SafetyCulture gets serious about company OKRs with dbt Cloud
SafetyCulture's data transformation layer was split across LookML and Airflow with no testing or dependency management, leaving stakeholders confused about where to find reliable data. Mistrust and rework slowed time-to-insights, while 97% of first-time users dropped off after 28 days without an analytical explanation.
How it works
Common implementation structure
How this type of workflow is generally built, generalized across documented cases — not tied to any one vendor's stack. Click any stage to read what happens there. Specific products that implement these stages appear in “Tools commonly seen” below.
Stage 1 · Scalability limit triggers rebuild
LookML's lack of testing, DAG support, and scalability triggered a move to rebuild the data transformation stack.
SafetyCulture rebuilt 80% of their data in dbt and cut average pipeline query time from 17 to 4 minutes. Data team eNPS rose from -20 to +69, an 89-point increase. New customer retention beyond 28 days more than doubled from 3 to 6.5%, with a projected 40% increase in MAU.
What failed first
An attempt to schedule and orchestrate LookML models with Airflow quickly ran into accessibility and architecture challenges: the team had to manually build all dependency graphs and ended up with the transformation layer across two tools that only a few people could operate.