Case Study

Microsoft Prediction Lab

A new platform re-invents survey research, allowing the public to make predictions about upcoming events and view the combined predictions of the crowd.

Microsoft Research desired to build a platform for predictive markets and polling, where users can submit their views and predictions regarding politics, sports and other subjects. With this entry into public surveys and polling, Microsoft realizes the first phase of their vision to revolutionize prediction market technologies via devices and artificial intelligence.


Microsoft Research had done some early proofs of concept to validate that their engine was correctly predicting outcomes. There were other, gambling-focused prediction engines on the market; by contrast, this site would take a gamification approach whose goal was to make predictions with a high level of confidence.

Microsoft Research quickly learned that the engine was sound, and that exploiting this engine’s power could be a way to monetize polling. But the product was not commercial-grade, and it needed to scale to be able to support millions of users. The data needed to be rationalized and migrated to the cloud. The team knew that in order to succeed, the engine needed a compelling user interface. Finally, targeting use for the November 2014 election cycle, the project needed to be completed within an extremely tight timeframe.


Microsoft Research chose BlueMetal for the architectural expertise to create commercial-grade software that could support millions of users, and for BlueMetal's ability to design a sticky, engaging user experience that is consumer-grade. BlueMetal's approach to user-led design changed the team's perspective about the project, resulting in more gamification of the site to make it more rewarding for the users. For this group of dedicated researchers, BlueMetal focused on building this software to a rigorous and truly iterative process, with cross-disciplinary teams, agile development methodology, and a high level of interaction with the client. The team delivered the project in just fifteen weeks.

The End User Website is an AngularJS Single Page App for cross-platform responsiveness. The pre-existing market engine is built in Java, and the solution is built on SQL Azure and Azure Cloud Services. The site leverages federated authentication, so users don't need to remember a separate password.

This solution allows Microsoft Research to prove out scalability during the upcoming midterm elections. In the future this could be used for a limitless variety of predictions. The website is just the first phase in a range of experiences that could encompass devices, Xbox, and other leading-edge platforms.

"It was a pleasure working with BlueMetal. They were professional, efficient with their time, and most important, built an engaging and sustainable platform with a tight timeframe."

David Rothschild, Economist, Microsoft Research