Legibility theater
When visibility determines what gets valued, organizations get better at looking productive.
Some of the most valuable engineering work is invisible to the systems that track engineering work. The engineer who closes twelve tickets shows up on every dashboard. The one who talked a product manager out of a feature that would have created six months of maintenance debt? Nearly invisible.
Organizations notice the first kind of work more, reward it more, and get more of it.
This isn’t a management failure. It’s a measurement problem that feels like it’s working. The dashboard is green. Velocity is up. Cycle time is down. Everything that can be counted is improving. The things that can’t be counted are quietly decaying, and there’s no dashboard for that.
The mechanism is simple. Leadership needs to understand what’s happening across teams they can’t directly observe. So they ask for artifacts: tickets, story points, pull requests, status updates. Reasonable enough. But the act of asking reshapes what gets done. Work that produces visible artifacts gets prioritized. Work that doesn’t gets squeezed into the margins. Design quality, knowledge transfer, the slow accumulation of shared understanding.
No one sends a memo saying “stop mentoring juniors and write more tickets instead.” It happens through a thousand small signals. Who gets praised in all-hands. What gets discussed in planning. Which work shows up in promotion packets. People are smart. They read the room.
The second-order effect is worse than the first. It’s not just that invisible work gets deprioritized. It’s that the people who were good at it adapt or leave. The senior engineer who spent half her time on design quality and knowledge transfer notices that none of it registers. She starts optimizing for what does. Or she finds somewhere that values what she’s good at. Either way, the organization loses a capability it didn’t know it was depending on. The metrics don’t capture the loss because they never captured the contribution.
And the loss is silent. Nothing on the dashboard gets worse. Tickets still close. A year or two later something breaks that used to not break, and it gets attributed to a reorg, a bad hire, a tough quarter. The cause never surfaces because the regime that caused it never had a name for what it lost.
Cities do this too. Optimize for what planners can see from above and you lose the messy knowledge embedded in how neighborhoods actually function. The street vendor who knows every regular. The implicit social agreements about shared space. These get rationalized away because they don’t fit the map. The map gets cleaner. The territory gets worse.
Organizations build their own maps. The board, the metrics, the rituals. Then they optimize for the map instead of the territory, and the two drift apart so gradually that no one notices until something breaks.
You can’t just stop measuring, though. Without visibility, you get a different failure mode: no accountability, no coordination, priorities drifting to whoever talks loudest. Most operators have lived both sides. They’ve worked at places where metrics ran everything and watched valuable work evaporate. They’ve worked at places with no metrics and watched the work drift into whatever the most persistent person wanted. Neither version is fine.
The deeper issue isn’t measurement. It’s the distance between decision-makers and the work. Small teams don’t have this problem. Everyone sees everyone else operate, and the invisible work is visible to the people working alongside it. The problem emerges at scale, when leadership needs to understand work it can’t directly observe.
Organizations that handle this well keep judgment in the loop. Metrics are signals for where to look, not the thing being managed. Leaders sit in design reviews, read the post-mortems, ask senior engineers who’s actually carrying the work. The dashboard is a starting point for investigation, not a verdict. This doesn’t scale as cleanly as a number. That’s exactly why the cheaper substitute keeps winning.
Most organizations, if they notice the problem at all, try to solve it by measuring more things. Add a metric for code review quality. Track mentoring hours. Score knowledge sharing. But this just expands the theater. Now people perform legible versions of the previously illegible work. The mentoring session gets a ticket. The design discussion gets a document. The artifact exists. Whether the underlying thing happened well is as invisible as ever.