Your posting says the experimentation team wants to run more tests on bestbuy.com without waiting a month for each read. That is the problem I have spent most of the last two years on at Target, and I would like to bring what worked to your team.
On Target's site search team I have designed and analyzed more than 40 A/B tests on ranking changes. Many of them were underpowered at our usual traffic split, so I added CUPED, using each shopper's pre-period behavior as a covariate, and our typical test length went from 21 days to 14 with the same error rates. I also wrote the readout template the team still uses, which puts the decision and the confidence interval before any chart.
Not every question can be randomized. When merchandising changed prices in one region, I estimated the sales effect with difference-in-differences across 212 stores, checked the pre-trends with the finance team, and they used the estimate to set fourth-quarter pricing. Before Target, I spent three years at Optum building readmission risk models on claims data, which taught me to document assumptions for reviewers who were not statisticians.
I work mainly in Python and SQL, with Snowflake and Airflow, and I am comfortable putting models into production alongside platform engineers. If it would help, I can walk through one of the search tests from design to decision in a first conversation.