The Metric That Incentivizes Waste in the Name of Productivity

Recently, I was talking to a friend who works as a software engineer, and he mentioned that his company had started evaluating performance based on token consumption. That conversation happened about two weeks ago, right around the time I started writing this piece.

My immediate thought was how quickly this could turn into a perverse incentive over time. Apparently, things are moving even faster than I expected. Just this week alone, I came across a video from OPrimoDev and articles from UOL and Ars Technica discussing this exact effect.

And this is far from an isolated case. I personally know of more than a dozen companies adopting this metric, many of which have already incorporated it into performance review cycles. While I do think this can be a useful indicator for measuring engagement or adoption, it is a terrible way to measure performance. At the end of the day, we are measuring activity instead of outcomes or value creation.

Metrics are not neutral. They shape behavior, often faster and more effectively than culture decks, internal speeches, or leadership messaging ever could.

Metrics are not neutral. They shape behavior, often faster and more effectively than culture decks, internal speeches, or leadership messaging ever could. Once an organization starts tracking usage volume as a performance indicator, it implicitly sends a message: using more matters more than using better.

The practical outcome is fairly predictable. People start interacting with tools even when there is no clear need, spend less time refining their inputs, repeat unnecessary cycles, and optimize for volume over efficiency. Not because they are bad professionals, but because they are responding exactly to the incentive system in place.

This kind of metric also creates a meaningful economic distortion. Instead of pushing the organization toward efficiency gains, it can encourage rising operational costs without a proportional increase in value delivered. It is a subtle but dangerous dynamic: the metric rises, the perception of productivity improves, but the actual impact remains flat or, in more extreme cases, declines.

The core issue is that effort-based or activity-based indicators are rarely good measures of performance. The number of commits, pull requests opened, tickets resolved, or tokens consumed is, fundamentally, an activity metric. In isolation, they tell us very little about real performance. The measurement layer needs to be elsewhere in the actual business impact. Everything beyond that is complementary.

Poorly designed metrics do more than create misleading interpretations. They also encourage behaviors that move organizations in the opposite direction of what they actually want to build. At some point, "productivity theater" starts replacing real productivity.

In the end, this discussion is not really about the technology itself. It is about the incentives we are creating for how these tools are used. More than anything, it is about distinguishing what is easy to measure from what actually matters.

And sooner or later, every company becomes a reflection of what it chooses to incentivize.