6. Goals
Maximize data value using
domain knowledge
Enable fast decision making
based on data
Reduce communications costs
Build a scalable data
architecture
7. Principles
Business context immersion and domain
knowledge building
Data is a first-class citizen in product
development and process design
Autonomy to build data pipelines
Responsibility for data quality
9. Data in product and processes
Business analysts in groomings and
plannings
Business analysts are responsible for the
squad outcome
Capturing data is not always cheap
10. Autonomy and responsibility
Transparency: all analysts know where the data
come from
Standards: we have guidelines for data
transformation and modeling
Peer-review: pull requests and code review
guarantee consistency and knowledge dissemination
Testing: transformations are tested, preventing
outages and wrong data
Simplicity: minimize number of tools
12. Raw data Master data
Transformation
Layer
Data Modeling
Data Consumption
Layer
Ingestion Layer
BigQuery
Federated Data
Source
BigQuery
Data Transfer
Service
20. Pros
- BAs have context, are independent, faster
and more effective
- Data engineers focus on platform instead
of ETL
- Engineering doesn’t bottleneck data
analysis
- Transparency. Everyone knows where the
data come from and how.
- People have access to all data and
transformations, avoid duplicate tables
Cons
- BAs are alone in the squads, require
chapter processes for knowledge
sharing
- BAs focused on squad context, could
miss big picture
- BAs code a lot. Requires training and
processes for scalability
- Works better with at least 1 BA per
squad
- Cloud costs are hard to track
Conclusion