Engineering KPIs

In a previous blog post I talked about the dangers of using ‘velocity’ as a KPI for Engineering, but what are good alternatives?

I propose that the following perennial KPIs measure a lot of what matter about Engineering success (in a SaaS/web app environment):

• Uptime
• Mean Time To Discover
• Mean Time To Resolution
• Deploy Duration
• Rollback Duration
• Ticket Cycle Time
• Build and Automated Test Duration
• Number of Deploys Per Unit of Time
• Incoming Production Defect Count Per Unit of Time

As with any metric, they don’t tell the full story about underlying causes, but :

• are easy to collect
• are leading indicators of potential problems
• have material business impact if they trend in the wrong direction

Note that these are separate from, and parallel to, project delivery and OKR metrics — the engineering KPIs are ones that are wholly or largely in Engineering’s control and are relevant regardless of what is being delivered.