Role: Product Manager · Tenure: 2021 – 2023 · Domain: Multilingual Consumer Social · Live Audio/Video · Creator Economy
Scale: ShareChat (~180M MAU) + Moj (~120M MAU) = 400M+ combined MAU, 15 Indic languages, 8B notifications/day, 1.2B streamed minutes/month, 32M+ creators — India's first profitable social media platform (2024)
ShareChat serves Bharat's non-English internet. Every product decision had to hold across 15 simultaneous language contexts, first-time smartphone users, and low-bandwidth environments — problems structurally different from English-first platforms.
I owned seven product areas end-to-end, each run on the same discipline: Problem → Diagnosis (research method) → Approach → Outcome.
Problem: One English-first UI architecture created friction across 15 Indic scripts — font-rendering failures, RTL/LTR conflicts, culturally misaligned iconography.
Diagnosis: Field research in Tier-2/3 cities (Patna, Bhubaneswar, Coimbatore) conducted in the user's native language; cohort analysis showed Bengali & Tamil D7 retention materially below Hindi despite similar acquisition volume; journey mapping found Hindi CTA copy reused verbatim in Bengali UI.
Approach: Ran Design Thinking sprints (Empathize→Define→Ideate→Prototype→Test) with native speakers embedded as co-designers, not testers. Rebuilt the UI component library with language-aware string containers and dynamic font scaling; defined a new north-star metric — Language Comprehension Score (LCS), measured via task-completion rate per language — as the shared number across design, engineering and localization.
Outcome: Bengali D7 retention +18pp; onboarding completion +31%; language drop-off 38% → 11%; localization turnaround cut from 14 days to 3 days per language (5x).
Problem: 8B notifications/day across 15 types, each team optimizing independently — a tragedy-of-the-commons dynamic. Users receiving >12/day had 3x higher opt-out.
Approach — 3 phases: (1) Rate Limiting — user-personalized daily caps via ScyllaDB at 400K reads+writes/sec, <5ms latency; (2) Right Routing — language/content/time-of-day preference signals (Bengali users get Bengali-language hooks, not Hindi-translated copy); (3) Intelligent Timing (ML) — send-time optimization per cohort (Bengali peak engagement 7–9PM vs. Hindi 10–11AM).
Outcome: ~5% CTR uplift across 10 languages; opt-out rate −28%; notification volume −22% while DAU contribution held at 18% — precision targeting outperforming volume inflation.
Problem: 1.2B minutes/month streamed but no monetization, retention hook, or discovery loop — organic growth with no financial reason for creators to return.
Diagnosis: Observed 200+ live sessions; mapped 3 user archetypes (hosts, engaged speakers, passive listeners); funnel analysis found a steep drop between gift intent and gift completion — the UI required manually selecting both receiver and gift, with no pre-selection.
Approach: Applied the anchoring principle — redesigned the gift flow with pre-selected receiver+gift, cutting the send flow to minimal steps (highest-ROI single change in the redesign). Layered gamification (Gifter/Creator/Opinion/Community Battles) on top of competitive psychology already latent in users. Introduced 3D culturally-inspired gifts (Taj Mahal, rose) as aspirational status tiers.
Outcome: 1B+ virtual gifts sent in 2022 alone; 3x revenue from virtual gifts; monthly unique transactors doubled since Dec 2021; ~$100M chatroom GMV at scale.