Gustav Söderström on Spotify, Product Strategy and the Value of Debate

Gustav Söderström became an entrepreneur almost accidentally. After graduating in electrical engineering and computer science in Stockholm during the aftermath of the technology crash, he could not find the job he wanted, so he and friends started a mobile messaging company.

Their product sent text over data rather than conventional SMS, which he says was much cheaper. The business attracted users in the Nordics but faced mobile carriers invested in SMS bundles. Yahoo acquired it to extend its desktop messenger to mobile, and Söderström worked there before returning to Stockholm.

In 2008 he met Spotify cofounder Daniel Ek and tried its desktop application. Söderström says the speed seemed almost too good to be true. Ek wanted someone to lead mobile product development before app stores and smartphones had achieved their later scale.

His previous founder experience gave him a company-wide ownership mindset, sometimes extending into colleagues' areas. He says the Spotify product, its technology-centered culture and the ability to debate with Ek attracted him. He also admits that not understanding the music industry's challenges made the move easier.

Söderström calls major changes such as cheap broadband, smartphones and machine learning “macro winds.” His advice is to recognize forces larger than any one company and reposition to benefit from them rather than try to stop them. This is his framework for strategy, not a guarantee that every trend will continue.

Spotify once gave desktop listening away and charged for mobile access. As smartphones spread, more people became mobile-only and had no entry point into the free product. Söderström says the company had to change a working business model by offering a free mobile tier, despite the risk of cannibalizing subscriptions.

Stebbings asks how the team overcame that fear. Söderström recalls using projections of smartphone growth to understand the risk of inaction. He says record labels were initially reluctant to license the change, and negotiations continued until Spotify's user growth had stopped and reversed.

While waiting for licenses, the team tested and built different concepts so it could launch quickly when agreements were signed. The account is Söderström's recollection of negotiations and company performance, not a complete independent record of the labels' positions.

Stebbings asks how to distinguish a lasting behavioral shift from a temporary one, using COVID-19 as an example. Söderström says Spotify chose to treat the collapse in car listening as an anomaly and continued investing for a return to normal, even when that return took much longer than the team expected.

As of the 2022 conversation, he says the company had recently seen home usage decline and car and mobile usage recover. He uses models to make assumptions shareable and predictions explicit, while acknowledging that any model simplifies reality and can be wrong.

To reduce overconfidence, he recommends using at least three models with different assumptions or dimensions. If they lead to similar conclusions, he believes the risk of a serious mistake may be lower. Even so, he says, no approach catches every surprise, and the pandemic illustrates that limitation.

Stebbings raises a product complaint: Spotify's radio feature is buried several clicks deep. Söderström says radio and personalized playlists can serve similar listening jobs, so the team has struggled to choose and combine user paradigms while keeping the interface simple.

He argues that adding a feature is easier than removing one because people value what they already use. He cites a behavioral-economics experiment to illustrate the point, though the conversation does not establish a universal numerical cost for removing features. Duplicate functions can also make the app harder to navigate.

When interpreting feedback, he suggests separating new users from existing ones. A change can initially disrupt an established habit while testing well with people who never learned it. He also wants to understand the job a user is trying to do rather than assume the requested feature is the only solution.

On competition, Söderström argues that a smaller company copying a larger competitor's strategy may end up with a weaker version because it lacks the same distribution, budget or assets. He prefers finding a position the larger player is not serving or does not want to serve because of its existing business model.

He acknowledges that a large player can reasonably copy a successful feature and offer it to its own users. His contrarian approach reflects Spotify's position relative to larger technology companies, not a rule that every product team should always do the opposite.

His example is Spotify's early free mobile tier. YouTube provided on-demand foreground video, but its free mobile experience stopped when backgrounded. Spotify instead licensed a background, shuffle-based experience so it could serve listening when the phone was in a pocket.

