Automating content with Nexrender: Spotify's approach

Discalaimer: This article is using materials from the original Spotify article.

We can make Spotify’s Performance Marketing more efficient by leveraging the power of engineering and content.

Spotify had a bold vision: to launch hyper-localized global advertising campaigns spotlighting regional artists and their hit songs. Their goal? To soar past industry standards with sky-high conversion rates, transforming listeners into paying customers with the expertise of their ads personalization team. A very ambitious goal indeed, particularly given the deliberate choice of video format as the primary creative driver for these campaigns.

Spotify owns an extensive data set including artists and their popularity metrics worldwide. While this wealth of data guided the selection of artists and songs for targeted regional advertising, figuring out if new users would like this music is a much more challenging task.

Recognizing this complexity, Spotify turned to cutting-edge machine learning algorithms to experimentally identify the most effective artists. These algorithms analyzed campaign performance metrics to pinpoint which artists resonated best with audiences, elevating the process from simply relying on popularity to being able to make precise predictions.

Orchestrating the generation of assets. Moving from Java templates to After Effects turned asset generation from something that could be done inline in an API call to something that needed rendering asynchronously. Scaling the render workers up and down in response to the volume of assets to be generated was also a challenge.

The first big challenge Spotify faced was content automation. Their solution? Using the power of Adobe After Effects to create animated ads, making the most of their team’s expertise with the software. But think about the huge number of videos the motion graphics team had to make. Each one had to be customized for different regions, languages and artists. The team had to build a really advanced rendering farm, with advanced automation capabilities.

Realizing how crucial creative production was, Spotify made it their top priority to solve this challenge first. They knew that without streamlined automation in this area, the whole project could fail.

The team identified Nexrender as the platform to automate video editing and never looked back on their decision. Nexrender seamlessly handled creative automation based on inputs from internal data and campaign metrics, empowering Spotify to efficiently generate and customize thousands of videos to fit specific parameters.

Our next option? Enter nexrender, an open source project that extends aerender into a first-class batchable system. Here, we can script file movements from network locations, specify multiple output formats all at once, and manage headless aerender nodes to slice up our batch jobs efficiently.

With nexrender, we were finally able to combine all the pieces that we wanted to automate the creation of visuals.

Conclusion

With Nexrender, Spotify tackled the challenge of ad content generation by automating their video creation process, allowing them to approach the remainder of the project with renewed confidence.

They developed a content ranking system to identify the optimal artists and songs for their campaigns. This involved cross-functional collaboration across engineering, marketing, and data science teams. Further, initial rankings were fine-tuned by a machine learning algorithm, which incorporated Spotify’s internal rankings and marketing statistics from campaigns. By iteratively refining their approach, they were able to make their performance marketing more efficient and handle the scale of the tens of thousands of ads running globally.