Jump to text
On this page
Copy and source
Published source confirmed at last check

Source changed 2026-11-18 21:11:39 UTC · snapshot created 2026-11-18 21:12:35 UTC · last check 2026-11-18 22:15:20 UTC

OPS.9 - Diagnose and Treat the Current Operating Constraint

Status: Stable

OPS.9:0 - Use This When

Use this Method when an operation is not producing its accepted result as required and someone proposes to remove a bottleneck, increase capacity, automate a step or accelerate a team. The question is which relation currently limits completion, and what, if any, treatment the evidence supports.

The first useful result is a bounded constraint account: the lost result and observation window, relevant observations, the supported limiting relation or still-live explanations, and a treatment, hold or precise inquiry. Compare a serious rival if it fits the observations and distinguishing it could affect the treatment or the grounds for the required diagnosis. The result can conclude that a proposed resource addition is unsupported while specifying a useful readiness or measurement action.

Use OPS.8 when a queue policy can be repaired without settling a wider constraint claim. Use OPS.10 when the immediate question is a load/capacity/service comparison with an already adequate causal basis. A safety, clinical, engineering or financial diagnosis remains with the qualified practice that supplies it.

OPS.9:1 - Problem frame

A queue can be long because service capacity is scarce, inputs arrive in bursts, prerequisites are incomplete, acceptance is delayed, or earlier defects return as rework. Several of these conditions can occur together. A resource may be idle for much of a week and still constrain a short critical window.

Here a current constraint is a supported limiting relation for one operating result under a stated arrangement and horizon. It can concern usable capacity, readiness, sequence, quality, access, a decision or another condition. A bottleneck identified from machine or staff utilization still needs this connection to the result and horizon.

OPS.9:2 - Problem

Local speed and utilization are easy to observe. The causal relation to accepted operating results is harder. An improvement can halve touch time while leaving completion unchanged because work still waits for acceptance, a complementary input or another resource.

The opposite error demands a complete causal model before making any useful decision. A small feasibility bound, a discriminating observation or a reversible probe can sometimes rule out a proposed treatment at much lower effort.

OPS.9:3 - Forces

ForcePractical tension
urgent treatment and adequate diagnosisDelay can be costly, but the wrong treatment consumes resources and can move congestion elsewhere.
one focal constraint and distributed delayFocusing effort helps, while several short waits or alternating regimes can defeat one-resource explanations.
local measures and accepted resultsTouch time, busy time and queue length can change without changing what the operation delivers.
useful probes and protected conditionsA probe can discriminate explanations only while preserving the relevant professional, acceptance and comparison conditions.
learning and migrationA useful treatment changes the arrangement; its success can make the former constraint cease to be current.

OPS.9:4 - Solution

OPS.9:4.1 - Fix the result and the shortfall

Name the operating result, its accepted unit, demand or commitment being compared, relevant subjects, horizon and acceptance criteria. Reuse current OPS.1–OPS.4 inputs where they fit. Distinguish accepted demand, service attempts and accepted results; a failed test or reopened case can consume capacity without adding a result.

State the shortfall in terms that matter to the question, such as fewer accepted results, unacceptable time to result, missed qualified service conditions, or another explicit operating consequence. Preserve the result definition and case mix in later comparisons. A changed definition reopens the account rather than silently improving the metric.

OPS.9:4.2 - Reconstruct where completion is lost

Follow representative matters through the relevant services and joins. Recover entry, eligibility, start, completion, return and acceptance events. Distinguish the processing, setup, starvation, blocking, unavailability or rework that bears on the shortfall.

Use the event sequence and the resource calendar together. A snapshot identifies a population at one time; it does not explain a week of waiting. A utilization average can hide a critical peak. Before interpreting a rate, identify missing events and how they limit the inference about the cause.

If the result is delayed by many small waits, inspect cadence, handoff and decision timing across the path and identify the waits that can accumulate.

Choose the time question before choosing the bottleneck measure. Sustained accepted output, the first accepted result by a deadline and the elapsed wait of a particular case can have different useful treatments. For accumulated delay, sum the actual waits along the consequential path and identify which can be removed together without merely moving work into another queue. A nonlimiting resource can still govern a first-result deadline; a faster sustained pace need not produce that first result sooner.

