Did Microservices Break DORA? • Dave Farley • GOTO 2024

Speed and quality are not trade-offs - you need speed for quality and quality to work at speed, and this approach is actually more cost-effective

Teams practicing loose-coupled architecture report higher burnout, likely due to the increased complexity and expertise required to implement microservices properly

The DORA metrics (stability and throughput) are the most reliable and universally applicable measurements for software delivery performance

High performers are significantly more likely to use version control (33%), practice continuous integration (39%), and implement continuous delivery (46%)

This year’s report showed surprising results where teams with less experience (under 16 years) doing both CI and CD performed worse than those only doing CD

Cloud usage predicts 14% higher chance of exceeding organizational goals, though the report lacks clear evidence comparing cloud vs non-cloud performance

True microservices depend on loose coupling, but many teams implement distributed systems they call microservices without proper decoupling

Security scanning and practices positively predict both organizational performance and software delivery performance, but require strong CI/CD capabilities

Most metrics become harmful when treated as targets rather than indicators - teams will game the system to hit targets

The DORA research approach provides valuable scientific evidence to guide software development practices, based on over 33,000 surveys