How good ideas become bad systems
Many systems that fail people today began by solving a real problem.
Pharmacy benefit managers began as a coordination layer: a way to process claims, manage formularies, negotiate with drug manufacturers, and reduce administrative friction. That was not fake value. It was useful. But over time, the same machinery that made pharmacy benefits legible also made them easier to optimize around rebates, preferred networks, prior authorization rules, spread pricing, and opaque contract terms. The problem is not that the original idea was rotten. The problem is that a useful function became an insulated system.
That pattern repeats.
A term clarifies reality. A profession forms around it. Metrics emerge. Infrastructure hardens. Participants begin defending the system as if they are defending themselves. Eventually, the system may begin consuming the very substrate it depends on.
This essay is a diagnostic map, not a sequence. The six pressures below do not always arrive in order. A system can become guilded before it becomes entrenched. Identity can fuse before the metrics fully drift. What matters is the profile: which pressures are active, how intensely, and what kind of intervention the system still allows.
The claim is not that institutions are evil. It is that helpful systems tend to protect themselves, and that self-protection can outlive the good the system was built to serve.
What follows are six pressures of institutional self-insulation.
1. Naming
A real phenomenon gets rendered legible. Someone gives it a word.
This is the honest stage. A pharmacist notices that formulary management can reduce waste. An educator observes that structured feedback improves learning. A therapist identifies that trauma produces measurable changes in behavior and physiology. A compliance officer names a recurring failure mode. A startup names a workflow that had previously lived only in email, spreadsheets, and tribal memory.
The abstraction maps to reality. The word does work in the world.
Naming is not the problem. Naming is the precondition for every system that ever helped anyone. A problem cannot be improved until it can be seen, and it often cannot be seen until someone gives it language.
The pressure begins later, when the name stops pointing outward at the reality and starts pointing inward at the profession that formed around it. The term becomes less a tool for understanding and more a membership badge.
Probe: What problem did this term originally help people solve? Does it still solve that problem, or does it now primarily organize a profession?
2. Guilding
The vocabulary thickens.
A specialized language emerges to describe the work, then to describe descriptions of the work, then to credential who is allowed to use the descriptions. Some of this is necessary. Expertise always compresses complexity. A cardiologist, actuary, pharmacist, trial lawyer, or software engineer needs terms that outsiders do not use every day.
But expertise and guilding are different.
Expertise compresses reality so people can act on it faster. Guilding compresses reality so outsiders cannot act on it at all.
In pharmacy benefits, this looks like the layered vocabulary of formulary tiering, MAC pricing, rebate aggregation, specialty carve-outs, accumulator programs, utilization management, prior authorization criteria, medical-vs-pharmacy benefit routing, and spread-retention mechanics. Each term may describe something real. Together, they create a system in which an intelligent employer, patient, physician, or policymaker often cannot understand the transaction without hiring an interpreter.
The same pressure appears in academic peer review. Double-blind review, response-to-reviewer letters, impact factors, tenure files, journal hierarchy, and disciplinary conventions began as quality-control mechanisms. But in many fields, they also became a credentialing economy. Outsiders cannot publish in the venues that count without first becoming insiders, and becoming an insider means absorbing a vocabulary whose mastery is itself the gate.
Guilding is not always malicious. Often it is just accumulated professional convenience. But the effect is the same: the system becomes harder to question because the language of questioning belongs to the people who benefit from the system.
Probe: Can an intelligent outsider understand this system without hiring an insider to explain it?
3. Metric Substitution
The system begins measuring what it can measure rather than what it set out to do.
Then the metric drifts loose from the mission.
Then the metric becomes the mission.
A prior authorization process exists, in theory, to prevent inappropriate use, reduce waste, protect patients, and preserve plan resources. But the operational system may end up optimizing for turnaround time, denial-rate management, documentation completion, queue closure, or audit defensibility. Those metrics are not meaningless. They matter. But none of them is identical to the original purpose.
The same drift shows up in research oversight. Institutional review boards exist to protect human subjects from research harm. Over time, the measurable artifact can become the proof of protection: completed forms, documented consent, approval letters, training attestations, protocol language. The revised Common Rule does exempt certain categories of low-risk research, including some benign behavioral interventions, which matters. But the institutional habit is still recognizable: when the process becomes the evidence of safety, participants learn to optimize around the process itself.
This is not an argument against measurement. It is an argument against losing the denominator.
