Design the analytics service that powers creator dashboards — impressions, reach, engagement rate, follower growth, and audience breakdowns — over selectable time ranges and per-post detail.

Operating context. Engagement events (impression, like, comment, share, profile visit) stream in at very high volume, each attributed to a creator's content. Creators open an interactive dashboard and query aggregates over ranges like today, 7 days, and 28 days, sliced by dimensions such as geography, follower vs non-follower, and content type. Raw events are enormous, so dashboards read pre-aggregated rollups rather than scanning raw data. Some events arrive late and must still land in the correct time bucket.

Out of scope. The ad-revenue and payout system, the live real-time view counter shown on an individual post, machine-learning audience predictions, and BI/data-export connectors. Assume other teams own those.

What to produce. A high-level design covering: event ingestion and attribution, the pre-aggregation strategy across time and dimensions, the storage choice that keeps dashboard queries interactive, how late-arriving and duplicate events are folded in, and the serving and caching path for the dashboard. Sketch the components and data flow; checkpoints will probe the rollup design and the query storage choice.