Bumble Scales Up While Keeping Costs Down

Bumble Inc. is the parent company of Bumble, Badoo and Fruitz. The Bumble platform enables people to connect and build equitable and healthy relationships. Founded by CEO Whitney Wolfe Herd in 2014, Bumble was one of the first dating apps built with women at the center, and connects people across dating (Bumble Date), friendship (Bumble BFF) and professional networking (Bumble Bizz). Founded in 2006, Badoo is one of the pioneers of web and mobile dating products. Fruitz, founded in 2017, encourages open and honest communication of dating intentions through playful fruit metaphors.

Bumble experienced rapid changes following its historic IPO in February 2021, when Founder and CEO Wolfe Herd became the youngest woman to take a company public. This change brought new demands on internal reporting systems for product analytics, business metrics and developer metrics. This rapid growth brings new users, new data and new demands on internal reporting systems for product analytics, business metrics and developer metrics. “All the way up to our leadership team, everyone wants reporting in a consistent and timely fashion, says Head of Data Vladimir Kazanov. “That uses a lot of resources, and we found our legacy data warehouse wasn’t really suitable.” 

As Bumble’s reporting needs grew, Kazanov and his team found that scalability was limited. “We had analysts as well as software and data engineers relying on data and reporting in our on-premises system,” Kazanov says. “And at times, we had to shut down internal user queries so they wouldn’t interfere with other processes. Due to limited compute resources in on-premises systems, we’d often see collisions between ETL and ad hoc queries.”

“Our previous system was a powerful compute engine, but it wasn’t entirely suitable as a data warehouse,” Kazanov says. “It wasn’t easy to maintain a historic archive, and licensing costs were rising as our data grew. That’s what got us interested in cloud platforms like Snowflake; storage is cheaper in the cloud.”

The data engineering team at Bumble assessed several cloud providers and platforms. “We actually assessed Snowflake on AWS in the past, and when we moved to Google Cloud Platform, Snowflake was there waiting for us as well,” Kazanov says. 

As an organization with a large Google Cloud Platform footprint, Bumble was drawn to Google BigQuery at first. “We did a light evaluation of BigQuery,” Bumble’s Head of ETL Team Alex Tolmachyov explains. “But it didn't support all of our queries, and we could see we’d need to spend a lot of time adapting our pipelines to make it work. Ultimately, we needed something with full SQL support.” 

By using Snowflake, Bumble gets a platform that works across cloud providers, supports SQL and offers the complete spread of features it was looking for. “Getting Snowflake up and running was really straightforward, and the documentation was good,” Kazanov says.

“Once you get everything into a platform like Snowflake, you can offer it to every business user,” Kazanov says. “Snowflake powers our BI reporting and visualization. So, it’s a platform for our data analysts as well as other users that need more intuitive interfaces through a visualization tool.” 

This democratized data estate will only grow as Bumble continues migrating systems and data — and explores using Snowflake as a data lake. “Our data lake in Snowflake is a useful source of raw data on almost every data point,” Data Platform Engineering Lead Daniel Tavares says. “We can keep this data for longer than before, even after ETL processes, to give internal users access to data in an unprocessed form.”

Bumble has also been experimenting with Snowflake’s collaboration features, and found some unconventional, yet powerful, use cases for the feature. 

“We often have product managers and developers asking us to assess and integrate different tools into our data ecosystem,” Kazanov says. “Naturally, we have to check what’s suitable. And Snowflake’s data collaboration capabilities make it really easy to do that in a secure way.”

For Bumble, the greatest benefit of the Data Cloud is having an efficient platform that can scale, without the need for compromise or complex workarounds. “By separating compute and storage layers, we’ve been able to scale easily,” Tavares says. “In the past, we had to run a series of optimizations on our on-premises systems to re-size clusters and reduce storage requirements. With Snowflake, it just works.” 

Crucially, Bumble can grow its business without any fear of encountering performance or usability bottlenecks. “We now have hundreds of users on Snowflake, and they can run all the queries they like,” Kazanov says. “We don’t need to worry about how it will affect our reporting, ETL or any other process. And we can do it all while freeing up engineer hours we previously had to spend deploying and maintaining hardware.” 

“Most importantly, our internal users love the interface,” Kazanov says. “They’ve been really enthusiastic about it; I’ve heard people saying how great the web interface is and how much easier it is to set up connections.”

Next, the team plans to migrate internal software to Snowflake to simplify its data estate. “We currently have about 15 internal products built around our data warehouses and data lake. Now, we want to unify this into one platform,” Kazanov says. 

“We’re migrating certain data engineering processes over to Snowflake as well,” Tolmachyov says. “There are other Snowflake features we’re keen to use. Time Travel, for example, will help us perform financial investigations and understand how figures are changing over time and why.” 

As the team explores these new features, they’ll have resources from Snowflake to help. And they feel confident in their solid foundation in the Data Cloud. “It can scale almost indefinitely,” Kazanov says. “With separate storage and compute layers, we can have as many internal users as we like. It’s all so clean and separate in Snowflake, which we really appreciate.”