Hospitals can optimize for procedure volume while community health worsens. Schools can optimize for graduation rates while literacy weakens. PBMs can optimize for rebate yield while patients, pharmacies, and plan sponsors struggle to understand the net effect. Compliance teams can optimize for completed training while the actual behavior remains unchanged.
The measurement was supposed to track the value. The value gets eaten by the measurement.
Probe: What is being measured? For whom? Over what time horizon? What disappeared from the denominator?
4. Entrenchment
The system becomes structurally hard to replace, even if many people agree it is failing.
This is where the pressure stops being mostly about ideas and starts being about infrastructure.
Moats form.
There are procurement moats: the system is buried in contracts that take years to unwind.
There are technical moats: the system has accumulated integrations, data feeds, workflows, permissions, edge cases, and reporting dependencies that no replacement can quickly match.
There are regulatory moats: the system has shaped, absorbed, or adapted to the laws that govern it.
There are moral moats: criticism of the system is treated as criticism of the people the system was supposed to protect.
Section 230 of the Communications Decency Act is a useful example outside healthcare. The statute says providers or users of an “interactive computer service” should not be treated as the publisher or speaker of information provided by another information content provider. What began as a liability architecture for online speech and moderation became load-bearing infrastructure for platforms, search, social media, creator economies, and the commercial internet more broadly. Reform proposals now run into stacked moats: legal doctrine, platform dependency, moderation logistics, political disagreement, and the fact that entire markets grew around the original settlement.
That is entrenchment.
The original tradeoff may or may not still be right. But by the time the system is entrenched, the argument is no longer happening on clean conceptual ground. It is happening inside a built environment.
The argument has been won or lost. What remains is the building.
Probe: Is the moat regulatory, technical, procurement-based, or moral? How many of these operate simultaneously? The more that stack, the harder the replacement becomes.
A Note on Sincerity
The most corrosive version of institutional decay is not cynical capture. It is sincere belief.
The PBM analyst may genuinely believe rebate aggregation lowers net plan cost. The clinician may genuinely believe prior authorization protects patients from low-value care. The IRB administrator may genuinely see paperwork as the proof of protection. The school administrator may genuinely believe the dashboard is the only way to manage accountability at scale.
These are not always bad-faith actors performing a script.
Often, they have absorbed the system’s account of itself so completely that they can no longer distinguish their professional integrity from the system’s continued operation. They defend the system past the point of evidence because to stop defending it would feel like repudiating their own working life.
That is what makes the pressures load-bearing.
A cynic can be bought out. A true believer has to be given a new way to remain honorable.
The next stage exists because of this.
5. Identity Fusion
Participants in the system can no longer distinguish between the system’s interests and their own professional identity.
Critique of the system feels like a personal attack.
This is what happens when a recovery culture treats any challenge to its doctrine as proof that the challenger is “in denial” or “not ready.” The framework absorbs its critics by making critique itself a symptom.
It is what happens in corporate cultures where questioning a core economic assumption is treated not as analysis but disloyalty.
It is what happens in professional guilds where leaving the orthodoxy is not an intellectual disagreement but a status death.
At Stage 5, argument is often a category error. You cannot reason someone out of an identity by presenting a cleaner argument. The argument may be correct and still fail, because the person is not only defending a claim. They are defending a self.
Only lived alternatives move people here.
Not abstract alternatives. Not white papers. Not “someone should build a better system.” Real alternatives. Visible alternatives. Inhabitable alternatives. Places where competent people can leave without feeling like they have betrayed their own past.
Probe: What identity does the current system offer its participants? What identity does the alternative offer? Is the alternative existentially plausible, or only intellectually defensible?
6. Substrate Consumption
The system begins degrading the underlying base it depends on.
This is the most dangerous pressure because it can remain hidden for a long time.
A PBM ecosystem still needs pharmacies to exist. It still needs patients to fill prescriptions. It still needs prescribers to tolerate the administrative burden. It still needs plan sponsors to believe the arrangement is working. But if reimbursement pressure, opacity, and administrative complexity squeeze independent pharmacies too hard, the system begins consuming its own access layer. The FTC has raised concerns that PBM concentration and business practices may affect drug access, affordability, and independent pharmacies; the PBM industry disputes many of these characterizations, but the substrate question remains unavoidable.
Schools show a parallel version. The institution depends on experienced teachers, but the profession absorbs workload, compliance, behavioral, safety, compensation, and cultural pressures until staffing stability itself becomes fragile. Federal education data showed that 8 percent of public school teachers left the profession from 2020–21 to 2021–22, while another 8 percent moved to another school. That is not just a labor statistic. It is a substrate warning.
