Automated Ad Bid Optimizer
Technology:
Python, Spark, Google Cloud Storage, Serverless computing, PID Algorithm
This company invested millions in marketing placement in different website. One of these websites has an intricate bidding system where the highest bidder gains the higher position in the ad placement. However, after several trials they figured out that depending on certain periods, higher investment doesn’t lead directly to higher placement. Without a clear understanding of how the ad algorithm works, their investments often felt like a shot in the dark. They sought a smarter investment strategy, one that wasn't flat across all periods but was dynamically adjusted based on data.
We crafted a predictive model designed to suggest bid prices taking into account seasonal factors as well as daily fluctuations in demand. By leveraging a lean model design and deploying it on Google Cloud, the solution was up very quickly for testing. Its direct impact was twofold: it allowed for more informed spending on ads in over 50+ markets and freed up manpower previously caught up in the tedious bidding process. An internal dashboard provided a clear comparison between our model and the old manual methods.