Challenge
Ströer, one of Germany's largest out-of-home and digital media companies, owns a supply side platform that processes millions of ads a day. With campaigns arriving from multiple sources, each with different priorities, types and schedules, pacing them is a non-trivial task. Ströer was looking for a partner who could help them understand their own data and let their AdOps team make sound decisions on scheduling a growing number of campaigns, increasing revenue without risking campaigns going unfulfilled.
Solution
We analysed Ströer's raw data and built a data aggregation pipeline that cleans it and extracts the most valuable signals for further processing. On top of that we gave them thorough documentation and new insight into their own data and traffic patterns. We then combined simple statistical models with neural networks (DeepAR) into a forecasting engine their AdOps team uses to predict future campaign spend from historical data.
Results
Ströer gained additional insight into their production traffic and secured further growth: forecasting now supports the decisions AdOps makes every day and lets the business operate at even larger scale.
