Case Study · Careers360 · Jul 2024 – Present
Polls &
Discussions
A complete UGC content creation and distribution system that generated 300,000+ votes and doubled platform engagement — built lean enough to run without ongoing editorial effort.
The Problem
Careers360 had strong informational content but pages were largely static, resulting in low dwell time, minimal return visits, and weak signals for search engine indexing.
Manually creating and mapping polls to the right pages at the right time was effort-intensive, making it difficult to scale across hundreds of exam, college, and result pages simultaneously.
My Role
As Product Manager, I owned end-to-end delivery — from identifying the opportunity through user research and competitive analysis, to defining requirements for the poll CMS, AI suggestion engine, auto-scroll behaviour, and entity mapping logic, through to working with engineering and design on execution and monitoring post-launch KPIs.
Key Insights
User Research
Students actively sought peer validation for exam and college decisions on external platforms — confirming latent demand for opinion tools Careers360 could own natively.
Platform Behaviour
Pages with interactive elements had significantly higher dwell time and lower bounce rates than static pages — polls presented a low-friction path to engagement.
Operational Reality
Manual scaling would require unsustainable editorial bandwidth. Building an AI suggestion engine was the only path to quality at scale.
Search Signal
UGC — especially comments tied to specific exams and colleges — generates long-tail keywords that improve crawl frequency and ranking.
Product Architecture
Poll Creation CMS
A dedicated internal CMS for the editorial team to create, schedule, and publish polls without engineering involvement — with configurable questions, answer options, target entities, and display rules.
AI Poll Suggester
An AI-powered suggestion engine integrated into the CMS. Based on the target entity (exam, college, article), it suggested contextually relevant questions and answer options — reducing creation time from hours to minutes.
Auto-Scroll & Sequential Discovery
Polls auto-scrolled to the next in sequence without manual interaction, extending session depth and increasing total votes per visit by mimicking familiar social-scroll behaviour.
Keyword & Entity Mapping
Each poll was mapped to entities and keywords, enabling automatic surfacing on relevant pages. A fallback keyword-based auto-fetch ensured no page was ever left without an active poll — making distribution self-sustaining.
Outcomes
"Polls and Discussions demonstrated that user-generated content — when designed around real student anxieties, operationally scalable for the team, and intelligently distributed — can drive outsized results with minimal ongoing effort."