Write a SQL query to compare experiment variants using grouped metrics and a window function for ranking.
Use the experiment_variants, experiment_assignments, and experiment_events tables. Calculate exposed users and distinct converting users for experiment 100, treating variants without assignments as having zero exposed users.
Output
- One row per variant in experiment 100.
- Columns:
variant_name, exposed_users, conversions, conversion_rate, and conversion_rank.
- Rank by conversion rate descending, with ties sharing a rank. Order by rank ascending, then variant ID.
