In the late eighteenth century, the rulers of Prussia and Saxony had a problem. Their treasuries depended on revenue from the crown forests, but the forests themselves were unreadable. An old growth forest was a riot of species and ages, tangled with underbrush, crisscrossed by villagers who grazed animals in it, gathered firewood from it, foraged in it, hunted in it, and occasionally worshipped in it. From the point of view of a fiscal administrator trying to answer a simple question, how much timber revenue can we count on next year, the forest was a mess. So the administrators did what administrators do. They simplified it.
Thus was born scientific forestry, one of the proudest achievements of the age. Foresters reduced the bewildering diversity of the forest to the one variable the state cared about, the yield of commercial timber, and then rebuilt the forest itself to optimize that variable. Out went the mess. In its place came the production forest: a single species, usually Norway spruce or Scotch pine, planted in straight rows, all the same age, with the underbrush cleared and the deadwood removed. The new forest was a triumph of order. You could count it, measure it, predict it, and manage it from a desk. For the first time, the state could truly see its forest.
The first rotation of trees grew magnificently. The second did not. What nobody understood at the time was that the first generation had been living off the accumulated capital of the old forest, the deep soil built by centuries of decaying diversity. Once that capital was spent, the system collapsed. The cleared underbrush and vanished deadwood had housed the birds, insects, and fungi that built the soil. The same age, same species rows proved to be an all you can eat buffet for the pests that specialized in spruce, and a row of identical trees falls to a storm the way a row of dominoes falls to a finger. Yields dropped by 20 to 30 percent, and German forestry science had to coin a new word for the worst cases: Waldsterben, forest death. The forest that the state could finally see was a forest that could no longer live.
Legibility and its discontents
This story opens Seeing Like a State, the 1998 book by the late Yale political scientist and anthropologist James C. Scott, and it is the founding parable for the book’s big idea. Scott’s word for what the foresters did is legibility. States cannot govern what they cannot read, so they relentlessly simplify the world into standardized, countable form: censuses, cadastral maps, permanent surnames, standard weights and measures, grid shaped cities. Scott, who retold the forestry story in his own words years later, is careful to note that none of this is inherently sinister. Simplification is the price of administration at scale. Vaccination campaigns and famine relief depend on legibility just as much as taxation does.
The trouble starts when the people at the top forget that their simplified representation is a representation. Alfred Korzybski warned nearly a century ago that the map is not the territory, but legibility failures go one step further and one step worse. The state does not merely mistake the map for the territory. It starts bulldozing the territory to make it match the map. The forest is replanted in rows. The winding old city is demolished for a grid. The blogger Venkatesh Rao, in a widely read essay on Scott’s idea, distilled the recipe for this kind of failure: look at a complex, functioning system, fail to see the logic behind its apparent chaos, conclude that the chaos must be irrational, and impose a rationalized order in its place. The mess, it turns out, was load bearing.
The knowledge that does not fit in the system
What exactly gets destroyed when the mess is cleaned up? Scott has a word for that too. He calls it metis, borrowing the Greek term for the practical cunning of Odysseus, described by Homer as Polymētis. Metis is local, experiential, hard won knowledge that resists being written down: the feel a nurse has for a patient who is about to crash, the mechanic’s ear for a bearing that is about to fail, the engineer who knows that the deploy script works except during the Tuesday backup window, the account manager who knows which name on the customer’s org chart actually makes the decision. Every organization runs on metis, and the proof is what happens when it is withdrawn. One of the oldest weapons in the labor movement is the work to rule action, in which employees do exactly what the written rules and job descriptions say and nothing more. Production promptly grinds toward a halt. The formal system, followed to the letter, cannot actually run the operation. It never could.
Leaders are little states
It is tempting to read Scott’s book as a story about governments, but every executive is running a small state. We govern through the same instruments: org charts, KPIs, OKRs, headcount models, capacity plans, engagement surveys, dashboards. Like the state’s maps, these are indispensable. One cannot lead a company of hundreds or thousands of people through direct observation any more than a Prussian finance minister could walk every hectare of forest. And like the state’s maps, every one of these instruments is a radical simplification that leaves the metis out. The danger arrives on the day we start managing the company to make it match the dashboard.
Once you have the lens, you see legibility projects everywhere in corporate life. Return to office debates are fought with badge swipe data, because presence is legible and contribution is not. Engineering leaders count story points and pull requests, because those are countable and judgment is not. Sales organizations track calls made and emails sent, because activity is visible and relationship depth is not. Utilization targets treat people as interchangeable capacity, because capacity fits in a spreadsheet and the chemistry of a specific team does not. None of these instruments is evil, and I have used most of them. But each renders the organization readable by ignoring precisely the things that make it work, and each invites the same second step the Prussian foresters took: having measured the legible slice, we begin to manage as if the legible slice were the whole.
