Min–Max Inventory¶

An inventory policy that manages stock inside a declared band — replenishing up to a maximum ceiling only once on-hand falls to a minimum reorder point, and doing nothing in between — collapsing continuous order decisions into two scalars per item.

Core Idea¶

Min–max inventory is an inventory-control policy that manages a stock level inside a declared band: when on-hand quantity falls to or below a minimum threshold (the reorder point), a replenishment is triggered that brings the stock up to or near a maximum ceiling; while quantity remains between the two thresholds, no replenishment action is taken. The mechanism converts a continuous control problem — deciding at every moment how much to order — into two scalar parameters per stocked item: the minimum, set high enough to cover expected demand during replenishment lead time plus a safety margin against demand and lead-time variability; and the maximum, set to limit excess holding cost and physical space. The operational logic is banded hysteretic control: the system tolerates stock anywhere within the band without acting, and takes a single large action (replenish to max) only when the lower boundary is crossed. The practical payoff is decision-cost reduction and thrash prevention — a stockkeeper managing hundreds of SKUs does not continuously optimize each one but monitors for boundary violations and acts at the minimum required frequency. The cost is within-band suboptimality: no inventory optimization is performed while stock moves between min and max. Threshold calibration is the central design problem: the minimum must cover demand during the replenishment lead time at the target service level (so it embeds safety-stock reasoning), while the gap between min and max sets the typical replenishment quantity and thereby determines order frequency and average inventory. The policy is formally related to the (s, S) policy in operations research — where s is the reorder point and S the order-up-to level — of which min–max is the practitioner's two-threshold formulation.

Structural Signature¶

Sig role-phrases:

the controlled stock variable — an on-hand quantity, with a unit and a monitoring cadence, that drifts with demand
the minimum threshold — the reorder point whose breach triggers replenishment; embeds safety-stock reasoning covering lead-time demand at a target service level
the maximum threshold — the order-up-to ceiling that caps holding cost and physical space
the replenishment action — the single large move that brings stock from below-min up to (or near) the max
the in-band inaction — the hysteretic rule: tolerate any state between thresholds without acting, trading within-band optimality for reduced decision cost and thrash
the band width — the min-to-max gap, the real lever: it sets the replenishment quantity and thereby order frequency and average inventory (the quantity is a consequence, not a parameter)
the lead-time-and-variance calibration — the placement rule that ties the minimum to replenishment lead time and demand variability, requiring recheck whenever supply timing shifts
the (s,S) correspondence — min–max as the practitioner's two-threshold reading of the operations-research (s,S) policy, importing its worked-out optimal-threshold results

What It Is Not¶

Not the same as safety stock. The minimum embeds safety-stock reasoning — it must cover lead-time demand at a target service level — but the two are distinct objects: the minimum is the trigger for action, the safety stock is the buffer baked into it against demand and lead-time variability. Conflating them leaves a stockkeeper unable to say whether a shortfall came from setting the reorder point too low or from sizing the buffer wrong.
Not a fixed-order-quantity policy. The replenishment quantity is not a parameter at all; it is a consequence — the gap between current stock and the maximum when the trigger fires. The lever that actually sets order frequency and average inventory is the band width, not either threshold alone, so reading min–max as "order a fixed lot" misses where the cadence is controlled.
Not continuous or proportional control. The policy is banded hysteretic control: it does nothing while stock sits anywhere inside the band and acts only when the lower boundary is crossed. Unlike a proportional or PID controller, the action depends only on whether the variable is in-band, not on its distance from a setpoint — the in-band inaction is the design, not a lapse.
Not an optimizer. Min–max deliberately accepts within-band suboptimality — no inventory optimization happens while stock moves between thresholds — buying decision-cost reduction and thrash prevention across hundreds of SKUs in return. An item whose holding or stockout cost justifies continuous optimization should not be left to drift inside a band; the policy's value is precisely that it stops optimizing.
Not distinct from the (s, S) policy. Min–max is the practitioner's two-threshold reading of the operations-research (s, S) policy — s the reorder point (minimum), S the order-up-to level (maximum). Treating them as separate ideas misses that the worked-out analytical results for optimal s and S apply directly to the two scalars on the shop floor.

