Gustav Söderström - How Spotify Thinks - [Invest Like the Best, EP.424]

Theme 1: From "Downlink" to "Symmetric Dialogue" — AI Reshaping the Consumer Product Paradigm

Gustav Söderström argues that the fundamental shift brought by generative AI is from "asymmetric machine recommendations" to "symmetric human-machine dialogue," which will completely transform the form of all consumer products.

1. Old Paradigm: Asymmetric Recommendation Systems

• Mechanism: Traditional machine learning (e.g., TikTok's feed) is essentially a "downlink" that is extremely wide (pushing large amounts of information to users), while the "uplink" is extremely narrow (only a few signals such as clicks, scrolls, and skips). The interaction between users and products is one-way, an "output" mechanism of the machine.
• Data Support: Gustav points out that even the "skip" signal is very coarse — "You skip a song because you hate it, or because you love it so much you're tired of it, or because you're working out and jazz doesn't fit. To us, all of that looks like a single 'skip' signal."

2. New Paradigm: Symmetric Conversational Interface

• Essence: Generative AI makes the "uplink" as rich as the "downlink," because users can directly input their needs using natural language (English). The product transforms from "a service you use" to "an entity you can talk to."
• Implication: "If you fast-forward five to ten years, almost every large consumer product will, to some extent, become a conversation." Gustav predicts that consumer products will evolve from "services" to "dialogues."

3. Concrete Example: AI Playlist Generation

• Mechanism: Spotify's already-launched AI playlist feature is a practice of this concept. Users can describe their needs in natural language (e.g., "I want a running playlist, EDM style, with big drops, 160 BPM"), and the LLM generates recommendations.
• Value: Gustav considers this "the first time we have obtained such a high-fidelity signal of user intent." It is equivalent to conducting continuous, in-depth qualitative user research on "nearly 700 million users." This helps the company "recreate a small part of your cerebral cortex on our servers."

---

Theme 2: The Art of Organizational Structure – Synchronizing Teams with the "Bets Board" Process

Gustav Söderström emphasizes that Spotify's "super app" strategy dictates a highly synchronized, flat organization, with a unique "Bets Board" process to allocate resources and ensure strategic focus.

1. Strategy Determines Structure: The Price of a Super App

• Mechanism: Unlike the Western "one app, one function" model, Spotify adopts the Chinese "super app" approach, integrating all content (music, podcasts, video, audiobooks) into a single app. This brings immense distribution advantages, but also means "everything is interdependent. You have to ship an app all at once. Everyone is a stakeholder. No one can act alone."
• Conclusion: This consumption strategy dictates that the company must be highly synchronized, thus requiring a robust planning process.

2. Core Tool: E-team (Executive Team) Meetings

• Structure: All 14 VPs meet for 3 hours every Tuesday, forming a "fully synchronized team."
• Core Principles:
• Real-time resolution: "You are not allowed to say, 'Let's take this offline' or 'We'll talk later.'" Because all decision-makers are present, dependencies can be resolved immediately, ensuring "any blocking issue lasts no more than 2.5 days at most."
• Detail-driven: VPs are not allowed to bring their direct reports to the meetings, forcing VPs to dive into the details themselves rather than relying on subordinate briefings. This also builds strong team alignment and a culture of candor.
• Data Support: This structure ensures that product, tech, and business teams all understand the company's big picture—product and tech people know the P&L and gross margins, while business people also understand AI and the codebase.

3. Resource Allocation Mechanism: The "Bets Board" Process

• Process: Every 6 months, the 14 VPs submit "bets" they want to make, as if pitching to a VC at a startup.
• Decision: Two co-presidents (Gustav and Alex Nordstrom) globally rank all proposals (usually 30–50), then allocate resources sequentially starting from #1 based on capacity, until resources are exhausted.
• Key: "The real trick is to do the ranking." Gustav argues that without ranking, claiming everything is important only pushes decision-making pressure down the organization, leading to internal conflict. Open and transparent ranking brings clarity and reduces friction.
• Risks and Challenges: The biggest risk is that "planning is expensive." The company has developed an internal tool for this. The 6-month cycle is a balance between "reaction speed" and "planning overhead."

---

Theme 3: Symbiotic Relationship with the Music Industry — R&D Department and the Scale Game

Gustav Söderström positions Spotify as the "R&D department" of the music industry, emphasizing that it has borne 15 years of losses to invest in industry growth, and that the key to resolving the "per-stream revenue" controversy lies in expanding the paid subscriber base, not in raising the royalty share.

1. Positioning: R&D Department and Capital Risk

• Historical Background: In Sweden, the music industry was on the verge of collapse due to piracy, which made the industry willing to take risks with Spotify. Gustav points out that the music industry itself has no R&D department, and Spotify is that R&D department.
• Data Chain: "If we look at the revenue trajectory, we distribute about 70%. But the remaining 30% we did not keep; we invested all of it, and even more, back into the music industry. We have been losing money for 15 years, while the music industry has been profitable during this period." This directly positions Spotify's losses as an investment in the industry.

2. Solving the "Per-Stream Payment" Debate: Scale is Key

• Viewpoint: The debate about "per-stream payment" being too low is a "red herring". Gustav argues mathematically: even if Spotify were a charity and distributed 100% of revenue to rights holders, the payment would only be 1.5 times the current amount.
• Core Strategy: The real solution is to "rapidly expand the number of paid music users". He believes that billions of people around the world should pay for music.
• Data Support: The paid conversion rate in the Swedish market is about 40%, which is a mix of mature and emerging markets. Over time, all markets will increasingly resemble Sweden.

