When Good Metrics Go Bad
For decades, engineering leaders have tried to quantify the value of their teams. This led to a generation of metrics based on activity: lines of code (LOC), commit frequency, and story points. These were always flawed, but they offered a semblance of measurement.
The core assumption was that more activity equalled more progress. In this model, a developer who wrote 1,000 lines of code was seemingly more productive than one who wrote 100. However, this often incentivized the wrong behaviours, rewarding verbose code over elegant, efficient solutions. As Bill Gates reportedly said, measuring software progress by lines of code is like measuring aircraft manufacturing by weight. These metrics were designed to be simple and countable, but they were never a true measure of value.
How Generative AI Breaks the Model
Generative AI has turned these flawed metrics into an absurdity. With tools like GitHub Copilot, developers can now generate thousands of lines of code in minutes, causing metrics like LOC and commit frequency to skyrocket. This creates the illusion of hyper-productivity without delivering any real business value. In fact, it can create downstream problems. A flood of AI-generated code can lead to bottlenecks in code review, integration, and testing. If you speed up one part of an assembly line without upgrading the rest, you don't get a faster factory—you get a pile-up. Some reports even show that while individual productivity feels higher, overall software delivery performance and stability can decrease as AI adoption increases, because teams are creating larger, harder-to-review batches of code.
The Shift from Output to Outcome
The new imperative for software teams is to prove their value by focusing on business outcomes, not engineering activity. The most effective engineering organizations are shifting their measurement systems away from what individuals produce and toward what the business achieves. The conversation is no longer about how much code was written, but about what impact that code had. Did it increase revenue? Did it improve customer satisfaction? Did it reduce operational costs? Teams that can draw a direct line from their work to these top-level business goals will be the ones that thrive. This requires a fundamental change in mindset, moving away from seeing the engineering department as a cost center and toward viewing it as a value engine.
The New Engineering Scorecard
Instead of vanity metrics, modern teams are adopting frameworks that measure system-level performance. The DORA metrics are a popular starting point, focusing on four key areas: Deployment Frequency (how often you release to production), Lead Time for Changes (how long it takes to get a commit to production), Change Failure Rate (what percentage of changes cause a failure), and Mean Time to Recovery (how quickly you can recover from a failure). Other frameworks like SPACE and DX Core 4 add layers to measure developer satisfaction, well-being, and collaboration. The common thread is a focus on speed, stability, and efficiency, which are much closer proxies for business value than lines of code ever were. The ultimate goal is to connect these engineering KPIs directly to business impact, such as revenue generated by a new AI-powered feature.
Redefining the Developer's Role
In this new era, the most valuable engineering skill isn't just writing code. It's the ability to effectively guide and collaborate with AI. This includes skills like prompt engineering, critically reviewing AI-generated output, and having the deep systems-level knowledge to know whether an AI's suggestion is brilliant or dangerously flawed. The focus shifts from manual labor to strategic oversight. Developers are becoming less like factory workers on an assembly line and more like pilots orchestrating a highly advanced machine. The human element—critical thinking, problem-solving, and a deep understanding of user needs—becomes more important than ever. Companies that recognize this will invest in training and create a culture that values judgment over pure output.
