I did this write up of a recent dbt meetup talk by an analytics engineer from Delivery Hero. They had dbt tests getting out of control triggering hundreds of alerts. Their solution is a mix of categorizing critical models, weighting alerts, and formalizing the response.<p>The video is here if you want to watch that directly, instead:
<a href="https://www.youtube.com/watch?v=Nk_K8mW-N9A" rel="nofollow">https://www.youtube.com/watch?v=Nk_K8mW-N9A</a><p>The whole subject of dbt tests, bloat or rot, is an interesting one. Here's a few stats from a couple of public facing dbt project:<p>- Mattermost: 194 models / 318 tests
- Cal-ITP (California Integrated Travel Project): 361 models / 941 tests<p>What's the model/test ratio in your dbt project? How do you handle alerts?