Disney+ Hotstar · Product Lead, Ad Platform · 2018 to 2020
Building the world's largest live sports ad platform
Built Hotstar's in-house ad server from zero and scaled to a then world-record 25M concurrent viewers and 3B+ ad impressions per day during the Cricket World Cup and IPL.
Peak concurrent viewers (then world record)
Ad impressions per day
Buffering & playback failures
Over-delivery (vs. 50%+ historical)
Context
Hotstar relied on third-party ad servers that couldn't handle the unique traffic shape of Indian live cricket: concurrency that swings from millions to tens of millions inside a single over.
Problem
The dynamics of streaming video in India are unique. Billions of viewers, mobile first viewing, and a rapidly expanding digital footprint. The one streaming product that delivered to this scale relied on third party vendors. When I joined the team in 2018, my first task was - build a live sports monetization engine that delivers the first personalized video ad in India's history.
Approach
01
Re-thought pacing and ad decisioning for live sports
Existing ad servers paced against traffic patterns, which inverts on live sports: viewership drops when results become predictable and spikes on cliffhangers. The result was massive over-delivery, missed advertiser targets, and zero ability to personalise ads at live concurrency. After observing sports viewership patterns, I designed a break-based pacing algorithm instead of a traffic-pattern one. This combined with cohort level decisioning system became the platform's structural advantage.
02
Built for operators and engineers
In tandem with building in house monetization Disney+ Hotstar was moving all match play-out operations in house. The live feed from the stadium directly piped into our office and here's where operators punched in breaks, and ad ops teams optimized campaigns in real time. I shipped a Match Management Dashboard with real-time impression visibility, giving live-ops the controls to intervene during a live game instead of after.
03
0→1→100 in 24 months
Took the platform from a blank repo to running every ad on the Cricket World Cup final at 25M concurrent viewers, with personalised live sports ads. An India first. The cohort based decisioning system has become a well adopted practice for live sports monetization at other major streaming platforms like Prime Video.
Outcomes
- →Delivered India's first personalised live sports ads at scale
- →Sustained 100M+ DAU and 3B+ impressions per month with sub-1% playback failure
- →All advertisers met delivery targets; over-delivery capped under 5% (down from 50%+)
- →Became the foundation for Hotstar's monetisation through subsequent IPL seasons
Reflections
The biggest lesson: domain insight beats infrastructure muscle. The break-based pacing call came from studying traffic patterns of historical games and correlating it with actual events on field. I still apply that pattern by sitting with the workflow before I architect for it.
Next case study
80% of Advertisers Were Missing 95% of the Inventory.
Amazon Ads