Part 1: case study plus SQL and Python.
The case: an e-commerce company wants to send personalized emails to boost conversion rate. How do you measure success? It was basically the standard A/B test process — define the metrics, figure out how to calculate the sample size, all of that. There was a follow-up at the end: suppose we ran the experiment and saw conversion rate go up 20%, we thought it was a huge success, so we launched it to everyone. Then a new marketing director came in and decided to rerun the experiment, and this time conversion rate only went up 2%. What happened, and what angles or data would you look at to figure it out? I got stuck on this one — curious what everyone else thinks.
For SQL, there were two questions. The first one: given a table with a column of ids like 1, 2, 15, 14, 13, and so on, sum up all the numbers that are smaller than or equal to the current row.
The second question had three parts, based on a user_txn table with columns like user_id, user_region, adopted_at, transacted_at, timestamp, transacted_region, and a few more I don't remember. The first part asked about adoption_rate and transaction_rate. The second part asked for the time from adoption to the first transaction. The third part was about cross-region sales — defined as a transaction where the region it happened in is different from the region of the user's first transaction. I had to find those transactions.
For Python, they basically gave me a dataset and had me do EDA live, build a stats model, then analyze the model's performance and talk about what could be improved.
Discussion
Loading comments…