Scope of Application¶

Min–max lives across the inventory-control subfields of logistics, manufacturing, and retail, plus the operations-research theory it reads onto; its in-domain reach spans the worked spectrum of stocked items, while the wider banded-hysteretic-control co-instances (cash management, autoscaling, reservoirs, homeostasis) are the general feedback-plus-hysteresis mechanism carried by feedback / reserve / boundedness, not this label.

Logistics and distribution stock — finished-goods and raw-materials stockkeeping run as a two-threshold reorder-to-ceiling band across many SKUs by exception.
Manufacturing materials control — component and consumable stocks held inside a band, the minimum embedding lead-time-and-variance safety reasoning.
Retail replenishment — store and warehouse SKUs reordered when on-hand hits the minimum, the band width setting cadence and average inventory.
Hospital pharmacy and crash-cart par levels — drug and consumable SKUs restocked to a maximum when they fall to par, scanned for boundary violations rather than continuously optimized.
Operations-research (s, S) policy — the formal analogue (s the reorder point, S the order-up-to level) whose worked-out optimal-threshold characterization transfers directly onto the two scalars.

Clarity¶

Naming the min–max band makes legible a set of distinctions that inventory practice habitually collapses into a single "stock level" intuition. The minimum is the trigger for action — the reorder point — not the same object as the safety stock, the buffer baked into that minimum to absorb demand and lead-time variability during replenishment; conflating them leaves a stockkeeper unable to say whether a shortfall came from setting the reorder point too low or from sizing the buffer wrong. The maximum is an order-up-to ceiling that caps holding cost and shelf space, distinct from the replenishment quantity, which is not a parameter at all but a consequence — the gap between current stock and the max — so the lever that actually sets order frequency is the width of the band, not either threshold alone. And banded hysteretic control is distinct from continuous-rate control: the policy's whole point is to do nothing while stock sits anywhere inside the band and to act only when the lower boundary is crossed, trading within-band optimality for a sharp reduction in decision cost and order thrash.

Once those are held apart, the SKU-management question changes shape. Instead of "how much should I order, and when?" — a continuous optimization the stockkeeper would have to re-run for every item every period — the policy reduces each SKU to two scalars and reframes the work as boundary-monitoring across hundreds of items. The frame also makes failure diagnosis crisp: a stockout decomposes into policy mis-specification (minimum set below what the lead-time demand and service-level target require) versus execution failure (the replenishment never fired), and an overstock into a maximum set too high versus a demand collapse after ordering. By exposing that the minimum embeds service-level-targeted safety-stock reasoning while the min-to-max gap governs cadence, the band tells a practitioner exactly which knob to turn for a given symptom — and clarifies that min–max is the shop-floor two-threshold reading of the operations-research (s, S) policy, so the analytical results for s and S apply directly.

Manages Complexity¶

Controlling a stock level is, in its raw form, a continuous-time decision problem: at every moment the quantity on hand has drifted, and in principle the stockkeeper must decide whether to order and how much — a problem that, run honestly across the hundreds or thousands of SKUs a real operation carries, is unmanageable, because it demands a fresh order-quantity-and-timing computation per item per period. Min–max compresses that continuous control problem onto two scalars per item — a minimum and a maximum — and replaces "when and how much to order" with a single boundary test: is on-hand at or below the minimum? While stock sits anywhere inside the band, the policy prescribes inaction, so the stockkeeper does nothing at all on the great majority of items on any given day and acts only where a lower boundary has been crossed. A continuous optimization over every SKU becomes boundary-monitoring across the catalogue plus one mechanical replenish-to-max action where a trigger fires. The decision load drops from per-item-per-period optimization to scanning for threshold violations.

