Spotify personas are the topic of much discussion by those in the product, design, and user research communities. Here, Olga Hörding, Mady Torres de Souza, and Sohit Karol explain how we developed our personas tool, how we use it today, and why it’s so useful for an autonomous, cross-functional organisation like Spotify.
Why not listen to our companion playlist for this article?
Here at Spotify, we often ask ourselves whom we’re designing for. And since listening to music is so universally popular, it might seem at first that the answer is ‘everyone’. After all, Spotify is available as a free and paid product. It can be used by anyone with a phone, computer, car, set of smart speakers, or many other devices. It’s present in over 79 markets and it offers experiences – like Daily Mixes – that are personalised to every single listener.
Yet designing for a mass, generalised audience isn’t likely to end up pleasing ‘everyone’. So in 2017, our team was challenged to create a better understanding of existing and potential listeners. We wanted to agree on how to differentiate the needs of these listeners and the problems our products could solve for them. We needed a solution that was durable and flexible enough to work for autonomous teams, working out of different offices, in different countries and on different parts of our products. And we were determined to put a face to our listeners – an identity that everyone at Spotify could recognise and talk about with ease.
We responded to this challenge by designing personas.
How did we craft the personas?
User-centred design has several schools of thought on how best to create and use personas. The general idea is that capturing and clustering the needs, goals, habits, and attitudes of existing and potential users helps to build a solid understanding of the problem space. For us, our personas tool is an example of a boundary object – a durable and reliable artefact that’s flexible enough to inspire discussions, share information, and adapt to the needs of the product development process. And we developed it in two phases:
Phase 1 (2017)
In Phase 1, we scoped our analysis to US listeners. We picked this market due to its size and the variety of listening behaviours that emerge from the way of life there – for instance, long commutes, suburban lifestyles, and so on. At the start, we discussed the idea of clustering behaviours gathered from our current data. But we moved away from this approach because it revealed only superficial knowledge about our listeners and concealed the reasons behind their behaviour. It also failed to help us understand why potential customers listen to music. So instead, we decided to study listeners of different ages, incomes, family types, lifestyles, music cultures, and more. We used a combination of diary studies and contextual inquiries to collect this data.
Early in the analysis, we noticed that people’s needs or reasons for listening to music were consistent, even in different clusters — that is, to kill boredom, to feel productive, to entertain themselves, etc. But what was different was their attitude towards music consumption, the value they saw in paying for music and their behaviours around devices in different contexts.
As a result, we ruled out the idea of clustering based on needs alone and used a combination of Alan Cooper's method and the Grounded Theory approach to build our personas instead. We transcribed our interviews minute-by-minute. Then, we coded and clustered them into needs, attitudes, device habits, contexts, and other dimensions in order to identify the best cluster combinations. Two tools — Mural and Airtable — were particularly useful during this phase.
Phase 2 (2018)
In our next phase, we built on a key Phase 1 insight – that when it comes to music listening, context matters. Sure, there’s value in creating abstract dimensions, such as needs and motivations. But ultimately, people use Spotify in the real world. Their device ecosystems, physical and mental abilities, and other contextual factors shape their listening choices. And so, combining the learnings from Phase 1 with a literature review of theories from sociotechnical systems and adaptive computing, we decided to focus Phase 2 on how people listen to music together.
In this phase, we sought to unpack the nuances and complexities that arise when people listen together at home, in the car, with kids, and more. And since this work built on our previous research, we once again kept our sampling within the US. We included roommates, empty nesters, partners with and without kids, households with toddlers, teenagers, and others. Our goal was to ensure we had an extensive variety of situations where people came together to listen to music.
Unlike in Phase 1, we followed up our diary studies and contextual inquiries with a bottom-up analysis using the Grounded Theory Approach. Qualitative coding revealed insights that we would have otherwise missed and resulted in the Listening Together Framework™, our tool to communicate the outcomes to a broader audience.
While people might have the same problems or needs, the existing habits determine the existing methods they use to address those problems. Attitudes determine how different people will adopt products designed to meet their needs.
Next, how should we represent our listeners?
Representing personas poses a tricky challenge: we want them to be relatable, but they’re not 1:1 matches with real people. Believable human traits and flaws help create empathy with problems and needs. But we don't want groups to be wrongly excluded based on the characteristics we've picked. So finding a balance is a crucial step if we’re to create useful and believable archetypes.
