We ran the OPO system on real traffic slices of a search advertising system. The goal was to improve system metrics
over the current parameter value (tuned using traditional methods) as much as possible. The objective function was a
formulaic representation of revenue and user experience on a search engine. The traffic slices were based on partitions
of search engine user space: each slice received traffic from about 200,000 users per day. The system ran for two weeks,
trying to optimize a parameter known to have a significant effect on search engine performance. The optimal parameters
generated by the OPO system demonstrated reasonable improvements in the primary measures of system performance:
revenue by 0.25% and click yield by 1.4%