Concord's data science team engaged to analyze the current state and develop a plan to make long-term keyword management more profitable and manageable. Our team implemented a combination of supervised and unsupervised learning data science models. For all uncategorized terms, the supervised learning model suggests existing categories and the unsupervised learning model suggests new categories. This revolutionary approach allows our client to focus their time on approving new term assignments and category suggestions, revealing a wealth of profitable new business categories for their search marketing program.
Our solution featured a simple interface using Google Cloud and Google Sheets powered by AWS Athena and Databricks to process the data and run models.
We successfully improved our client's keyword search categorization to reduce manual effort, save cost, and identify profitable new search categories.
- Developed new algorithm that sorts tens of thousands of keyword terms in under five minutes
- Saved the marketing and analytics team over 120 hours of weekly effort spent on the problem
- Discovered two new profitable topics while running the initial test algorithm