3. Competitive Difference: The Better the Product, the Lower the Per-Stream Payment

• Paradox: "The better our product, the lower our per-stream payment appears to be." Because Spotify's paid subscriber engagement (listening hours) is more than double that of competitors, and its churn rate is half that of competitors. Under the "pay per user" royalty model, the higher the engagement, the lower the per-song rate.
• Conclusion: Therefore, competitors' "per-stream payment" is higher because their product is worse and users listen less.

---

Theme 4: The Power of Good Explanations – From "Pattern Recognition" to "Causal Theory"

Gustav Söderström, deeply influenced by physicist David Deutsch, emphasizes fostering a culture of "good explanations" internally, requiring teams to not only know "what works" but also understand "why it works", which offers greater scalability and lasting value than empiricism.

1. Theoretical Origin: David Deutsch's "Good Explanation"

• Criteria: A "good explanation" should not only be falsifiable but also be "scalable" (able to explain phenomena at different scales) and "hard to vary" (you cannot easily replace its components without undermining its explanatory power). Gustav believes that most conspiracy theories are implausible precisely because they are "too easy to change".

2. Implications for Product Management: From Intuition to Science

• View: The "intuition" or "magic" popular in product development is actually just "inexplicable pattern recognition", i.e., "seniority". Gustav argues that if someone cannot explain their intuition, it means "you are not smart enough to articulate it clearly yourself".
• Value: A communicable "causal theory" is more valuable than an inexpressible "pattern recognition". Because a theory can be learned and applied by the entire organization, and ultimately lead to "leapfrog" breakthroughs rather than incremental progress along the original path.
• Case Study: The Birth of the Mobile Free Tier
• Problem: With the proliferation of smartphones, a large number of new users had no computers, only phones, and Spotify's free model (based on computers) became ineffective.
• Pattern Recognition: Imitate YouTube, create a free model of "on-demand playback with ads".
• Good Explanation: Charlie Hellman started from first principles. He analyzed and found that at the time, 91% of music listening was "background" behavior (screen off). The core user need was "music in the pocket as background sound". Then he discovered that 50% of paying users also used the "shuffle" feature, meaning "shuffle" was not a mandatory behavior for paying users. So, they designed a tier: "a playlist that can be shuffled, phone in pocket, free forever". This was highly counterintuitive, but ultimately succeeded without severely cannibalizing paying users.

---

Mentioned Targets

Target	Guest Attitude	Key Data
Spotify	Bullish (Long-term value creator)	Nearly 700M MAU, nearly 300M paying subscribers; podcasts and audiobooks increase user retention and willingness to pay; first annual profit in 2025; paid over $10B to the music industry.
Netflix	Neutral (as a comparison reference)	Its "content differentiation" model was tried by Spotify (exclusive podcasts), but proved unsuitable for Spotify.
YouTube	Risk Warning (as a competitor)	In the mobile free era, it is the "obvious" choice for pattern recognition, but Spotify bypassed this path through "good explanation".
Pandora	Neutral (historical reference)	It is a form of free mobile listening, but only allows listening to radio stations, not one's own curated playlists.
Joe Rogan	Neutral (as a case study)	Previously signed an exclusive agreement with Spotify, but Gustav considers this "a poor strategic decision" because the low production cost of podcasts makes the "exclusive" model contrary to industry trends (platformization/distribution).
Cursor	Neutral (tool/trend)	Coding efficiency improvement (7%), but mainly affects writing new code, with limited impact on large companies; a greater impact lies in allowing non-developers to interact with infrastructure via the MCP protocol.

---

Judgments Worth Remembering

1. "AI is either a business model change or just a product change." (Gustav Söderström)

Support: Compared to the mobile era (which required changing free/paid business models), AI's impact on Spotify is currently at the product level (e.g., AI playlists). However, high inference costs are a new variable that may lead to more subscription tiers based on "inference volume."

2. "We position ourselves as the R&D department of the entire music industry. We lost money for 15 years, while the entire music industry was profitable during that time." (Gustav Söderström)

Support: This directly highlights Spotify's symbiotic relationship with the industry and explains the strategic rationale behind its long-term losses—essentially a form of industrial capital investment.

3. "The debate over per-stream payment is a red herring. The real solution is to rapidly expand the paying user base." (Gustav Söderström)

Support: Even if the company were charitable, the payout rate could only increase by 50%. The real way to grow the pie is to get billions of people to pay, rather than raising the existing revenue share ratio.

4. "The better our product, the lower our per-stream payment looks." (Gustav Söderström)

Support: Because user engagement is more than double that of competitors, under the "total pool" allocation model, higher listening hours dilute the per-song payout. This is the paradox of product success.

5. Three elements of a "good explanation": scalable, compatible with all known explanations, and hard to change. (Gustav Söderström, derived from David Deutsch)

Support: This is the philosophical framework Gustav uses for internal decision-making. He used this framework to defeat the "pattern recognition" of imitating YouTube and created the unique "mobile free shuffle" model.

6. "No saying 'talk offline.' All decision-makers are present, and dependencies must be resolved in real time." (Gustav Söderström)

Support: This is a core principle of Spotify's E-team (Executive Team) meetings, aimed at ensuring that a highly synchronized organization is not delayed by any blocking issue for more than 2.5 days.

7. The core of the "Bets Board" process is "global ranking." (Gustav Söderström)

Support: Not ranking means pushing conflicts down to the organization. A public, transparent ranking (from 1 to 44) provides absolute clarity for the entire company, avoiding internal friction caused by resource competition.

8. "The real mistake is trying to defend your past decisions." (Gustav Söderström)

Support: When explaining why Spotify gave up its podcast exclusivity strategy, he admitted it was a "bad strategic decision" and quickly pivoted to embracing the "platform" and "distribution" model.