We scaled a Bubble app to millions of users

Can you build an app on Bubble.io with millions of users?

As Bubble Gold Partners, we’re among the biggest advocates for the technology.

But that kind of question stops us in our tracks. The answer is quite unsatisfying, “Yes and No”.

Eric, our client and new owner of Folkd.com came to us with a proposition. He wanted us to rebuild Folkd, a niche social media platform with millions of users, using Bubble.

To be honest, our first answer was a definitive “No”. From our vantage point, an app with millions of concurrent users could never work on Bubble. However, Eric convinced us to at least try. Since the site had been abandoned and abused in the past, we had a lot of work cut out for us.

Folkd: a brief history

First of all, let’s give you some background about what Folkd does and how it got to this point.

Folkd.com is social bookmarking website launched in 2006, competing at the time with Delicious and Digg (who is old enough to remember those?!). Social bookmarking got taken over by other social networks like Twitter. Bookmarking links did not require a dedicated website anymore. And Folkd seemed dead in the water.

Folkd never got any significant updates but was still a popular website. With a domain authority of 79, it makes folkd.com the 15,000th most powerful website on the internet. It was put up for sale in 2022 and bought by our client.

Folkd has about 90 milion pages and more than 65 million bookmarks, each with its own page. Despite being stuck in 2006, it still had 500.000 people sign up every year.

As it is basically impossible to cold start a social network today, these numbers are very attractive!

But Folkd was in dire need of a facelift.

Dissecting the challenge

In general, a social network is probably the hardest type of app to build. While simple at its core, the UX is incredibly complex.

Despite the previous owners’ general apathy towards Folkd, they were still getting around 500,000 signups per year. With those numbers, we needed to build an app that accommodated high volumes of users immediately. This was not an MVP.

Another thing about social networks is that for the most part, nobody needs to use them. The expectations around User Experience (UX) are through the roof. Good design is part of how you convince users to keep coming back.

Just think about your feed in any social network app. The work behind the scenes to serve you just enough good content to make you keep scrolling for more is mind-blowing. And there are plenty of other little elements that we’ve all grown accustomed to. For instance, many button-clicks are animated.

And that doesn’t even cover the technical aspect. We needed to allow millions of users to post content in real-time, with moderation, tracking, and tagging. Then, we needed to enable interactions between users and posts, allowing users to reply, share, and like.

You can see why we were reluctant to take on such a challenge.

How we rebuilt Folkd

We design and develop using no-code and low-code technology. We specialise specifically in Bubble.io. We know very well what types of applications fit Bubble and which do not. A social network is not one of those app types.

This is not something you would build on a no-code platform. Especially for Folkd where we are migrating millions of users and posts. This is not an MVP.

We evaluated different solutions like Bubble, Bubble + a third party backend, WeWeb and a third party backend. We eventually settled on the combination: Bubble + Xano + Coalias.

Bubble gives us the flexibility and speed to design a pixel-perfect, modern application. It comes with a lot of complex functionality out of the box like user management, signups, login, and password management. Bubble's backend did not prove to be fast enough to manage importing millions of records so we opted for Xano.

Xano takes care of all the heavy lifting. From post creation, feed algorithms for different user interests, post and user interaction to behind-the-scenes stuff like content moderation, ranking updates and payment management. Even transactional email.

Serving millions of pages to Google

Coalias is a service that we used to manage the enormous amounts of redirects and the huge sitemap. Folkd has a lot of pages. Way more than the default 50K that Bubble supports in a sitemap. We also had to deal with millions of redirects. All this requires a custom Cloudflare integration and that is not available on the shared cluster subscription plans. But CoAlias does and their setup is amazingly easy. Coalias provides these services by running a proxy allowing Bubble apps to leverage custom Cloudflare functionality. This allows us to create a multi-million page sitemap and manage redirects at scale.

It also provides a Prerender service which allows search engines to crawl, and index, pages at a much faster rate as it serves a different, much leaner, page to search engine spiders.

To answer the question at the beginning of this article: Yes we can build an app that scales to millions of users on Bubble.io, but we also need to use third parties. We did evaluate weweb.io for the front-end too but it does not make things easier. With WeWeb you can customize the sitemap, but it is very complex as it requires you to export the app and host it yourself. We felt that Bubble was still the best choice for a no-code solution.

Cleaning up Folkd's reputation

Before Eric came along, Folkd was abused by all sorts of nefarious actors. This brought them to a bad corner of the internet. Some users would spam links to illegal downloads, porn, and other unsavory things.

The last part of our job was to freshen up Folkd’s reputation. We needed to institute measures to ensure that the content posted on Folkd was clean. To achieve this, we used 3 different tools: Neverbounce, SafeDNS, and Webpurify to help with this.

From an SEO standpoint, we managed to usher Folkd back to the legal part of the internet. This caused an initial dip in traffic, but with new users piling in to try out the platform, Folkd is back on course. In part, we can also thank the work we did to optimize site speed and allow long-form posts.

Why not build Folkd in native code?

By now, the obvious question that might have popped into your head is why not simply build this app in native code? Especially given how much it stretches Bubble’s technical capabilities.

Simply put, even with all the scale challenges, it’s far faster to use Bubble than any native code solution. The development team is smaller, and the technical product becomes much more transparent for Folkd. Given Eric’s own technical experience, we had no problem communicating about the application.

Shipping Folkd: The Results

We were now ready to launch.

While daily posts were at the same level as before, the actual content was coming from real people who were using Folkd for its intended purpose – sharing bookmarks with other users.

In the 24 hours following its launch on Bubble, Folkd received over 1,500 new signups and 4,000 new links shared. Currently, user management and page load operations take up about 2.5 million workflow units per month.

Following the launch, we’ve continued to support Folkd with an in-house team. In terms of monetization, Folkd uses a combination of subscriptions and ads. Served through Monumetric.

While there certainly were hurdles, we were able to build Eric a spam-free version of Folkd that could be revitalized by thousands of new users. So, if you were ever wondering, you can build a highly complex app with millions of users on Bubble. With a bit of help.

Update June 2024

We will write a separate piece on Bubble.io and SEO, what we've done to overcome Bubble's SEO limitations and how it had an impact on a site that needs to rely completely on SEO for it's survival. As of now, folkd.com is scaled down to allow Google to fully crawl and index the site and run break-even.