String Match of User Comments
When the organization's data platform didn't support Machine Learning, analysts had to download all the text data and run the model manually on the local desktop monthly, which is not cost effective. I came up with an innovative way to automate the reporting.
​
1. I gathered insights from product managers/ user researchers/ designers/ engineers to understand what issues they would like to address and came up a list of key words that user may mention when they experience certain difficulties. The key word will trigger the mention of the topic and shown as the mentioning rate of each topic to represent urgency of user pain point.
2. I then configured the survey database as the following, scheduled data refresh and oozie exported the tbl_survey_comment table to the data warehouse.
3. I also connected the above mentioned table to the Tableau server and created a self-served dashboard so that different teams can find the mentioning rate of tailored topics. I also connected the survey data with product data, salesforce data so that different teams could filter the result based on their needs. Users can also simply click on certain data point, and tableau will direct them to the physical comments immediately.
This dashboard has been the most frequently viewed project in the product analytics team since built. It provides easy access to a variety of stakeholders to understand users' needs and wants, which is key for user experience success.
This dashboard also inspired a cross team collaboration between the product analytics team and data science team to build and train ML categorization models. I led the part of data warehousing and providing BI solutions.