Stratified Treatment¶
Essence¶
Stratified Treatment is the archetype of treating unlike cases differently on purpose. It begins with a simple observation: a population, system, or workload may look like one set only because the differences inside it are being ignored. If those differences change what will work, then a single treatment can be both inefficient and unfair.
The archetype does not stop at classification. Naming strata is only preparation. The solution pattern is complete only when each stratum receives a treatment policy suited to its risk, need, capacity, response pattern, or constraint, and when the system monitors whether those differences remain justified.
Compression statement¶
When a population or system contains distinct strata with different needs, risks, capacities, responses, or constraints, segment them and apply stratum-specific treatment to improve fit while preserving fairness, accountability, and a defensible baseline.
Canonical formula: heterogeneous field + actionable stratum differences + stratum-specific treatment policy + fairness/review guardrails -> better treatment fit without arbitrary segmentation
When to Use This Archetype¶
Use Stratified Treatment when uniform treatment predictably misfits the field. Typical signs include simultaneous over-service and under-service, repeated informal exceptions, subgroup failure hidden by aggregate success, or costly intervention capacity being spent where it has little effect.
It is especially useful when sameness would not preserve the real invariant. In education, the invariant might be a shared competency rather than the same worksheet. In public services, it might be usable access rather than identical application channels. In cybersecurity, it might be acceptable risk rather than identical controls on every asset.
Do not use it merely because categories are available. The categories must matter for action. If no treatment changes, this is classification. If only the order of service changes, this is queue discipline or priority admission. If only a sample is balanced across groups, this is stratified review or sampling.
Structural Problem¶
The structural problem is heterogeneous need under a uniform rule. The system is trying to govern, support, protect, inspect, educate, or serve cases as though they were interchangeable, but they differ in ways that change the appropriate treatment.
This creates a recurring mismatch. Some cases receive too much burden or intensity. Others receive too little support or protection. Operators begin creating informal exceptions because the official rule is too blunt. Meanwhile, averages hide which strata are succeeding and which are failing.
The deeper tension is between accountability and fit. A system needs rules stable enough to be explainable, but it also needs enough differentiation to avoid treating unlike cases as if they were the same.
Intervention Logic¶
Stratified Treatment turns hidden heterogeneity into governed differentiation. First, the system names the shared purpose: safety, capability, service quality, access, resilience, learning, or another invariant. Then it identifies the differences that matter for treatment. It defines strata, assigns cases to them, attaches treatment policies to each stratum, and monitors outcomes.
The treatment difference may be a different resource level, threshold, cadence, burden, safeguard, channel, support, or pathway. A high-risk asset may be patched faster. A high-need case may receive intensive support. A learner with a different readiness profile may receive different scaffolding. A vulnerable applicant may receive extra navigation assistance.
The key is that differential treatment must be justified. A stratum is not just a label; it is a claim that a different treatment better preserves the system's purpose.
Key Components¶
Stratified Treatment converts hidden heterogeneity in a population, workload, or asset set into governed differentiation, and its components form a chain from why-strata-exist to how-strata-are-judged-to-still-work. The Classification Basis states the dimension along which cases meaningfully differ for treatment — risk, need, capacity, exposure, vulnerability — and rules out categories that are easy to observe but irrelevant to action. The Stratum Definition makes each layer administrable by fixing boundaries, evidence requirements, and edge-case handling. The Assignment Rule then determines how a specific case enters a stratum, including how missing information, uncertainty, and overrides are handled — without it, operators apply strata inconsistently or shape membership for convenience. Together these three components establish what the strata are and how cases are placed into them.
The remaining components convert membership into actual differentiated action and protect the system against the predictable failure modes of unequal treatment. The Treatment Policy is the core intervention component: it specifies exactly what changes by stratum, and without it the system is only classifying. Resource or Threshold Differentiation translates membership into operational difference by varying support intensity, escalation cutoffs, response time, or maintenance cadence. Because the archetype deliberately treats cases differently, a Fairness Policy names the legitimate basis for differentiation, sets a minimum floor no stratum falls below, and flags prohibited distinctions. Monitoring Feedback checks whether outcomes actually improve, whether any stratum is being neglected, and whether the assignment rule has drifted, while a Review or Appeal Path makes individual misassignments and boundary harms correctable rather than absorbing them silently.
