Design an analytics dashboard for an AI chat app

Design an analytics dashboard for an AI chat application similar to ChatGPT. The dashboard will be used by Product, Engineering, and Support teams to track overall system health, user behavior, and business performance. Core capabilities Near real-time visibility into: Usage DAU / WAU / MAU Active sessions Messages per user Retention and cohort analysis Performance End-to-end latency (P50 / P95 / P99) Time-to-first-token Streaming token latency Reliability Error and timeout rates Rate-limit events Quality signals (proxies) User feedback (thumbs up / down) Conversation abandonment rates Cost Tokens generated Estimated cost by model and tenant Metrics should be filterable by model, region, platform, tenant, and application version, with support for common time ranges (last 15 minutes, 1 hour, 24 hours, and custom ranges) and drill-down exploration. Non-functional requirements Data freshness Operational metrics: near real time (1–5 minutes) Business and retention metrics: hourly or daily updates acceptable Reliability & availability High availability with graceful degradation under partial failures Security & privacy Proper handling of PII Role-based access control Auditing and access logs Deliverable Propose a complete end-to-end system design, including: Instrumentation and event collection Data ingestion pipelines Storage and aggregation layers Query patterns and access paths Dashboard APIs and serving layer