Substrate consumption often gets offloaded downstream.
Patients wait longer. Teachers burn out. Pharmacists absorb more hostile interactions. Researchers avoid messy topics. Nurses become the shock absorbers for broken throughput. Families become the care coordinators. Small operators close. The surface system remains intact because the pain is distributed across people with less power to refuse it.
This stage can persist far longer than it should because the dashboard still works.
The intervention here is narration plus a ready alternative. The consumption has to be documented clearly enough that replacement has legitimacy when people start looking for exits. But critique alone is not enough. Critique without a ready alternative produces awareness without agency. It generates voice without exit.
Probe: What is being depleted so this system can preserve its own motion? And is there a functioning alternative ready to absorb the people who are ready to leave?
The Hierarchy of Fights
The six pressures are not a timeline. They are a profile.
A system can experience identity fusion before regulation hardens. Metric substitution and entrenchment often emerge together. Substrate consumption can begin while the system still appears morally confident and financially successful.
The interventions, however, do have an order.
At Stage 2, you are fighting obscurity. Explain better.
At Stage 3, you are fighting measurement. Force the question: measured against what?
At Stage 4, you are fighting infrastructure. Build elsewhere.
At Stage 5, argument is a category error. Only lived alternatives move people.
At Stage 6, you are fighting inertia. Narrate clearly and absorb the migration.
Most people collapse these into a single battle and wonder why their argument fails.
They bring explanation to an infrastructure fight.
They bring evidence to an identity fight.
They bring outrage to a substrate fight.
They bring a better metric to a system whose procurement moat makes replacement impossible.
Knowing which fight you are in determines which tool you reach for.
The Translator
There is a figure who appears at every stage but belongs to none of them: the translator.
The translator keeps the abstraction tied to lived reality.
The translator makes the guild’s vocabulary legible to outsiders.
The translator asks what disappeared from the denominator.
The translator identifies which moats are stacked and where those moats do not apply.
The translator models defection without demanding renunciation.
The translator documents substrate consumption clearly enough that people leaving the system have somewhere to land.
This role matters because institutional decay is often protected by illegibility. People feel the harm before they can name the mechanism. They know the process is absurd, but not which part is structurally necessary, which part is regulatory theater, which part is vendor lock-in, which part is moral camouflage, and which part is simply habit.
The translator’s job is to separate those layers.
But the translator role has its own failure modes.
The first failure is consultancy: building a career on explaining systems you never change. The translator becomes a tour guide for dysfunction, a diagnostician with no treatment, well-paid to describe problems in ways that never threaten the system’s operation.
The second failure is academy: retreating into description for description’s sake, where the rigor of the analysis becomes a substitute for contact with the world the analysis was supposed to improve.
The third failure is the translator’s own guilding: the moment the translator’s vocabulary becomes the new credential. The framework becomes a moat. The diagnosis becomes another system people need permission to use.
The work, when it stays honest, is closer to a craft than a profession.
Read the system. Build the alternative. Keep the lens free. Keep it legible. Keep it usable by people who will never know your name.
The test is not whether the framework spreads.
The test is whether the people it touches are freer afterward.
Using the Companion Tool
The companion tool turns this essay into a profile.
Pick a system: a payer, a school, a startup, a professional guild, a compliance function, a therapy culture, a public agency, a platform, a hospital department, a benefits administrator.
Score it across the six pressures.
The point is not to certify the diagnosis. The point is to ask better questions:
Which pressures are active?
Which are acute?
Which fight are you actually in?
Is the work explanation, measurement, replacement, defection, narration, or absorption?
A radar chart will not fix a broken institution. But it can prevent a common mistake: treating every institutional problem as if it needs the same intervention.
Some systems need translation.
Some need better denominators.
Some need a new forum.
Some need an exit.
Some need a place for the exhausted to land.
Build Anyway
The pressures are not going away.
New systems are being named right now that will be guilded by 2030, fused with the identities of their participants by 2035, and consuming their substrate by 2040.
That is not pessimism. It is the operating condition of any system durable enough to matter.
The work is not to find the institution that will never decay. The work is to build with the decay arc in view.
Name carefully.
Measure against the mission.
Keep the language open.
Avoid moats that only insiders can cross.
Let people critique the system without losing their identity.
Watch the substrate.
And when the system begins to preserve itself at the expense of the thing it was built to serve, do not stop at critique.
Build a better elsewhere.
Make it legible.
Make it inhabitable.
Make it ready.