Because the two scalars carry specific meanings, the analyst reads the policy's behavior off them rather than re-deriving inventory dynamics per item. Order frequency and average inventory are not separately chosen — they follow from the band width, since the gap between current stock and the maximum is the replenishment quantity and the min-to-max span sets how often the trigger recurs; service-level performance follows from where the minimum sits, because the minimum embeds the safety-stock buffer that must cover lead-time demand at the target service level. This is what makes failure diagnosis a fixed decomposition instead of an open investigation: a stockout resolves to either policy mis-specification (the minimum was set below what lead-time demand and the service target require) or execution failure (the trigger fired but the replenishment never arrived); an overstock resolves to either a maximum set too high or a demand collapse after the order was placed. The analyst reads the symptom and the band tells them which knob is implicated — raise the minimum for a coverage shortfall, widen or narrow the band for a cadence problem, check execution for a phantom order. The within-band suboptimality the policy accepts (no optimization happens between thresholds) is the price paid for this collapse, and naming it tells the analyst exactly what they have given up. Because min–max is the practitioner's two-threshold reading of the operations-research (s, S) policy, the analytical results that characterize optimal s and S transfer onto the two scalars directly, so the qualitative outcomes a stockkeeper reads off the band rest on a worked-out theory rather than intuition.

Abstract Reasoning¶

Min–max licenses a set of moves on any banded stock-control problem, every one of them routed through the two scalars and the band they define. Diagnostic (the signature move) — symptom-to-knob decomposition: a failure does not present as a vague "inventory problem" but resolves, via the policy's structure, into a small fixed set of causes the analyst can discriminate. A stockout splits two ways — either policy mis-specification, the minimum set below what lead-time demand at the target service level requires, or execution failure, the trigger fired but the replenishment never arrived — and the analyst tells them apart by asking whether the reorder ever launched. An overstock splits the same way — a maximum set too high, versus a demand collapse after the order was placed — discriminated by whether demand held after replenishment. The move is to read the symptom and infer which of these the band implicates, rather than open an undirected investigation. Interventionist — turn the right scalar: each diagnosis points to a specific lever, and the analyst predicts the effect before acting. A coverage shortfall means raise the minimum, because the minimum embeds the safety-stock buffer that must cover lead-time demand; an order-frequency or thrash problem means change the band width, because the min-to-max gap sets the replenishment quantity and thereby how often the trigger recurs and how much average inventory the item carries; a phantom shortfall with a correctly-set minimum means fix execution, not parameters. The characteristic prediction couples a parameter move to two outcomes at once: widening the band lowers order frequency but raises average inventory, narrowing it does the reverse — so the analyst reasons about the tradeoff the width controls, not about order-quantity and inventory-level as separate dials. Boundary-drawing — when the band is the wrong instrument, and where its results hold: the policy buys decision-cost reduction by accepting within-band suboptimality — no optimization happens while stock moves between thresholds — so the move is to recognize when that trade is wrong: an item whose holding or stockout cost is high enough to justify continuous optimization should not be left to drift inside a band, and an item managed across a catalogue of hundreds, where the binding constraint is the stockkeeper's attention, is exactly where banded inaction pays. A second boundary is the lead-time assumption baked into the minimum: the reorder point covers demand during the replenishment lead time, so when that lead time lengthens or grows variable, the minimum must move with it or the band silently loses its service-level guarantee — the move is to recheck the minimum whenever supply timing shifts, not to treat it as a set-once constant. Borrow the theory — read min–max as (s, S): because the practitioner's two thresholds are the operations-research reorder point s and order-up-to level S, the analyst's move is to import the worked-out characterization of optimal s and S directly onto the band rather than tune by intuition, reasoning from the established results to where the minimum and maximum should sit given demand variance, lead time, and the cost ratio.

Knowledge Transfer¶

Within inventory control min–max transfers as mechanism: the same two scalars, the same band-width tradeoff, the same symptom-to-knob failure decomposition, and the same lead-time-and-variance calibration of the minimum apply across logistics, manufacturing, and retail stock, and across the worked spectrum of stocked items from raw materials to finished goods to hospital pharmacy par levels and crash-cart restocking. The diagnostics (a stockout splits into policy mis-specification versus execution failure; an overstock into a too-high maximum versus a post-order demand collapse), the interventions (raise the minimum for a coverage shortfall, change the band width for a cadence or thrash problem, fix execution for a phantom shortfall), and the boundary checks (recheck the minimum whenever lead time lengthens or grows variable) carry intact; only the SKU and the lead-time distribution change. And because min–max is the practitioner's two-threshold reading of the operations-research (s, S) policy, the analytical characterization of optimal s and S transfers directly onto the two thresholds, so the within-domain results rest on a worked-out theory rather than intuition.