Stebbings asks how the logic changes now that Spotify also has video. Söderström says the service still emphasizes long sessions often consumed in the background. For video podcasts, users may mostly listen but bring the screen up occasionally to see the speakers or a moment being discussed.

Asked where Spotify should have moved faster, he mentions lyrics. He says the obstacle was often licensing rather than a deliberate contrarian choice: product teams in music must build within the rights they can obtain and afford. He presents that constraint as difficult but also as a challenge to innovate.

He credits TikTok with a major innovation in music use. In his telling, Musical.ly discovered that people wanted to make dance videos with short music clips, then developed the use case and licensed rights later. He contrasts that with the up-front licensing obligations a larger platform faces.

His “lower on a taller mountain” metaphor describes leaving a highly optimized product or business model for one with more long-term potential. The first move may make core metrics worse and require extensive spending before the new approach reaches parity with the old one.

Söderström says patience is especially difficult at a public company, where management and investors may ask why a year of work has only restored the old level. He argues that continuing to iterate can eventually unlock improvements unavailable on the previous path, though the result is uncertain.

Stebbings asks whether podcasts were such a case. Söderström distinguishes them: adding podcasts did not initially reduce existing metrics in the same way, but it took more than a year to assemble a catalog and reach the features users expected from a podcast player.

The free mobile launch is his clearer example of losing altitude. As free desktop intake fell, the apparent paid-conversion rate had looked unusually high. Once mobile free listening launched, new free users surged and the conversion percentage fell sharply before monetization had time to catch up.

Söderström says he expected conversion to improve because Spotify had observed a relationship between free listening and paid conversion, which he summarizes as “the more you play, the more you pay.” His real concern was whether the new free experience was good enough; labels and investors, he says, were more concerned about conversion and advertising.

His phrase “talk is cheap” deliberately reverses the idea that teams should settle disagreements only by writing code. He argues that structured debate among skilled people can expose a flawed idea before an expensive build and experiment. The key, he says, is allowing people to disagree without fear.

Stebbings objects that conversation can be costly for startups whose advantage is speed: endless theorizing may prevent shipping and learning. Söderström agrees that debate needs limits and clear ownership, but distinguishes moving quickly from moving in the right direction.

He says music licensing may have trained Spotify to think carefully because a wrong licensing decision could constrain the product for years. He also admits Spotify has sometimes been unclear about ownership, which can slow decisions. In his view, debate is useful when it builds alignment and still ends in a decision.

For a small company, he recommends hiring people willing to debate and drawing on outsiders when necessary. He suggests time-boxing a substantial topic to an hour or two: in his experience, that can prevent months of work on the wrong thing, though it is a judgment about trade-offs rather than a measured guarantee.

The host asks how junior staff can safely disagree with an influential leader. Söderström says leaders have to model the behavior: Ek can be challenged intensely, and Söderström tries to show disagreements with direct reports in front of their teams without punishing them.

He also invites quieter people to contribute, then expects the group to decide and, where needed, disagree and commit. He notes cultural misunderstandings: a Swedish colleague's silence can be interpreted by an American as agreement when it actually means disagreement.

Remote discussions had some unexpected benefits in his experience. Video calls gave debate more structure and reduced some domination by individual voices; he also found walking conversations over audio useful during COVID. He had expected strategy discussions to suffer more than they did.

Rather than divide his organization into completely independent lanes, Söderström tries to synchronize teams around one coherent user experience. He compares the work to synchronized swimming: the listener does not care which team built each part, only whether the application works as a whole.

The downside is blurred accountability. He says people may avoid decisions when everyone seems responsible, so teams still need a clear owner. That owner must bring the strategy to the group, explain it and respond to challenges, balancing agency with a shared direction.

Stebbings asks how Söderström has reinvented himself as Spotify scaled. Early on he led by example, even staying with teams working weekends. That style stopped scaling across many teams, so he had to learn to coach other leaders instead of trying to be personally involved everywhere.