| Component | Description |
|---|---|
| Classification Basis ↗ | The classification basis states why strata exist. It might be risk, need, capacity, response likelihood, complexity, exposure, severity, or vulnerability. The basis must be relevant to treatment choice. A category that is easy to observe but irrelevant to intervention will produce false precision. |
| Stratum Definition ↗ | A stratum definition makes each layer, tier, band, or segment administrable. It clarifies who belongs where, where boundaries sit, and what evidence is required. Strong stratum definitions reduce arbitrary assignment while still leaving room for review of edge cases. |
| Assignment Rule ↗ | The assignment rule determines how a case enters a stratum. It handles missing information, uncertainty, overrides, and evidence quality. Without an assignment rule, operators may apply strata inconsistently or manipulate membership for convenience. |
| Treatment Policy ↗ | The treatment policy is the core intervention component. It specifies what changes by stratum: support, resources, inspection cadence, escalation threshold, response time, channel, safeguards, burden, or pathway. If nothing changes by stratum, the system is only classifying. |
| Resource or Threshold Differentiation ↗ | Many stratified systems work by changing intensity or thresholds. Higher-risk strata may have lower escalation thresholds. Lower-capacity strata may receive more support. High-criticality assets may receive faster maintenance. This component translates membership into operational difference. |
| Fairness Policy ↗ | Because the archetype deliberately treats cases differently, it needs a fairness policy. This policy states the legitimate reason for differentiation, the minimum floor no stratum falls below, and any distinctions that are prohibited or require special review. |
| Monitoring Feedback ↗ | Monitoring asks whether the strata still work. It checks whether outcomes improve, whether a stratum is neglected, whether boundary cases are harmed, and whether the assignment rule has drifted. Monitoring also supports reclassification when risk, need, or capacity changes. |
| Review or Appeal Path ↗ | A review or appeal path protects against assignment errors and boundary harms. It is especially important when stratified treatment affects access, burden, safety, opportunity, or dignity. Review does not eliminate rules; it makes them correctable. |
Common Mechanisms¶
| Mechanism | Description |
|---|---|
| Risk Stratification Protocols ↗ | Risk stratification protocols assign cases to risk bands and connect each band to action. They are common in clinical care, cybersecurity, infrastructure maintenance, and regulation. They implement the archetype only when the risk band changes treatment, not when it merely produces a score. |
| Clinical Risk Banding ↗ | Clinical risk banding is a domain-specific mechanism that changes follow-up, monitoring, screening, treatment intensity, or escalation based on patient risk. The archetype is not the medical scoring tool itself; it is the stratum-specific care policy that follows from the band. |
| Tiered Service Catalogs ↗ | A tiered service catalog documents what different strata receive: response times, support levels, escalation paths, eligibility, or safeguards. It is a mechanism under the archetype because it packages treatment policies in an operator-facing artifact. |
| Differentiated Instruction Plans ↗ | In learning systems, differentiated instruction plans vary scaffolds, practice, pacing, examples, or challenge level across learner strata. They instantiate Stratified Treatment when the differentiation preserves a learning objective rather than permanently tracking learners into unequal expectations. |
| Vulnerability-Based Support Workflows ↗ | A vulnerability-based support workflow directs extra assistance, protection, or simplification toward strata with higher exposure or lower capacity. It is often used in public services, disaster response, health outreach, and case management. |
| Stratum-Specific Threshold Schedules ↗ | A threshold schedule lists different cutoffs for escalation, review, inspection, eligibility, or intervention by stratum. It is useful but dangerous near boundaries, where small measurement differences can create large treatment differences. |
| Fairness Audits by Stratum ↗ | A fairness audit checks whether the differentiated treatment is producing intended fit without unacceptable under-service, exclusion, stigma, or disparate harm. It is a mechanism for maintaining the archetype, not a replacement for the treatment policy. |
Parameter / Tuning Dimensions¶
The first tuning dimension is stratum granularity. Too few strata recreate one-size-fits-all treatment. Too many strata create fragmentation and operational burden.
The second is assignment strictness. Strict thresholds improve consistency but can create cliff effects. Softer boundaries allow judgment but can become arbitrary.
The third is treatment contrast. If treatment differences are too small, stratification adds bureaucracy without benefit. If they are too large, boundary errors become harmful.
The fourth is reclassification cadence. Static strata are stable but can lock cases into stale labels. Dynamic strata are adaptive but can create churn, gaming, and administrative load.
The fifth is fairness floor strength. A strong floor prevents abandonment of lower-priority strata, but it may reduce the capacity available for high-need strata. A weak floor can make stratification look efficient while quietly degrading service.
Invariants to Preserve¶
The most important invariant is the shared purpose. Different treatments should still serve a common aim, such as safety, capability, access, learning, or resilience.
A second invariant is the minimum service floor. No stratum should fall below a defensible baseline simply because another stratum receives more attention.
A third invariant is classification integrity. Assignment should be valid, auditable, revisable, and explainable enough for the stakes involved.
A fourth invariant is fair justification. The system should be able to explain why each treatment difference exists and why that difference is legitimate.
