Fail Fast, Fail Smart, Succeed

on YouTube ↗29m

" summary

Describes building and launching "Spotify Now" where a New Zealand cohort test shows early promise then a full press rollout to 1% cohorts in UK SE DE flips results to -1% retention so the feature harms engagement; investigation finds an A/B testing bug and conflicting experiments that masked effects; outcome is months of targeted experiments, partial feature salvage and eventual small gains, illustrating the tradeoffs of secret big launches versus staged percentage rollouts and fast learning.

AI-written summary · about

" chapters (6)

1. 0:00Start
2. 3:19The lesson of Clippy
3. 10:14Extract the lessons from failure
4. 13:22Catalog the lessons
5. 15:35Minimize the cost of failure
6. 19:12The lesson of Spotify Now

" related talks

• 52mHow do we do product at SpotifyRachel Dubois · Agile on the Beach 2024
• 29mThe Spotify model - Does copy paste implementation workVili Mileva Yankova · DevOpsDays Eindhoven 2022
• 43mThe #Spotify model - Does copy/past implementation work?Vili Mileva Yankova · Update Conference Prague 2022
• 11mBreaking the (developer) silosRaul Herbster, Elin Nilsson · LeadDev Berlin 2019
• 01h 04mYou are not Spotify, and you don’t want to beRay Arell · HUSTEF 2022
• 43mHow Autonomy Saved One of Spotify’s Most Loved FeaturesJoakim Sunden · GOTO Aarhus 2023
• 48mFrom Operational to Strategic AgilitySteve Denning · Agile India 2018
• 43mHow Autonomy Saved One of Spotify’s Most Loved FeaturesJoakim Sunden · GOTO Copenhagen 2022