case study: 对将要launch shopping feature在pins界面上，metrics咋选，实验怎么design，别的metrics drop了怎么给pm推荐\n. .и\n指标上从 success metrics (DAU, user time spent)，intermediate metrics (num_clicks_of_new_feature，stay_time_in_shopping）考虑\n\n实验design提到了spill over effect, learning effect 怎么解决\n\ncoding: sql: 忘了是啥了，反正不难。。python：用到了dict mapping，dataframe/series sort，null value handling\n\n不给stackflow语法强行回忆。。这年头都chatgpt写代码了差点以为挂在coding轮