Company

Primephonic is a Dutch-American music streaming service that focuses solely on classical music. Founded in 2017 by an international team of classical music lovers, Primephonic has made it their mission to fix the streaming problem classical music has on the major platforms like Spotify and Apple Music. Constantly improving their app and web player since the launch to improve user experience, Primephonic is now available in 154 countries, and was earlier this year selected as one of the 100 best apps in the world.

Challenge

The team launched a Minimum Viable Product (MVP) version of their app. This initial version was built with the essential features and relied on market research, as well as “hunches” from the product team.

Six months into the product launch, Primephonic noticed that although subscribers to the service were increasing, app downloads were going down. At the time the teams only had access to marketing data from Appsflyer, app stores metrics and a custom solution to account for music that was being streamed for regulatory compliance purposes. None of these solutions were giving them the answers they needed to understand what their users were doing in the app.

In addition, being one of the first streaming services dedicated exclusively to classical music, meant there was only so much they could learn from existing players in the industry. Every new feature being developed was so customized to the classical music domain, that there was always a high degree of risk with each release. “We needed a way not only to measure if we were on the right track, but also as a way to decrease uncertainty in the future roadmap.” says Henrique Boregio, Chief Technology Officer at Primephonic.

As Primephonic was scaling up, new hires from product, data and even marketing teams started requesting data they didn’t have access to, leading Technology to investigate a product analytics solution. Having used Mixpanel’s starter version during the MVP development phase, Primephonic naturally evaluated the benefits of the Enterprise version, comparing it with Amplitude. The simplicity to both implement the solution and surface meaningful and actionable insight from it was the biggest decision factor in ultimately choosing Mixpanel.

Solution

Primephonic has tackled its most complex challenges using Mixpanel, from how to display a myriad of search parameters in an intuitive manner to improving daily curated playlists recommendations for app stickiness purposes, as well as transforming their conversion funnel to support their growth strategy. “When we build a new feature, we think about who we’re building it for and whether we’re leaving any users behind. Before Mixpanel, we never had a way to link it to our data.” comments Henrique. All of these milestones were achieved without complicated coding, any use of raw data or going through detailed specifications, thanks to Mixpanel’s simple interface.

“Most questions can be answered with what Mixpanel has to offer”, says Henrique. Before, teams were often making decisions based on “hunches” which led to opinionated conversations from the passionate team of mostly active users at Primephonic. As the solution to getting unbiased, data-driven insights, Mixpanel helped the streaming service shift its mindset and focus on real user data.

During the discovery phase, a cross-functional team worked together to define the exact events they wanted to track. As they had previously used Mixpanel, they understood the pitfalls of not putting in the effort upfront to define these events. “We brought it down to specific questions e.g. what’s the difference between a user that converts at the end of the trial period and one that doesn’t.”, highlights Henrique. “There was a lot of trial and error with the first few events, and we refined to get to what we wanted.”

Primephonic used the standard SDK documentation to implement Mixpanel, leveraging the “Sample tracking plans” as a foundation for their plan, which enabled them to move at speed. “Those documents really helped in highlighting potential issues with data consistency, or avoiding over-tracking. There was also one related to music, which we used as a template.” emphasizes Henrique.

Stakeholders across the company were able to dive immediately in and start creating their own dashboards, making communication across teams very smooth, especially for long-running KPIs which tend to remain stable over time. “The fast and intuitive interface provided by Mixpanel, decreased the barrier of entry, which in turn greatly contributed towards building a data-driven culture.” points out Henrique.

Results

From 26% to 80% of signups completed. When they first implemented Mixpanel, Primephonic started by analyzing the signup flow and were surprised to see a very low 26% completion rate. Deep diving into the funnel steps immediately brought to light that their web-based registration process, which took users out of the app, was a key root cause in drop-offs. “We always knew this was not the best user experience, but we could finally put a number to it. This made it clear to us that we needed a native signup implementation.” specifies Henrique.

Following the initial jump to 45% completion rate, Primephonic continued experimenting with different combinations and iterating on smaller feature changes until they landed on the best journey, resulting in an 80% completion rate. By keeping updated Mixpanel dashboards throughout all of these iterations, they were able to precisely measure if they were having a positive or negative impact, and adjust accordingly.

Proving user personas assumptions. The product and marketing teams conducted qualitative user interviews to segment their users into three well-defined personas, representing beginner, moderate and advanced users based on their overall knowledge of classical music. Primephonic leveraged this knowledge extensively to develop new content and features, but had never actually been able to test it properly.

When they released a new onboarding flow asking users to select a category, they immediately saw a distinction in how each group used the app using Mixpanel. “For instance, we saw how beginner-level users interacted more with our curated playlists, while advanced users preferred to spend their time either performing more advanced searches, or looking for very specific recordings” observes Henrique.

Surfacing gaps with Flows. Although Primephonic reports using the dashboards feature more regularly, Flows is the functionality that generates the most interesting insights to them as they’re able to dive really deep into their onboarding process, thus highlighting gaps which can then be fixed quickly.

For example, the product team was able to adjust the persona self-identification step of the conversion funnel, by pinpointing through Flows how users were going back and forth in the process as they were intimated by the questions about their knowledge. “Flows allowed us to pick up on this right away, and we saw an improvement in the conversions once we updated our onboarding questions.” confirms Henrique.

Validated north star metric. As a music streaming app, Primephonic was measuring app engagement by tracking commonly used metrics such as time spent on the app or which songs were added to the library. With Mixpanel, they realized the volume of minutes media had been played per day was more representative of stickiness.

“We could have a user that streams 10 hours a day, but only opened the app once, for 10 seconds, just to play his custom-made playlist.” says Henrique. By changing their mindset and tracking “media minutes played”, they are now able to capture these scenarios and draw more accurate conclusions.

Single source of truth across teams

Mixpanel is used by four separate teams across Primephonic including product, technology marketing and music content. Depending on the team, users are either logging in a couple of times a week to keep an eye on things, or using it more extensively during the development of a new feature.

As the solution flexes to varying functional needs, each user type is able to visualise data in their preferred way on personalised dashboards, whilst still using the same source of data. For example, product teams look at the adoption of new features, impact on retention and usage, whereas the music content team will monitor the engagement on beginner playlists vs advanced.

Combined features with marketing

In addition to monitoring conversion across registration funnels and different pricing plans using Mixpanel, the Primephonic marketing team benefits from the push and in-app notifications feature to highlight new features and appreciates being able to set them up without having to deal with the back-end.

“When we chose Mixpanel, we had these elements separate in our backlog and getting them both is a big win.” says Henrique, with his team currently in the process of integrating Mixpanel and Appsflyer to connect both sources of data.

Lessons learned from using Mixpanel for product analytics

As a seasoned Mixpanel power user, Henrique feels he has grown as a data-driven leader and has surfaced key recommendations for his peers:

• Start small and prioritize getting the basics right. “Chances are whatever you get out of the box, will be good enough for over 90% of your needs”
• Spend time streamlining events.“You always end up with too many events at first and too much data you don’t know what to do with.”
• Define an events naming convention to ensure they always reflect the data you’re looking for. “Different teams build things in different ways so you might think you’re tracking the same thing but you’re not!”
• Clarify your requirements to avoid over-planning your analytics strategy to the point of paralysing it and not finding the answers you need. “It’s very easy to get pulled into the hype of buzzwords such as AI, predictive analysis, A/B testing, big data, AI, etc. and to assume that you need all of it.”
• Allow yourself to be surprised! Success is defined by the users and the objective is that the product is as sticky as possible for each user. “Be open to any finds, as counter-intuitive as they may be, as long as they’re not damaging the brand.”