For that reason, we arbitrarily picked genders, names and appearances that matched the range of people we interviewed. While personas exist independently from these traits, they were fundamental to make them memorable as people. And deciding which human characteristics to include in each of the personas was especially challenging. To do so, we reduced the representation of personas to keywords, colours, symbols, and energy levels reflecting their enthusiasm for music. This exercise helped us navigate through the variations of poses, facial features, clothing, and visual styles we created.
To balance out these specific traits, we used flat illustrations with our brand colours, giving them a more abstract look. Avoiding a too-realistic representation made the material easy to refresh with evolving illustrative styles. It was also much easier to reproduce in high or low fidelity, since sketching a specific pose or picking a colour palette would be enough to refer to a persona.
How did we share our work?
We didn’t wait until our personas were complete before sharing them – we actually started thinking about communication as soon as we began our research. We spent a lot of time testing our asset ideas in pilot workshops. The goal was to integrate with our existing practices seamlessly. And by following our team needs, we crafted a communication strategy for Personas that includes digital assets, physical assets, and workshops.
Traditionally at Spotify, we create Google presentations when reporting back research – and sometimes, these get lost amongst all the many other presentations produced! But this time around, we envisioned our personas work to be relevant for at least a couple of years. So we created an interactive website, shared across Spotify offices through announcements and posters. Having a digital source of truth for the research was especially handy whenever we needed to update the study or add new learnings.
Raising awareness about the personas was useful, but we didn't want to stop there. We wanted to create fun, playful ways for the teams to incorporate them into their workflows. So we created assets that teams could use on their own, whether they were running one-hour mini-workshops or design sprints over several days. These assets were made available through our personas website.
Our team hanging out with the personas cardboard cutouts and the card game we've created to share the insights.
One of the most powerful modalities for learning that emerged during our pilot workshops was ‘learning by doing’. So the user research team hosted workshops with product teams and helped them to use personas in a way that was relevant to their specific areas.
What was the impact?
Since our teams are so autonomous, we realised right from the start that the personas would be relevant to all of them in different ways and at different stages of their work. For that reason, no one was mandated to use personas. Yet, as a reliable, durable, and carefully designed information artefact, we’ve seen many teams beyond the product organisation adopt them into their work and vocabulary over time – including those across Marketing, Content, and Brand.
For instance, teams that want to create features from scratch can now choose their personas, map out the existing opportunities, pick a direction, and start ideating from there. Although personas don’t replace user research, they can help us create educated hypotheses and save us time – meaning we don’t need to run foundational research every time we want to explore a new topic within the music listening experience. Our teams can now focus their resources on diving deeper into problems from the level set by the personas.
Equally, when teams are more focused on maintaining features, they can now map out their work and see how different personas would use it. They can create mental model diagrams for different personas and discover how they experience their journeys. And in doing so, they can refine the features to better fit certain ways of listening to music, while making sure they don’t alienate others.
Crucially, the personas are slowly becoming a part of our internal vocabulary – a means of helping teams to select and identify which ways of listening are being affected. We can’t optimise a feature for every single one of our listeners. So today, it’s common to see teams having their product roadmaps centred around specific personas instead.
A long process, with long-lasting results
Sometimes, in order to move fast, you have to move slow. Foundational research initiatives, like the development of personas, take time and are resource-intensive. Yet the learnings benefit us long into the future. And here are just a few of them:
When in doubt, over-communicate. We need a regular cadence to share details and progress around the organisation – this might add overhead, but it ensures alignment and transparency. We used Facebook Workplace, Slack, and emails to keep the stakeholders updated throughout the process.
Keep your disciplines close. Our process had to move quickly from behavioural analysis to fieldwork, then straight onto asset creation and scoping needs, attitudes and habits, through the use of surveys. The speed we moved was only possible by having design, user research, and data science integrated throughout the process.
Know your audience. Adopting new frameworks may be a significant change for some product teams. So we spent lots of time getting to know their workflows, running pilot workshops, and inviting them to fieldwork sessions in order to build trust and reduce any potential resistance to change.
As Spotify continues to grow, we expect to expand and adapt our personas for markets outside the US, as well as broadening out our area of study to also include podcasts. There are exciting times ahead and plenty more work to be done – we’re looking forward to the next chapter in the story of Spotify personas. :)
Mady Torres de Souza
Senior Product Designer
Mady designs with the Home Consumer Electronics team. She converts oxygen into activities like obsessive cooking and taking electronics apart.