Beyond inventory the honest characterization is a (B) shared abstract mechanism, and min–max is a strong instance of it because the control logic genuinely recurs as a co-instance across substrates rather than as mere resemblance. Strip the inventory vocabulary and what remains is banded hysteretic feedback control of a continuous variable: act only when a lower boundary is crossed, push to a ceiling, and tolerate any state inside the band without acting — trading within-band optimality for a sharp cut in decision cost and thrash. That structure recurs literally in cash and treasury management (operating cash held inside a band, drawing on credit below the minimum and sweeping excess above the maximum — the Miller–Orr model is the analytical formalization), in cloud autoscaling (add capacity above a high-water mark, shed it below a low-water mark, the band preventing scale-up/scale-down thrash on small fluctuations), in reservoir and dam outflow rules (level held in a band balancing flood risk above and supply risk below), and in biological and engineered homeostasis (a thermostat's hysteresis band, glycemic or blood-pressure titration to a target range). These are not analogies but the same banded-control mechanism on a different controlled variable, which is why an autoscaling engineer setting thresholds, a pharmacist setting par levels, and a treasury team setting a cash band are recognizably solving one problem. But the mechanism that travels is the general one, already carried by the catalog: feedback (the closed loop, here specifically hysteretic rather than proportional/PID — the action depends only on whether the variable is in-band, not on distance from a setpoint), buffering and reserve (the held quantity), system_slack (the min-to-max gap as one form of slack), and boundedness (the in-band property the policy produces). The home-bound cargo is everything inventory-specific: the minimum's safety-stock content covering replenishment lead-time demand at a service-level target, the (s, S) calibration, the SKU-catalogue attention constraint that justifies banded inaction. So the cross-domain lesson should carry the general banded-hysteretic-feedback pattern (feedback plus reserve plus hysteresis plus boundedness), and "min–max inventory," as named, should stay the inventory-control instance of it, sitting alongside (s, S), EOQ, kanban, and JIT rather than being promoted to cover every hysteresis band (see Structural Core vs. Domain Accent).

Examples¶

Canonical¶

Take one SKU with average demand of 20 units per day and a replenishment lead time of 5 days. Expected demand during lead time is 20 × 5 = 100 units; to guard against demand and lead-time variability at the target service level, add a safety buffer of, say, 30 units, so the minimum (reorder point) is set at 130. The maximum, chosen to cap holding cost and shelf space, is set at 300. The policy then does nothing while on-hand sits between 130 and 300. The moment stock falls to 130, a single replenishment fires, bringing on-hand up to 300 — an order of roughly 170 units. With demand of 20 per day, that 170-unit band depletes in about 8.5 days, so the trigger recurs roughly every 8–9 days.

Mapped back: On-hand stock is the controlled stock variable; 130 is the minimum threshold, embedding the lead-time-and-variance calibration (100 units of lead-time demand plus a 30-unit safety buffer). 300 is the maximum threshold. The order fired at 130 is the replenishment action, and doing nothing between 130 and 300 is the in-band inaction. The 170-unit gap is the band width — the real lever, setting both order quantity and the ~8.5-day cadence as consequences, not chosen parameters.

Applied / In Practice¶

Hospital pharmacies run min–max as "par levels" in automated dispensing cabinets on nursing units (Pyxis, Omnicell, and similar systems are the common deployments). Each drug line in a cabinet is assigned a par (minimum) and a maximum. Nurses withdraw doses through the day; when a drug's count falls to its par, the system flags it on a restock report. Pharmacy technicians then refill only the flagged items back up to the maximum, working the thousands of line items across dozens of cabinets purely by exception rather than counting every drawer.

