How to improve complaint resolution

You are given three related tables for a Wayfair-style e-commerce customer-support dataset. `complaints` - `complaint_id` STRING - `customer_id` STRING - `order_id` STRING - `product_category` STRING - `issue_type` STRING - `complaint_channel` STRING - `complaint_created_at` TIMESTAMP - `region` STRING `resolutions` - `complaint_id` STRING - `resolution_type` STRING - `first_response_at` TIMESTAMP - `resolved_at` TIMESTAMP - `agent_id` STRING - `escalated_flag` BOOLEAN - `refund_amount` DECIMAL(10,2) - `replacement_sent_flag` BOOLEAN `service_ratings` - `complaint_id` STRING - `customer_service_rating` INT Rating is from 1 to 10, where 10 is the highest satisfaction. - `rating_submitted_at` TIMESTAMP Each `complaint_id` refers to one customer complaint and may or may not have a submitted rating. You do **not** have a full orders table, so be careful to distinguish complaint **volume** from complaint **rate**. The business goal is to improve customer experience while controlling support and refund costs. Answer the following: 1. Provide 3 data-driven insights you would look for in the data. 2. Provide 3 recommendations linked to those insights. 3. Choose 1 prioritized recommendation and explain why it should be implemented first. 4. Explain how you would measure success after rollout. Specify primary metrics, guardrail metrics, and whether you would use an A/B test, phased rollout, or quasi-experimental design. 5. Discuss confounding, selection bias in the ratings data, and key implementation risks.