The Twenty Minute VCSpotify CTO Gustav Söderström: TikTok's Music; How Olivia Rodrigo Gamed the Algo | 20VC #936
CHAPTERS
- 0:00 – 3:09
From involuntary entrepreneur to Yahoo acquisition
Gustav recounts how a lack of jobs after the IT crash pushed him into founding a mobile messaging startup. He explains the product idea (data-based texting vs. SMS), the carrier dynamics, and why selling to Yahoo made strategic sense at the time.
- •Graduating into the post–IT-crash job market and starting a company by necessity
- •Building messaging over data as a cheaper alternative to SMS
- •Carrier resistance because SMS bundles were a core revenue stream
- •Selling to Yahoo to leverage its global messenger and carrier relationships
- •Early perspective on big-company dysfunction at Yahoo
- 3:09 – 4:09
Why founder experience changes how you lead product
Harry probes how being a founder translates to stronger product leadership. Gustav emphasizes ownership, holistic thinking, and the tendency to care about the whole system—not just the product surface area.
- •Founder mindset creates stronger accountability and end-to-end ownership
- •Thinking like a CEO even when you’re not the CEO
- •Holistic perspective: questioning the ‘why’ behind decisions
- •Tension: caring about ‘everyone else’s stuff’ can be annoying but valuable
- 4:09 – 5:59
Joining Spotify in 2008: product disbelief, culture fit, and “useful naivete”
Gustav explains what convinced him to commit to Spotify early: the product felt almost too good to be real, and the culture valued intense debate and meritocracy. He also credits ignorance about the music industry as a hidden advantage that made the leap possible.
- •Spotify’s speed and UX felt ‘too fast to be true’
- •Attraction to Daniel Ek’s humble, non-hierarchical debating style
- •Tech-centric, meritocratic, idea-driven culture
- •Naivete about music industry complexity reduced perceived risk
- •Early mobile expertise was a key entry point
- 5:59 – 7:54
Macro winds: stop fighting inevitability, reposition fast
Gustav defines “macro winds” as unstoppable societal/technology shifts like broadband, smartphones, and machine learning. The lesson: accept reality early and reposition so the wind becomes a tailwind rather than a headwind.
- •Macro winds are larger than any company’s control
- •Examples: cheap broadband, smartphones, machine learning
- •Companies often waste time trying to resist inevitable shifts
- •Winning move is early acceptance + strategic repositioning
- •Speed of adaptation matters more than initial stance
- 7:54 – 9:33
Smartphones forced a business-model rewrite: making mobile free
He details Spotify’s original monetization—charging for mobile while desktop was free—and why the smartphone era broke it. Spotify had to redesign the model for mobile-first and mobile-only users, which unlocked exponential growth.
- •Early model: free desktop to build habit; paid mobile for on-the-go listening
- •Smartphone adoption created mobile-first/mobile-only users with no free tier
- •Growth slowed, signaling the model mismatch
- •Repositioning to free mobile was risky but necessary
- •Result: tailwind effect and exponential user growth
- 9:33 – 11:10
Fear of cannibalization and label constraints: how change actually happened
Harry asks how Spotify got comfortable cannibalizing a working model. Gustav explains the role of data (notably smartphone growth projections) and that the hardest stakeholder alignment was with record labels, which forced delays until growth pain made change unavoidable.
- •Using external data/projections to build conviction (smartphone trajectory)
- •Biggest blocker wasn’t internal—labels resisted freemium expansion
- •Waiting until growth stalled/declined to secure needed licenses
- •Parallel prototyping/A-B testing so launch could happen fast once licensed
- •Stakeholder negotiation as a core product constraint in music
- 11:10 – 15:30
Temporary vs. permanent behavior shifts: COVID modeling and multi-model thinking
They unpack how to tell durable shifts from temporary anomalies, using COVID as the case study. Gustav advocates explicit modeling, accepting models are always wrong, and using multiple models (different dimensions) to reduce catastrophic error.
- •COVID vs. mobile: reversibility made COVID harder to interpret
- •Decision to keep investing in ‘normal returns’ (e.g., car listening)
- •Models simplify reality; by definition they’re wrong
- •Risk comes from missing key dimensions (e.g., disease spread)
- •Using multiple models (Berkshire-style) to cross-check conclusions
- 15:30 – 19:01
Why Spotify Radio is buried: feature duplication and the pain of removing things
Harry challenges the UX choice around Radio being hard to find. Gustav explains Spotify’s historical paradigm (playlists) vs. the radio paradigm, and how overlapping features doing the same ‘job’ create complexity—yet removing features is disproportionately painful due to loss aversion.
- •Playlists vs. radio as competing user paradigms for similar outcomes
- •Different mental models by user cohorts (radio familiarity vs. younger users)
- •Feature creep is easy; removing features is psychologically costly
- •Loss aversion: users value what they have more than what they never had
- •Need to streamline while balancing competing use cases
- 19:01 – 20:44
Which customer feedback to trust: new vs. existing users and underlying jobs-to-be-done
Gustav shares how Spotify interprets passionate user feedback without overfitting to loud voices. He highlights testing on new users to avoid legacy habit bias, and reframing feedback around the true user goal rather than the requested feature.
