Bumble’s Optimisation Journey Behind ‘Best Photo’ Feature

Dónal Keane, Senior Data Scientist at Bumble Inc, shared insights about how the dating app’s machine learning models quickly identify a user’s best photo. 

Speaking at the recent Data Science Festival, Keane explores how the app was able to boost engagement by optimising its data analysis. You can watch his presentation below:

Any dating app developer will tell you that photos play a significant role in online dating and a user’s dating success. But what can developers do when users don’t arrange their profile photos in an optimal way?

Dónal Keane dives into this issue, highlighting two important points. Firstly, oftentimes dating app users don’t actually recognise which of their photos is the most appealing to others. 

Secondly, a user’s journey on a dating app may be fairly short, meaning that their photo arrangement needs to be optimised quickly in order to ensure their user experience is maximised.

In his presentation at the Data Science Festival on the 14th of October, Keane explores some of the experimentation and analysis models that Bumble deploys to quickly and effectively find a user’s best photo.

It is not as easy as displaying each user’s photos for an equal period of time, and then analysing which is the most successful. Bumble learns on-the-go from the experimentation, ensuring that less popular photos don’t receive unnecessary focus.

From these product optimisations, Keane highlights that user activity, engagement, matching metrics and Bumble’s revenue all increased as a result of finding users’ ‘Best Photos’, faster and more efficiently. 

Read the official description of Dónal Keane’s presentation here.

Photo courtesy of the Data Science Festival.