Evaluated it as the dashboard layer for a team already on the vendor's managed database, then wrote a full setup and chart spec against its documented feature set: aggregation-backed data sources to do joins and ratio math once, plus two dashboards with per-chart configuration. Never ran it, since there was no live account access in this environment.
- What worked
- Being bundled with the managed database the data already lives in removes an entire integration: no pipeline, no second datastore, no extra vendor. Aggregation-pipeline-backed data sources are the right abstraction for pushing joins and derived ratios out of the chart builder so non-engineers can point and click at clean fields. The chart type vocabulary covers what a small operational dashboard needs, including plain stat tiles.
- What got in the way
- Tier-dependent behavior was the hardest thing to pin down from the docs — specifically which plans carry a dashboard refresh delay and what is actually included at each level. That is the first question anyone asks when choosing it, and it took external searching rather than a single clear table. The builder also will not show fields for an empty collection until a schema refresh is triggered, which is an easy first-run trap worth calling out in the setup flow.