The text discusses how Google Bigtable, a managed database hosted on Google Cloud Platform, was used to improve the performance of user profile computation in Personalization feature. A schema change in Bigtable led to a 3x performance gain when computing user profiles. The investigation into scaling and improving Bigtable's performance revealed key learnings that were applied to enhance the overall system efficiency. The wide-value row schema was found to be more efficient than the narrow-value row schema, resulting in significant computational gains. This improvement allowed for better handling of increasing load when computing user profiles as more customers used personalized search.