SLAs, SLOs and alerting

A data platform has internal customers, and customers want commitments. The posting talks about data work that is reliable and easy to build on, and about tools that solve real stakeholder needs. One way to answer that is to run the platform like a product, and SLAs are how you write the promise down.

The junior mistake is one blanket promise for everything. The experienced move is a tiered promise, because agent-training data and revenue numbers do not deserve the same reliability budget as an analyst's exploratory scratch table.