Angel One, a prominent Indian financial services platform, has successfully leveraged Apache Pinot to handle high-capacity loads and deliver ultra-fast query times across its diverse business verticals. By implementing Pinot, Angel One has resolved previously difficult analytical challenges, enabling real-time personalization, automated campaign management, and seamless user onboarding.
The Challenge
Angel One operates across multiple complex business verticals, including equity, futures and trades, commodities, loans, insurance, B2B, and B2C. Each of these verticals has its own distinct requirements for analytics, monitoring, and decision-making panels. The company needed a highly scalable, real-time database capable of powering user-facing features while simultaneously driving internal operational dashboards.
The Solution
Apache Pinot has been instrumental in shaping several critical systems at Angel One:
- Consumer Experience: Angel One uses Pinot to provide users with a personalized view for stock trading and related activities. By calculating trends and popularity in real time, the platform customizes the user interface based on individual behavior, thereby improving the overall user experience and maximizing ROI.
- GRIP (Notification System): GRIP manages promotional and marketing campaigns. Pinot powers a real-time dashboard that allows campaign managers to make instant decisions and detect anomalies. Furthermore, GRIP utilizes a phased campaign mode where subsequent phases are only executed if the previous phase hits specific performance thresholds. This automated behavior lessens user fatigue and maximizes return on investment.
- PRISM (User Onboarding): The user onboarding journey involves numerous steps, including KYC and stock exchange registration. Pinot provides clear visibility into this funnel, alerting the team to user drop-offs and anomalies. Crucially, PRISM features a self-healing system powered by Apache NiFi working on top of the Pinot database. When it detects an issue with a user’s onboarding, it automatically triggers workflows to fix the problem without manual intervention, allowing users to onboard quickly.
Scale and Performance Metrics
The implementation of Apache Pinot handles massive loads across five different business tenants. Although the numbers are continually growing, baseline operational metrics include:
- Ingestion: ~100k peak Transactions Per Second (TPS).
- Query Volume: Over 2 million queries per day.
- Latency: The P99 latency for user-facing analytics is under 100 milliseconds, with average query times being even lower.
Open Source Contributions
In addition to using Pinot, Angel One has actively given back to the open-source community. Their engineering team has contributed to the Pinot codebase by improving the correctness and accuracy of Group By queries, enhancing multi-tenant support, and contributing to the roaring bitmaps repository used by Pinot and other databases.
Business Impact and Future Outlook
Apache Pinot’s high-load capabilities have enabled Angel One to build real-time features that would have otherwise cost significantly more and taken much longer to implement. Having started this journey just a year ago, Angel One plans to expand Pinot’s footprint to shape its entire business across all verticals
