Professional Services business metrics
How can a business metric stay explainable all the way down to the records behind it?
- Business data
- Batched refresh
- Metric storage
- Cohort analysis
- Record drill-down
The product
I worked on an analytics application for Professional Services leaders, with metric cards, historical views, cohort filters, and account/opportunity drill-downs. The goal was to make the selected population consistent across every view.
Architecture shaped by access
The project started with a Workers architecture, moved to a functional Gadget implementation while data access was constrained, and returned to Workers when the required opportunity data became available. The later design used D1 storage and Workflows for batched retrieval and ingestion.
The difficult part
Reliable comparisons require shared metric definitions, consistent cohort scope, and explicit treatment of small samples. A configuration-driven catalog was intended to make future metrics easier to add without rebuilding the application.
State at handoff
The core product and data architecture existed, but the complete Workers refresh path still needed end-to-end validation and production hardening. AI summaries and the fully verified metric/cohort builders remained future work. This account is based on my retrospective handoff.