Mapped back: Each drug's count is the controlled stock variable; the par is the minimum threshold whose breach is the replenishment action trigger, and it must cover usage during the pharmacy's restock lead time — the lead-time-and-variance calibration. The maximum caps drawer space. Above par, the cabinet takes no action — the in-band inaction — so technicians scan for boundary violations instead of continuously optimizing, exactly the decision-cost collapse the policy buys across a large SKU catalogue.

Structural Tensions¶

T1: Decision-cost reduction versus within-band suboptimality (what the collapse to two scalars costs). The policy's whole value is that it stops optimizing: it converts a continuous order-quantity-and-timing problem into a boundary test, so a stockkeeper managing hundreds of SKUs monitors for threshold violations instead of re-running an optimization per item per period. But the same inaction that saves attention also forgoes every optimization opportunity while stock drifts inside the band — a favorable price, a demand signal, a consolidation chance mid-band all go unexploited. The trade is not incidental but constitutive: attention savings and within-band optimality are bought and sold in the same move, and an item whose holding or stockout cost is high enough to reward continuous optimization is exactly the item on which banded inaction leaves the most on the table. Diagnostic: Is the binding constraint the stockkeeper's attention across many SKUs (banded inaction pays), or the per-item holding/stockout cost of this one item (continuous optimization pays)?

T2: Interface parsimony versus hidden semantic load (two scalars that are not two simple numbers). Reducing each item to a minimum and a maximum is the source of the policy's tractability — but the parsimony of the interface disguises how much reasoning is compressed into those two numbers. The minimum is not a bare trigger: it embeds safety-stock reasoning that must cover lead-time demand at a target service level, so a shortfall could come from the reorder point being too low or from the buffer inside it being mis-sized, two different faults wearing one number. The replenishment quantity, meanwhile, is not a parameter at all but a consequence of the band width. The tension is that the same two-scalar simplicity that makes the policy usable at scale invites practitioners to treat the numbers as atomic when their content is compound — and misreading which meaning a number carries misroutes every fix. Diagnostic: When this minimum failed, was it the reorder trigger set too low or the safety buffer baked into it sized wrong — and is the quantity being treated as a chosen parameter when it is a consequence of the band width?

T3: Band width as a single lever versus the coupled outcomes it moves (order frequency and average inventory together). The min-to-max gap is the real control, but it is one dial wired to two outcomes that cannot be set independently. Widen the band and each order is larger, so the trigger recurs less often — fewer orders — but average inventory rises and holding cost with it; narrow it and orders come frequently at low average inventory but with more thrash and ordering cost. There is no band width that delivers both infrequent ordering and low inventory; the policy forces a position on that curve rather than letting the two be chosen apart. The lever that makes the policy simple is exactly the lever that fuses two objectives a manager might wish to optimize separately. Diagnostic: Is the current pain a cadence/thrash problem or an average-inventory problem — and is the fix acknowledging that moving the band width trades one against the other?

T4: Set-once robustness versus silent staleness (the calibration that erodes without a symptom). The policy's operational appeal is that once the band is set it can be left alone — the stockkeeper just watches for boundary crossings. But the minimum's service-level guarantee rests on an assumption baked in at calibration time: that it covers demand during the replenishment lead time. When lead time lengthens or grows variable, or demand shifts, the band keeps functioning mechanically while quietly losing the guarantee it was sized to deliver — the failure surfaces only as a stockout, after the fact. The virtue (leave it alone, monitor boundaries) is precisely what lets the calibration go stale unnoticed, because nothing in the day-to-day operation flags that the world under the parameters has moved. Diagnostic: Have the lead time or demand distribution that this minimum was sized against shifted since it was set — and is anything in the process rechecking that, or is the band assumed constant?

