Slack: From internal tool to global platform
Originally a gaming company, Tiny Speck built an internal chat tool while developing its game Glitch. The game struggled; the chat tool showed unexpected promise.
Signals
- Glitch had low traction despite heavy investment
- Internal teams relied on the chat tool daily
- Early external testers responded positively
Experiment
- Shared the internal tool with small outside groups
- Collected feedback and usage data to validate demand
- Measured adoption against the game
Result
- Strong adoption and clear product-market fit
- Public launch in 2013 and rapid B2B SaaS growth
- Acquired by Salesforce for $27.7B in 2021
Lesson
A purposeful pivot is not failure but evolution: listen to signals, test hypotheses before committing, and align purpose with a real customer problem.
Spotify: Empowering autonomous squads
Spotify reorganised engineering into small, self-organising squads, grouped into tribes, with chapters and guilds for shared practice.
Signals
- Rapid growth created bottlenecks in traditional structures
- Hierarchical overload slowed innovation and delivery
- Teams needed autonomy to respond to user needs
Experiment
- Small cross-functional squads with full ownership of a product area
- Squads choose their own agile approach
- Tribes, chapters and guilds to keep alignment
Result
- Faster decisions and release cycles
- Alignment without centralised control
- A highly aligned, loosely coupled model widely studied and emulated
Lesson
Clear missions, autonomy over how, and lightweight alignment structures let teams innovate faster without losing coherence.
Netflix: From DVD rentals to continuous streaming innovation
Netflix pivoted from DVD rental to streaming and then content creation, adopting a product-centric, continuous delivery model.
Signals
- Release cycles for DVD software could not keep up with streaming
- Hulu, Amazon and later Disney+ demanded constant innovation
- Customers expected low latency and personalisation
Experiment
- Long-lived product teams per area such as personalisation and playback
- Automated testing, deployment and rollback pipelines
- Budgeting moved from projects to products
- A/B testing at scale
Result
- Hundreds of deployments per day with minimal downtime
- Continuously optimised personalisation
- Global scale, 260M+ subscribers in 2024
Lesson
Moving from projects to products unlocks continuous evolution. Long-lived teams with autonomy and data compound small improvements into strategic advantage.
Duolingo: Gamification driving global learning engagement
Duolingo designed its product around gamified engagement, turning a long-term skill into a daily habit with streaks, XP, leaderboards and badges.
Signals
- Language apps suffered high drop-off after a few weeks
- Learners needed constant motivation over years
- Classroom-style apps failed to hold attention
Experiment
- XP points for lessons and drills
- Streaks for daily engagement
- Leaderboards for friendly competition
- Levels and badges, tuned through A/B tests
Result
- Daily active users grew to 20M+ globally
- Learners with streaks were 3x more likely to stay active
- Habit loops kept many learners daily for years
- IPO in 2021 at about $6B valuation
Lesson
Gamification fuels energy and habit formation. Blending intrinsic and extrinsic motivators sustains engagement and adoption.
Airbnb: AI-powered search and personalisation
Airbnb embedded AI and machine learning across its marketplace: search ranking, recommendations, dynamic pricing and fraud detection.
Signals
- Users overwhelmed by listing options abandoned searches
- Conversion suffered when results were not relevant
- Host visibility did not always match guest relevance
Experiment
- ML model to rank search dynamically
- Signals such as past bookings, location, seasonality and host responsiveness
- A/B tests against older rules-based ranking
- Personalised recommendations and pricing tools for hosts
Result
- Significantly better booking conversion
- More relevant results and fewer abandoned searches
- Better visibility and pricing for hosts
- AI embedded across recommendations, fraud detection and support automation
Lesson
AI works best when deeply embedded in the flow of value creation, not treated as a side experiment.
These are public company stories analysed in the ValueFlow book as illustrations of each flow. They are not ValueFlow client engagements.