Overview

Could not load KPI summary. 404 Not found: Table project-ad7a5be2-a1c7-4510-82d:dev_claims_analytics.fct_claims was not found in location us-central1; reason: notFound, message: Not found: Table project-ad7a5be2-a1c7-4510-82d:dev_claims_analytics.fct_claims was not found in location us-central1 Location: us-central1 Job ID: b79b613f-f169-404d-8065-506e472c4f7b
Overhead view of a desk: a loss development triangle worked out by hand on graph paper, a mechanical pencil, an enamel mug, and a closed laptop with a coiled cable.

An actuary who builds his own data infrastructure

I trained as an actuary. Loss development, IBNR, pure premium: that is the mathematics this dashboard is doing. What sits underneath it is engineering. A warehouse, an orchestrated pipeline, a streaming intake, a pricing model, all on Google Cloud, all built here.

In insurance those two skills usually live in two different people, and the handoff between them is where most of the time goes. Keeping them in one pair of hands is the point of this project. You cannot design a claims table well without knowing what a development year is, and a reserve you have no way to query is not a reserve you can defend.

If the vocabulary here is unfamiliar, Methodology walks through the whole platform and shows how an actuarial question turns into a table. Every page below opens by saying what it is for and how to read it.

Every number below comes from a live SQL aggregation against the dev_claims_analytics and dev_claims_reports datasets in BigQuery. The portfolio covers five lines of business across all 32 states of Mexico.