T5: Hysteretic inaction as thrash-prevention versus trend-blindness (stability bought by ignoring information). Doing nothing while stock sits inside the band is what prevents order thrash on small fluctuations — the action depends only on whether a boundary is crossed, not on distance from a setpoint, so noise does not provoke churn. But that same insensitivity discards real information: a demand trend developing mid-band is not acted on until stock happens to hit the minimum, by which point the trend may have outrun the buffer. The policy's stability comes from being deaf to everything short of a boundary crossing, and that deafness is indiscriminate — it filters out noise and signal alike. Robustness to fluctuation and responsiveness to genuine change are the same knob turned opposite ways. Diagnostic: Is the in-band quiet filtering out meaningless fluctuation, or is it also suppressing a real demand trend that will breach the buffer before the boundary check fires?

T6: Autonomy versus reduction (an inventory policy or an instance of banded hysteretic control). "Min–max inventory" carries genuine home-domain cargo — the minimum's safety-stock content covering replenishment lead-time demand at a service target, the (s, S) calibration, the SKU-catalogue attention constraint that justifies banded inaction — and within inventory that apparatus does the work. Yet strip the vocabulary and what remains is banded hysteretic feedback control of a continuous variable, a mechanism that co-instantiates literally in cash management (Miller–Orr), cloud autoscaling, reservoir outflow rules, and homeostasis. Those are not analogies but the same structure on a different controlled variable, already carried by feedback (hysteretic, not proportional), reserve/buffering, system_slack, and boundedness. The tension is between a named policy whose inventory-specific calibration earns its place beside (s, S), EOQ, and kanban, and the recognition that the control logic which reaches autoscaling and treasury belongs to the general banded-feedback parents. Diagnostic: Resolve toward the feedback/reserve/hysteresis/boundedness parents when carrying the lesson to autoscaling, cash bands, or homeostasis; toward named min–max when safety-stock, lead-time, and (s, S) calibration are doing the work.

Structural–Framed Character¶

Min–max inventory is mixed on the structural–framed spectrum — a designed operational policy pinned to inventory vocabulary, but one whose portable skeleton is an evaluatively neutral control mechanism that genuinely recurs (and runs observer-free) in nature, so it holds the middle. The criteria pull in both directions. On evaluative weight it reads structural: a stock level held inside a band is neither good nor bad, "min–max" convicts nothing, it merely names a control rule. But human-practice-bound and institutional origin point framed: min–max is an engineered policy — a shop-floor decision rule that presupposes SKUs, stockkeepers, replenishment orders, and a service-level target — and it is the practitioner's two-threshold reading of an operations-research construct, the (s, S) policy; the named entry is an artifact of inventory-management practice, not a fact nature exhibits on its own. Vocab-travels points framed: the operative terms — reorder point, safety stock, order-up-to level, replenishment lead time, (s, S), SKU catalogue — are irreducibly inventory-control and do not float free of the stockroom. Import-vs-recognize is bimodal in the entry's own telling: within inventory control the policy transfers as recognition of the same mechanism across logistics, manufacturing, retail, and pharmacy par levels; beyond it (cash management, autoscaling, reservoirs, homeostasis) the control logic is a real co-instantiation of the general banded-feedback mechanism rather than mere metaphor — but what co-instantiates there is the parent, not "min–max inventory."

The portable structural skeleton is a single one: banded hysteretic feedback control of a continuous reservoir — act only when a lower boundary is crossed, drive to a ceiling, and tolerate any state inside the band, trading within-band optimality for reduced decision cost and thrash. That skeleton is genuinely substrate-general and even nature-recognized: it is the same mechanism a thermostat's hysteresis band and physiological homeostasis run without any designer watching. But it is exactly what min–max instantiates from its umbrella primes — feedback (here specifically hysteretic, not proportional/PID), reserve / buffering (the held stock), system_slack (the min-to-max gap), and boundedness (the in-band property produced) — not what makes "min–max inventory" itself travel: the cross-domain reach to autoscaling and treasury bands belongs to that banded-feedback family, while the domain-accented specifics — the minimum's safety-stock content covering lead-time demand at a service level, the (s, S) calibration, the SKU-catalogue attention constraint — stay home. Its character: an evaluatively neutral banded-hysteretic-control mechanism, structural and even nature-recurring in skeleton, but realized here as a human-designed inventory policy stated in stockroom vocabulary, leaving it mixed rather than a free-floating prime.

