Metrics and Telemetry Framework
Replaced a fragmented instrumentation system with a unified framework that the team could trust.
Stage.in's data team was spending a significant portion of its time debugging reporting discrepancies rather than building. Without a unified instrumentation and metrics framework, errors were difficult to trace, reporting inconsistencies were common, and the team lacked confidence in the numbers they were producing.
Design a metrics and telemetry framework that would give the data team a single, reliable layer for all instrumentation and reporting, dramatically reducing the time lost to debugging and the volume of errors reaching stakeholders.
The foundations laid at Stage.in were the equivalent of what had been built at Blinkit, with one crucial difference: they were designed from the ground up rather than retrofitted onto a legacy system. At Blinkit, the instrumentation and modelling overhaul had to contend with years of accumulated technical debt and organisational inertia around existing definitions. At Stage.in, starting from scratch meant the canonical event schema, metric definitions, and validation rules could be established as first principles before any data began flowing through them — producing a much cleaner result with the same level of effort.
- 01Designed the unified event schema from scratch: standard naming conventions, required vs optional properties, and canonical metric definitions covering all key product and business surfaces.
- 02Implemented Rudderstack as the instrumentation layer with automated validation at ingestion — schema tests running at point of collection and alerting on malformed or out-of-schema events.
- 03Built core dbt transformation models establishing clean, documented fact and dimension structures from the validated event data.
- 04Configured Metabase dashboards on top of the modelled data, giving content, product, and executive teams reliable self-serve reporting from day one.
- 05Documented the full instrumentation framework so engineering teams could add new events independently, against the schema, without data team involvement.
- 06Established a review process for new metric definitions: proposed events reviewed before instrumentation, preventing schema drift at the source.
- 90% reduction in time spent debugging reporting issues
- 50%+ reduction in reporting errors reaching stakeholders
- Data team redirected from firefighting to building
- Engineering teams able to instrument new events independently against the schema, without data team involvement
Start a conversation.
Every engagement begins with a focused discussion of your current data environment and priorities. To schedule an initial consultation, reach out directly.
Get in touch