OPS.9:4.3 - State the mechanism and resolve live alternatives

Express the candidate explanation as a relation: “Under this demand and access pattern, the final test cannot process the required eligible load before the horizon,” or “Late prerequisite delivery leaves the booked rig without eligible packages.”

Consider alternatives that could fit the available observations and matter to the requested diagnosis or treatment. Examples are insufficient usable rig time versus incomplete kits; scarce acceptance effort versus evidence repeatedly missing at acceptance; or slow processing versus arrivals concentrated near the deadline. Keep only those still plausible under the current grounds. A supported bound or directly established condition can settle the bounded question without constructing another rival.

If the available observations fit more than one explanation, state what they establish and which causes remain possible. When the question requires further discrimination, identify what result would count against each explanation and whether a distinguishing observation can be obtained. If it is unavailable, report that limit and use any treatment or hold already justified.

OPS.9:4.4 - Obtain the least burdensome discriminating basis

Select the basis appropriate to the claim and decision:

  • A resource/precedence bound can establish infeasibility under explicit demand, capability, calendar and acceptance assumptions. Show a feasible schedule as well when claiming that a proposed treatment makes the result achievable.
  • Event-history replay can reveal starvation, blocking, rework and timing relations. Its observational explanation remains limited by event quality and unobserved conditions.
  • A bounded field probe changes one consequential mechanism under the applicable professional Method and permission. State the comparison cases or windows, preserved conditions, interference, expected observations and stopping rule.
  • A specialist causal or scheduling analysis is warranted when uncertainty, confounding or coupled behavior could alter a consequential treatment or the grounds for the diagnosis, and simpler evidence is insufficient.

For a probe, keep the same accepted-result criteria and record demand, mix, resource availability and consequential conditions. Shared queues or common providers can affect nominal comparison cases too. Use an appropriate time-blocked, matched or other justified comparison; a convenient before/after label does not remove interference or confounding.

Current FPF C.16 governs the measurement chain and C.28 governs what causal use the basis supports. For the proposed use, identify the measurements or observations and their professional basis. Use a suitable existing result or obtain the missing evidence if the judgement requires it.

When a utilization ranking leaves the diagnosis unresolved, compare feasible ways of investigating the limiting mechanism. Active-period analysis can expose shifting production bottlenecks from a qualified history of machine states; sensitivity analysis can compare throughput response under a validated model. Obtain their actual state definitions, time resolution, observation window and performance criterion before interpreting the result. Price instrumentation, event reconciliation, missing/censored intervals, implementation and validation, as well as continuing observation. A five-job feasibility bound may settle a fixed deadline without any of that apparatus. Repeated regime changes that make one aggregate ranking misleading can justify it. Use the diagnostic result to identify a possible treatment or a useful probe. To claim an effect of the proposed investment, establish its consequences at the relevant operating boundary.

OPS.9:4.5 - Compare treatments from the current arrangement

If a treatment decision remains, compare the current arrangement with a feasible change that acts on the supported mechanism. Candidate treatments include protecting usable constraint time, changing release or dispatch, completing prerequisites earlier, reducing avoidable setup or rework, obtaining capacity, or changing a domain Method.

Account for implementation effort, transition loss, displaced work, downstream congestion, affected commitments and reversibility. Use C.11.CRC for the finite comparison and C.11 for the choice when several admissible alternatives remain.

A treatment is supported only to the scope of the evidence. A deterministic scenario can show feasibility; it does not establish actual field effectiveness. A probe can support a bounded effect without showing that the same treatment remains useful under another demand regime.

OPS.9:4.6 - Return the decision and remeasure

Return the result definition and window, observations, supported explanation or unresolved rival, basis and limits, any selected treatment, probe or hold, the authority for that decision, the affected uses and any observation that would reopen the conclusion.

After a treatment, inspect the accepted result and the actual load, waiting, quality and resource consequences. Reopen when demand, mix, eligibility, access, acceptance, recovery or a coupled structure changes. If another condition now limits completion, direct the next improvement decision to that condition.

OPS.9:5 - Archetypal Grounding

OPS.9:5.1 - PumpWorks: readiness or rig capacity?

PumpWorks coordinates weekly evidenced releases. In the constructed history, five eight-hour rig-access windows contain 20 test hours, 4 setup hours, 4 unavailable hours and 12 hours with no eligible job. Twelve matters appear on local boards at the final snapshot, but only four ready test packages require eight rig-hours.

