Summary
With the massive opportunity in the market, the focus remains on Acquisition, i.e. getting new users to register and use the service. The struggle becomes getting a larger volume of registrations each quarter, at lower rates (cost per registration). While manual optimizations are necessary and effective, the scale of data and the required frequency of changes become huge challenges.
Challenge
Spotify is one of the top music streaming services in India in 2022, a notable achievement since its launch in India in 2019. To compete with the other top players, the primary performance objective is to maintain a healthy MAU number which is achieved by running a combination of Acquisition, Retention and Reactivation. With the massive opportunity in the market, the focus remains on Acquisition, i.e. getting new users to register and use the service. The struggle becomes getting a larger volume of registrations each quarter, at lower rates (cost per registration). While manual optimizations are necessary and effective, the scale of data and the required frequency of changes become huge challenges. To deal with this challenge, we decided to integrate Scibids, an AI and Machine Learning tool that is able to crunch data and influence bids to optimize towards performance goals.
Objective
In the race to acquire the most MAUs, aggressive new user acquisition campaigns are imperative for Spotify to compete with the top players. Our primary focus is on how to increase the scale of our existing campaigns while making them more efficient, a formidable undertaking. One of the important pillars of our performance media mix are Programmatic campaigns. While most app install (+ registration) campaigns focus one driving acquisitions, Programmatic has a more targeted approach. Using audience buying, the Programmatic approach allows us to reach out specific audiences that are most likely to convert or move down the conversion funnel, where other channels can capture the conversion. Making this channel efficient was our top priority as everytime we tried increasing the scale, the cost-per-registration (CPR) would shoot up. Unlike many other performance channels, programmatic campaigns provide lots of data signals on performing and non-performing audiences. Manually analyzing these signals and optimizing the campaigns while essential, is a time and effort intensive process. Dealing with both issues became the task at hand.
Strategy
First step was to have a clear understanding of the type of data signals. Then identifying the ones that best informed optimisation decision making. Once done it was important to on-board and integrate a tool that helps take care of the tedious analysis that could help the executions specialist make quicker, strategic decisions. After much delibration the team decided to go ahead with Scibids. Scibids' role consists of 3 main actions: Conducting detailed analysis and creating learning datasets, to identify the variables most likely to impact the campaign. This allows the team focus more on Strategy rather than data crunching. It is also able to take in 3rd party data signals, a feature not available in the DSP. Using machine learning to recommend optimal bidding to influence expenditure at an auction level. It builds multiple models based on many variables, the simplest being Geo, App/Url, Device, Time of Day, they make the best use of DV360's Custom Bidding feature. Repeating the analysis and optimisation recommendations every few hours, upto 12x a times. Allowing for faster and more frequent optimizations resulting in a smarter campaign. We progressed with a systematic approach to start with A/B testing to make sure not to shock the existing campaign. Then increase spends on the Scibids-supported campaigns to test efficiency when scaled. We then started making use of DV360's Custom Bidding feature when the feature came out in H2 2021. With enough experience and data under our belt the last update was to start exploring new SSPs to recognize and target new, untapped audience buckets.
Data
Spotify is one of the top music streaming services in India in 2022, a notable achievement since its launch in India in 2019. To compete with the other top players, the primary performance objective is to maintain a healthy MAU number which is achieved by running a combination of Acquisition, Retention and Reactivation.
Solution
Results
The performance over one year proved the decision to onboard Scibids to be a huge success. In Q3'21, after the A/B test, we pushed the campaign and doubled spends which resulted in a 25% increase in CPR only Looking at the success, we further increased spends by 6x in the subsequent quarters resulting in a 33% higher CPR, another very positive result Exploring additional SSPs helped the campaign drive 50% more regs than those driven from Google AD Manager at the same CPR