Another transition came with deep learning around 2014–15. He returned to the mathematics, code and research papers so he could understand the technology he was responsible for. He says it took substantial time from family life, but he felt a technology leader needed enough depth to evaluate the change.

Stebbings asks why he did not simply hire an expert. Söderström says he is an engineer at heart and would feel unable to lead credibly if he could not understand the work his team was discussing. He applies similar reasoning to learning about newer technologies.

The host asks how he explains the cost of work to his family. Söderström describes trying to see his wife in the areas where she excels and allowing her to see his work, so they can renew respect for each other. He says the same can help with children, but it does not replace spending time together.

Asked about their relationship, he points to shared values, complementary strengths and a long period of knowing each other before marriage. He offers this as a personal account rather than a universal recipe.

On leadership weaknesses, he says the job changes as a company grows, even when the title stays the same, and that he still feels like an imposter at times. His company-wide perspective can be valuable but can also intrude on colleagues' responsibilities, so he tries to communicate diplomatically.

Söderström recalls Spotify twice experiencing negative growth. His most vivid near-death story concerns an early dispute with a record label over the free tier. He emphasizes that labels represent artists and seek value for their rights, even though negotiations can frustrate product teams.

He says one label threatened to withdraw its catalog from both free and paid users unless Spotify changed the free offering. Spotify built a switch to remove the music if negotiations failed, accepting that missing tracks could break playlists and badly damage the user experience.

With a midnight deadline approaching, he recalls the team preparing for the loss of a substantial share of the catalog. A call shortly before the deadline said an agreement had been reached. He says Spotify took the risk because it believed a free tier was essential to converting people who did not yet think they wanted to pay for music.

Asked what he would build without licensing constraints, he imagines a more interactive music experience in which users can experiment with elements of songs, analogous to learning from open-source code. This is a hypothetical product idea, not a claim that current rights permit those uses.

Stebbings asks whether Olivia Rodrigo changed song structure to suit algorithms. Söderström agrees with the broader idea that incentives and streaming behavior can affect music, citing choruses moving earlier and new listening categories such as sleep music. The captions do not provide evidence about Rodrigo's own intent or a specific song, so that individual claim remains unverified.

He predicts podcast and audio use could someday reach a market as large as social networking because communication and listening are basic human behaviors. As a Spotify executive, he has a direct commercial interest in that forecast; it is not a demonstrated market size.

He describes Spotify's shift from users curating playlists, through editorial curation, toward recommendations informed by machine learning. He says the company made that bet several years before the interview and increasingly thought of itself as a recommendation-based service.

Asked which competitor he respects, he praises large technology companies for repeatedly changing their products and business models as new technologies arrive. He wants Spotify to think similarly about music, podcasts, audiobooks and other formats, rather than define itself permanently by one category.

For a product he admires, he chooses Simply Piano. He likes its licensed popular music, staged learning and game-like feedback, saying it allowed him to play recognizable songs while gradually increasing difficulty. This is his personal user experience, not an independent efficacy assessment.

Stebbings points out that a great product is not automatically a great investment. Söderström agrees that he once undervalued business models and now believes the most disruptive innovations can combine new technology, new customer behavior and a different way of funding the business.

On crypto and Web3, he is excited by the technology but uncertain about a durable business model. He notes the gap between promises of trustless cooperation and the practical problem of trusting participants, and speculates that the technology might eventually matter more as infrastructure than as a consumer feature.

Asked what he would change about product management, he wants to demystify it. He provocatively calls product “100% science” and says good leaders should explain the models behind their instincts so others can learn, rather than protect an aura of unrepeatable genius.

Stebbings disagrees, arguing that visionary product leaders cannot simply be replicated. Söderström replies that making reasoning explicit can spread judgment across a company, even if doing so is difficult; the exchange remains a difference in perspective, not a settled scientific finding.

He closes by praising Shaq's ability to connect technology, artists and the music industry, including around Spotify's US launch. Stebbings thanks him, and both return to their shared enjoyment of discussing products.