Spotify Product Analytics

Product teams drown in reviews but starve for priorities. I treated the Spotify app's review corpus as a roadmap input: what are users actually asking for, and what's quietly driving them to churn?

• Mined 20,000 real user reviews with an NLP pipeline (NLTK + VADER) to score sentiment and surface recurring themes.
• Clustered complaints and requests into themes, then mapped each to a potential roadmap bet and churn-risk signal.

• A prioritized, evidence-backed view of what to build next and where churn risk concentrates.
• The kind of synthesis a PM can take straight into a planning meeting — opinion backed by 20k data points.