Structural Core vs. Domain Accent¶

This section settles why min–max inventory is a domain-specific abstraction and not a prime, and — with no separate section for the point — carries the case for its domain-specificity as well.

What is skeletal (could lift toward a cross-domain prime). Strip the stockroom away and a thin relational structure survives: a continuous reservoir is held inside a band by hysteretic feedback — take no action while the level sits anywhere between a lower and an upper bound, act only when the lower boundary is crossed, and drive back to the ceiling — trading in-band optimality for a sharp cut in decision cost and thrash. The portable pieces are abstract — a controlled quantity that drifts, two boundaries, a triggered corrective move, and a wide zone of tolerated inaction. That skeleton is genuinely substrate-portable, and — unusually — it runs observer-free in nature, which is why the entry hands its cross-domain reach to a family of general primes: feedback supplies the closed loop (here specifically hysteretic, keyed to in-band/out-of-band rather than distance from a setpoint, not proportional/PID); reserve and buffering supply the held quantity; system_slack supplies the min-to-max gap; and boundedness supplies the in-band property the policy produces. But that banded-feedback composition is the core min–max shares, not what makes it min–max inventory.

What is domain-bound. Almost all the operative content is inventory-control furniture and none of it survives extraction intact: the minimum's safety-stock semantics (a buffer covering demand during the replenishment lead time at a target service level); the maximum as an order-up-to ceiling capping holding cost and shelf space; the replenishment action and its lead-time-and-variance calibration; the (s, S) correspondence that imports operations-research optimal-threshold results onto the two scalars; the symptom-to-knob failure decomposition (stockout, overstock); and the SKU-catalogue attention constraint that justifies banded inaction in the first place. The decisive test: remove the stockkeeper, the SKUs, the replenishment orders, and the service-level target — carry the band to a thermostat or a reservoir — and the safety-stock content, the (s, S) calibration, and the attention-economics rationale all fall away, leaving the bare banded-feedback loop that is no longer min–max inventory but its parent mechanism. The distinctive cargo is exactly the inventory-specific calibration that does not lift.

Why this does not clear the prime bar. A prime is a relational structure whose vocabulary travels and whose cross-domain transfer is recognition of the same mechanism, not analogy. Min–max's transfer is bimodal. Within inventory control — logistics, manufacturing, retail, hospital pharmacy par levels — the two scalars, the band-width tradeoff, the failure decomposition, and the (s, S) calibration carry intact as mechanism, because every case supplies the one substrate it needs: a stocked item drifting with demand under a replenishment lead time, with only the SKU and lead-time distribution changing. Beyond inventory — cash management (Miller–Orr), cloud autoscaling, reservoir outflow rules, physiological homeostasis — the control logic is a real co-instantiation of the general banded-hysteretic-feedback mechanism rather than mere metaphor; but what co-instantiates there is the parent, not "min–max inventory," and importing the inventory label would smuggle in safety-stock and (s, S) content those substrates do not have. And when the bare structural lesson is wanted cross-domain — hold a reservoir in a band, act at the boundary, tolerate the rest — it is already carried, in more general form, by feedback, reserve / buffering, system_slack, and boundedness, the primes min–max composes. The cross-domain reach belongs to those parents; "min–max inventory," as named, belongs beside (s, S), EOQ, kanban, and JIT as the inventory-control instance, its stockroom baggage staying home.

Relationships to Other Abstractions¶

Current abstraction Min–Max Inventory Domain-specific

Parents (3) — more general patterns this builds on

Min–Max Inventory is a kind of Hysteresis Prime

Min–max inventory is hysteresis specialized to stock control with a lower reorder trigger, an upper reset level, and in-band inaction.
Min–Max Inventory is part of Service Level Domain-specific

Min–max inventory contains a target service level that calibrates the safety-stock component of its minimum reorder threshold.
Min–Max Inventory presupposes Stockout Domain-specific

A min–max policy presupposes the stockout condition whose lead-time risk gives the minimum reorder point its lower-bound meaning.

