implement.”
They track users on Google Analytics and with a homegrown analytics module. They use Google Analytics for high-level channel data. The homegrown analytics is used to handle attribution to certain channels. They believe that this provides them with much better attribution.
They are familiar with the session stitching problem in analytics, but they have not implemented the solution to that. As a result, they see these visits as two sessions. They were not bothered too much by this as they knew about the issue.
One of the issues was that they were the earliest adopter in Southeast Asia. In fact, they were implementing AMP prior to Google India supporting AMP.
They have issues understanding what AMP pages are indexed or not. Not all the pages show up, and they don’t know why. Even Search Console does not provide data on what’s indexed vs. not. The site: command does not help. Gupta says that “Google should invest in this and come up with a solution.”
They are quite happy with their implementation, and the metrics shared above show very clear, strong results.