Consider one of the famous corporate legibility project of the last few decades: stack ranking at Microsoft. The system, popularized by Jack Welch at General Electric, forced every manager to sort every team onto a bell curve, a fixed percentage rated top, a fixed percentage rated average, a fixed percentage rated poor, regardless of how the team had actually performed. As a legibility device it was brilliant. Talent, that most tangled and illegible of forests, was rendered into tidy, comparable rows. Leadership could see the whole company’s performance at a glance. And like the Prussian forest, the scheme quietly destroyed the ecosystem underneath it. When Kurt Eichenwald investigated for Vanity Fair in 2012, every single current and former employee he interviewed named stack ranking “the most destructive process inside of Microsoft”. Engineers avoided joining teams with strong performers, because being ranked against stars meant landing at the bottom of the curve. Collaboration, the invisible soil of a software company, eroded. Employees competed with each other instead of with competitors. In November 2013, Microsoft abandoned the system, with human resources chief Lisa Brummel announcing there would be no more curve and no more ratings. The years that followed, under a new CEO and a culture rebuilt around learning rather than ranking, produced one of the great turnarounds in business history.
Notice the shape of the failure. Stack ranking did not fail because the data was wrong. Every ranking was dutifully, accurately recorded. It failed because the model was a fiction, a forced distribution imposed on teams that did not naturally contain one, and because managing to the fiction destroyed the real, illegible thing, trust between colleagues, that made the enterprise productive. The rows were straight. The forest was dying.
The first rotation problem
I wrote recently in Only Variety Beats Variety about Ashby’s law, the principle that a system must contain as much variety as the environment it is trying to manage. Legibility campaigns are variety destroyers by design, which is exactly why they feel so good to run. And in The Alarm That Went Silent I explored how instruments can fail without our noticing. Scott adds a darker layer to both. Even when your instruments work perfectly, they only ever report on the legible slice of your organization, and the act of optimizing that slice can consume the illegible machinery that makes the whole thing go.
The cruelest part of the forestry story is the timing. The first rotation thrived. For decades, the evidence said the scheme was working, because the new forest was silently drawing down capital nobody could see. Organizational legibility schemes behave the same way. The aggressive standardization shows immediate efficiency gains while it burns off accumulated trust. The rigid process rollout ships its first projects on time while the informal networks that used to catch problems quietly disband. The surveillance style productivity metrics look great for several quarters while your best people, the ones with the most options, update their resumes. The early returns validate the scheme right up until the soil is exhausted, and by then the people who planted it have often been promoted for their tidy rows.
Leading without flattening
The answer is not to burn the dashboards. Scott himself was clear that legibility is unavoidable; a leader who refuses to simplify cannot lead at scale at all. The answer is to hold your simplifications with appropriate humility and to deliberately protect what they cannot see. A few practices help.
Treat every metric as a compression, and ask what got compressed away. When a number moves, the question is not only why but also what this number cannot show me. Before optimizing anything to a target, name at least one vital thing the target is blind to, and decide how you will watch it.
Go to the territory. Toyota built this into its management system with the concept of gemba, the actual place where value is created, and the expectation that leaders regularly go there to see for themselves. Skip level conversations, sitting with support tickets, watching a real customer struggle with your product, joining an on-call rotation for a day. These are not nostalgic gestures. They are how you audit the map against the ground.
Protect pockets of productive illegibility. Somewhere in your company is a fixer whose real job appears on no org chart, a Slack channel where the actual coordination happens, a team whose messy process produces suspiciously good results. Before you standardize any of it, ask what the mess might know that the model does not. And when a new legibility scheme posts spectacular early numbers, remember the first rotation, and ask what stored and invisible capital it might be spending.
A challenge
This week, pick the single dashboard, metric, or report you rely on most, and go find the territory behind it. Sit with the people whose work becomes that number. Find the resident holder of metis and ask them what the report always gets wrong. Then come back and look at your tidy rows of data with new eyes. The Prussian foresters were not fools or villains. They were diligent professionals with excellent spreadsheets, and they killed the thing they were measuring. The forest your dashboards cannot see is the one keeping you alive. Walk in it.
I spent 4 years at Microsoft during the stack ranking era (I was there 2008-2012). Came in as a contractor, and opted to remain a contractor in large part because I saw the trauma that stack ranking imposed on FTEs.