Hierarchy paths (6) — routes to 5 parentless roots

Min–Max Inventory → Hysteresis → Path Dependence → Dependency

Show alternative paths (5)

Not to Be Confused With¶

The (s, S) policy. Not a rival concept but the same policy in operations-research vocabulary — s the reorder point (the minimum), S the order-up-to level (the maximum). Min–max is the practitioner's two-threshold reading of it, so the worked-out analytical characterization of optimal s and S applies directly to the two shop-floor scalars. Tell: if the terms are "reorder point" and "order-up-to level" it is (s, S); if "min" and "max" it is min–max — the object is identical, and treating them as separate ideas forfeits the theory that already solves it.
The (Q, r) / fixed-order-quantity reorder-point policy. A genuinely distinct policy: reorder when stock hits point r, but then order a fixed quantity Q rather than topping up to a ceiling. Min–max's replenishment quantity is a consequence — the gap between current stock and the maximum when the trigger fires — not a chosen parameter. Tell: does replenishment bring stock up to a fixed ceiling regardless of how far below it fell (min–max / order-up-to) or add a constant lot Q (Q, r)? The two diverge whenever stock undershoots the reorder point by varying amounts.
Economic Order Quantity (EOQ). A cost-minimizing formula for the optimal order lot size, balancing ordering cost against holding cost — an optimizer that computes a quantity. Min–max is a banded control policy that deliberately stops optimizing between thresholds; it sets a band, not a cost-optimal lot. EOQ can inform the band width, but it answers "how much per order" whereas min–max answers "when to act and to what ceiling." Tell: is the tool computing a cost-optimal quantity (EOQ) or defining a monitor-and-trigger band with tolerated inaction inside it (min–max)?
Two-bin system / kanban. The two-bin system is a physical implementation of reorder-point logic (when the first bin empties, reorder while drawing on the second) — closely kin to min–max but realized as bins rather than counted thresholds; kanban is a pull replenishment signalling discipline paced to consumption. Min–max is the general two-scalar policy of which the two-bin card is one concrete, calibration-light embodiment; kanban answers a different question (authorize replenishment on consumption) and need not carry an order-up-to ceiling. Tell: is replenishment triggered by a physical empty bin (two-bin, a min–max embodiment) or a pull card signalling downstream consumption (kanban)?
Safety stock. The buffer baked into the minimum to absorb demand and lead-time variability during replenishment — a component of the reorder point, not the trigger itself. Min–max is the whole two-threshold policy; safety stock is one summand inside its lower threshold. Conflating them leaves a stockkeeper unable to say whether a shortfall came from the reorder point being too low or the buffer being sized wrong. Tell: is the quantity the level that fires the order (the minimum) or the variability cushion inside that level (safety stock)? Part versus whole of the lower threshold.
The banded-hysteretic-feedback family (the umbrella it instances). The substrate-neutral mechanism — act only at a boundary, drive to a ceiling, tolerate any in-band state — carried by feedback (hysteretic, not proportional), reserve / buffering, system_slack, and boundedness, and co-instantiated literally in cash management (Miller–Orr), cloud autoscaling, reservoir outflow rules, and thermostat/homeostatic control. Min–max is the inventory-control specialization keyed to safety-stock and (s, S) content. Tell: those parents carry the cross-domain reach to autoscaling or cash bands — treated more fully in earlier sections — while "min–max inventory" is the stockroom instance whose lead-time and service-level calibration does not travel.

Neighborhood in Abstraction Space¶

Min–Max Inventory sits in a crowded region of the domain-specific corpus (12^th percentile for distinctiveness): several abstractions share nearly its structure, so a description that fits it tends to fit its neighbors too.

Family — Inventory & Threshold Accumulation (5 abstractions)

Nearest neighbors

Stockout — 0.89
Economic Order Quantity — 0.87
Backorder — 0.87
Service Level — 0.87
Make-to-Stock — 0.86

Computed from structural-signature embeddings · 2026-07-12