A fifth invariant is movement where appropriate. When the relevant condition changes, the system should allow reclassification rather than turning a temporary stratum into a permanent label.
Target Outcomes¶
A successful Stratified Treatment design improves fit. Cases receive the level, type, timing, and pathway of treatment more appropriate to their condition.
It also improves visibility. Instead of hiding failures inside averages, the system can see outcomes by stratum and adjust policies where needed.
It makes exceptions more governable. Informal workarounds become explicit, reviewable treatment policies.
It can improve resource use. Scarce capacity moves toward cases where different treatment changes outcomes, while baseline protections prevent neglect.
Tradeoffs¶
The core tradeoff is fit versus simplicity. A stratified system is more complicated than a uniform one. It requires assignment, boundary handling, review, and monitoring.
Another tradeoff is equal treatment versus equitable fit. Treating people or cases differently can be fair when circumstances differ, but the same structure can also disguise arbitrary inequality.
A third tradeoff is stability versus adaptability. Stable strata are easier to understand and administer. Adaptive strata are more responsive but increase churn and ambiguity.
A fourth tradeoff is transparency versus privacy. People may need to understand why they are assigned to a stratum, while the indicators used for assignment may be sensitive.
Failure Modes¶
Bad stratum design is the most basic failure. If the strata do not reflect relevant treatment differences, the system becomes a confident misclassification machine.
Label lock-in occurs when cases cannot move after their situation changes. This is common when strata become identities or administrative conveniences instead of treatment-fit tools.
Cliff effects occur when a small difference near a threshold causes a large treatment difference. Boundary uncertainty zones, blended treatment, or review can reduce this harm.
Stigma arises when stratum names or treatment differences become status markers. Neutral language, privacy protection, and clear movement paths help reduce this risk.
Opaque discrimination occurs when differential treatment hides illegitimate distinctions. Fairness policies, audits, and appeal paths are essential safeguards.
Abandonment of low-intensity strata occurs when resources concentrate on high-risk or high-need strata while the baseline erodes. A minimum service floor is the main protection.
Neighbor Distinctions¶
Stratified Treatment is distinct from Priority-Based Admission because admission decides who enters or who goes first; stratified treatment governs what different strata receive.
It is distinct from Canonical Classification because classification names categories; stratified treatment links categories to different actions.
It is distinct from Access Control because access control blocks or permits use; stratified treatment can operate after access is granted and may differentiate ongoing support, burden, or monitoring.
It is distinct from Queue Discipline Design because queue discipline changes order, while stratified treatment changes treatment substance or intensity.
It is distinct from Gradient-Guided Intervention because gradient-guided intervention may follow continuous variation, while stratified treatment discretizes relevant variation into governed strata.
It is distinct from Stratified Sampling / Review because sampling by stratum is about evidence quality. It becomes treatment only when the review changes intervention policies.
Variants and Near Names¶
Risk-Stratified Treatment uses risk as the primary dimension and changes monitoring, protection, escalation, or intervention intensity accordingly.
Vulnerability-Based Support uses lower capacity or higher exposure as the basis for extra assistance or protection.
Tiered Service Levels package stratum-specific treatment into service tiers, response levels, or support catalogs. These are useful mechanisms but should not be mistaken for the parent archetype.
Differentiated Instruction is a learning-domain variant. It should remain in merge review because future learning-family drafts may treat differentiated pathway design as a full archetype.
Dynamic Restratification makes movement between strata part of the operating loop. It may become a standalone candidate if feedback-driven stratum movement proves structurally distinct.
Near names include tiered treatment, differentiated treatment, segmented intervention, stratified care, service tiers, and risk bands.
Cross-Domain Examples¶
In healthcare, patients may be grouped by risk so that higher-risk groups receive closer follow-up and lower-risk groups avoid unnecessary burden.
In education, learners may receive different scaffolds or practice paths while preserving the same competency target.
In cybersecurity, high-criticality assets may receive stricter controls and faster response windows than low-criticality assets.
In social services, high-complexity cases may receive intensive case management while stable cases receive lighter support and periodic review.
In infrastructure maintenance, assets can be stratified by consequence of failure and condition, changing inspection cadence and repair thresholds.
In regulation, facilities with higher hazard profiles may receive more frequent inspections and stricter reporting requirements.
Non-Examples¶
A labeled segmentation table with no action differences is not Stratified Treatment. It is classification.
A queue that serves urgent requests first is not necessarily Stratified Treatment. It may be priority admission or queue discipline if the treatment itself does not change.
A stratified audit sample is not Stratified Treatment unless the results feed into different treatment policies.
A prestige tier that gives favored groups better service without a defensible relation to need, risk, capacity, or shared outcome is not a good instance of this archetype. It is segmentation without the solution logic.