Ruth Garcia presented on using simple machine learning models in an ads manager. Online advertising spending for mobile has grown significantly, with a 76.8% compound annual growth rate for mobile compared to 15.4% overall. The ads manager aims to balance increasing revenue and engaging users. Various machine learning models were considered for click prediction, including logistic regression, random forests, and neural networks. Challenges addressed categorical values through one-hot encoding and hashing tricks. Model performance was evaluated offline using metrics like precision at 1, mean reciprocal rank, and AUC. The talk concluded with lessons on starting lean, communicating machine learning requirements upfront, and balancing exploitation and exploration in ads delivery.