- •Segmenting analysis: new users vs. existing users behave differently
- •Legacy habit can make improvements look worse in metrics short-term
- •Decision: is it worth retraining the existing base if globally better?
- •Translate feature requests into underlying intent (what job is being done?)
- •Avoid adding features that worsen overall complexity
- 20:44 – 23:02
Competing with giants: why doing the opposite can be the winning strategy
Gustav argues smaller players shouldn’t mimic big competitors’ strategies because those strategies exploit the incumbents’ strengths. Instead, find counter-positions the larger company can’t or won’t pursue, while acknowledging that copying can work for dominant platforms.
- •Incumbents’ strategies often align with their unique advantages
- •Copying usually yields a ‘worse version’ with fewer resources/assets
- •Look for counter-positioning that’s unattractive to the incumbent model
- •Exception: very large players can credibly copy emerging features fast
- •Spotify remains smaller than many competitors, so contrarian thinking helps
- 23:02 – 25:06
The YouTube contrast case: background shuffle vs. foreground on-demand
He gives a concrete example of contrarian strategy: instead of copying YouTube’s foreground, on-demand video experience, Spotify optimized for where YouTube’s free tier failed—background listening. This led to a free mobile tier centered on shuffled playlist playback in the background.
- •YouTube’s strength: foreground discovery + on-demand + video
- •YouTube free weakness: no background playback (stops in pocket)
- •Spotify bet on background listening as the dominant behavior
- •Licensed a background-shuffle free tier (not foreground on-demand)
- •‘Works where YouTube doesn’t’ as deliberate positioning
- 25:06 – 29:12
Spotify video strategy for podcasts—and when copying would’ve helped
Harry asks how Spotify’s background-first philosophy adapts to video. Gustav positions video as an enhancement for long sessions (talking heads), not a shift into silent, foreground-first content, and notes areas Spotify was slow (e.g., lyrics) due largely to licensing constraints.
- •Podcast video is possible because it’s not constrained like music licensing
- •Primary use case remains long background sessions with occasional glance-ups
- •Video supports context/identity/moments (e.g., notable visual clips)
- •Admitting slowness on some obvious features (lyrics)
- •Music product innovation constrained by licensing ‘lowest common denominator’
- 29:12 – 34:14
TikTok’s reinvention of music and the “taller mountain” mindset
Gustav credits TikTok (originating from Musical.ly) for meaningfully changing how music is used—through participation and creation—enabled by being small enough to experiment before licensing. He then explains why big companies must sometimes ‘lose altitude’ when jumping to higher-potential paradigms.
- •Musical.ly’s trial-and-error led to discovering dance/clip PMF
- •Being small allowed experimentation before heavy licensing costs
- •TikTok created a new interactive format for music usage
- •“Lower on a taller mountain”: moving paradigms often drops metrics short-term
- •Public company pressure makes temporary metric declines harder to endure
- 34:14 – 37:19
Where Spotify ‘lost altitude’: free-tier launch shock and recovery dynamics
Harry asks for a concrete moment of painful metric regression. Gustav describes how launching a massive free mobile tier tanked conversion rates initially—even as MAUs soared—forcing simultaneous investment in ad monetization and patience for conversion to catch up.
- •Pre-launch optics: conversion looked high because free intake declined
- •Post-launch reality: MAUs surged; premium conversion dropped sharply
- •Stakeholder anxiety: labels/investors focused on monetization collapse
- •Internal belief: ‘the more you play, the more you pay’ correlation
- •Risk centered on whether the contrarian product was good enough at scale
- 37:19 – 46:40
“Talk is cheap”: structured debate as a leverage tool (and how to make it safe)
Gustav defends long-form debate as cheaper than coding and a way to avoid shipping the wrong thing, citing Socratic reasoning and even historical examples. He explains how to structure debates, create psychological safety across cultures, and why Zoom sometimes improved discussion dynamics.
- •Debate can prevent expensive engineering misfires and slow A/B loops
- •Difference between being fast vs. being right; ‘fast to nowhere’ risk
- •Tactics: time-box 1–2 hour ‘empty it out’ sessions; pull in external brains
- •Psych safety via leader modeling: visible disagreement without punishment
- •Cultural differences (US hierarchy vs. Swedish norms); Zoom adds structure
- 46:40 – 1:27:34
Synchronizing leadership vs. swim lanes: scaling, reinvention, and near-death moments
Gustav describes choosing ‘synchronized swimming’ over strict org swim lanes to produce a coherent single-user experience—while warning about ambiguity in accountability. He then covers personal reinvention (from leading by example to coaching, returning to school for ML), Spotify’s label standoff near-death story, and closes with broader reflections on interactivity, algorithms shaping music, and product/business-model thinking.
- •Org design: optimize for one cohesive user experience, not internal boundaries
- •Failure mode: unclear responsibility leads to stalled decisions; balance is key
- •Personal scaling: shift from hands-on example to coaching leaders; deep-learning retraining
- •Near-death: preparing to ‘pull the switch’ and remove a major label catalog
- •If constraints vanished: make music interactive (a ‘GitHub for music’); incentives shape song structure; great products need durable business models