A proposal to buy more rig time cites the twelve cards. The practitioner identifies the claimed result—accepted evidence-complete releases—and distinguishes it from test attempts and ready packages. The history supports a late-readiness explanation, but it does not rule out insufficient rig time in a later peak window.

The diagnosis identifies what was observed, what might explain it and what to do next.

ElementBounded result
observed losstwelve open hours without an eligible job; accepted-release shortfall still requires its own event and acceptance account
candidate explanationprerequisite delivery misses usable rig windows
serious rivaleligible demand exceeds usable rig time in the relevant deadline windows
discriminating next observationtimestamp package eligibility, usable rig slots, starts, results, returns and final acceptance for the same release population
bounded probewhere engineering and test permissions allow, prepare the next-test inputs earlier for a comparable class; preserve test/acceptance criteria and record mix, access and shared-resource effects
present decisiondo not infer a rig-capacity treatment from card count; operate the supported ready-queue policy and obtain the missing discriminatory basis

If earlier readiness reduces starvation but accepted releases remain unchanged, that result does not establish completion improvement. Inspect acceptance waiting, rework and the next consequential coupling. If usable access is then reduced by ProviderChange-P8, re-evaluate the capacity explanation under the new window.

OPS.9:5.2 - A narrow capacity treatment with a feasible witness

A different constructed service must finish five ready jobs within twelve hours. All upstream inputs are complete. Each job requires two hours of final verification and then ten minutes of certification; the certificate service has reserved slots after each possible verification completion. Acceptance requires both results, with no failures or rework in this stated scenario.

The verifier is available for eight hours. Verification completions occur at hours 2, 4, 6 and 8, so no more than four jobs can be accepted. Extra upstream processing capacity cannot change that bound: the five inputs are already ready.

The resource owner can supply two additional verifier-hours within the same twelve-hour horizon. With that change, the fifth verification completes at hour 10 and certification at hour 10 plus ten minutes. The comparison therefore returns a feasible five-job treatment candidate under the stated conditions. Its selection still uses the supplied resource cost, permission and protected consequences.

If certification availability later falls to two reserved slots, the old verifier treatment no longer supports five accepted jobs. The same result definition now exposes another limiting relation. Adding further verifier time cannot repair that missing certification capacity.

OPS.9:5.3 - Faster automation without more accepted cases

An operation automates a repeatable step and halves its touch time, but the same accepted-case rate remains. The practitioner keeps the accepted-case criteria, mix and observation window fixed and inspects waiting and returns at acceptance.

Two possible causes are insufficient acceptance capacity and repeated gaps in the evidence presented for acceptance. Hiring more acceptors addresses the first. Completing the evidence before presenting the case addresses the second. If a diagnosis or treatment choice depends on their contributions, use the least burdensome comparison that can establish them. A faster automated step alone establishes neither cause.

OPS.9:5.4 - The first result and sustained output choose different changes

Three independent serial resources A, B and C take 3, 5 and 2 hours per job. Inputs are ready at 0; transfers are immediate; buffers and downstream acceptance do not restrict this constructed deterministic case. A job is accepted only after C. The first completion is at 10; subsequent completions are five hours apart, so the fifth is at 30.

Two qualified arrangements are available before the window. An upstream fixture reduces A to 2 hours, leaving 5 and 2 elsewhere. Alternatively, moving qualified assistance from C to B produces durations 3, 4 and 4. Each has supplied permission and preserves the same accepted result; the move’s slower C is part of its actual cost, not an omitted side effect.

ArrangementFirst accepted resultFifth accepted resultPreparation burden outside the service window
Current 3/5/21030none
Upstream fixture 2/5/2929two qualified staff-hours, with a reversible fixture change
Reassigned assistance 3/4/41127three staff-hours of preparation and a changed C assignment

For one accepted result due by 9, select the fixture if its preparation can actually be supplied without displacing protected work. It changes the deadline result although B still limits the sustained pace. If the receiving requirement changes to five results by 28, that fixture is insufficient and the reassigned arrangement supplies a feasible alternative. A financial ranking needs the comparative costs. A probability of timely service needs evidence beyond these fixed durations.

Accumulated waits can make another noncapacity treatment useful. Add four evidenced 15-minute handoff waits before acceptance of the first current job: it now finishes at 11. A coordinated handoff arrangement that removes those four waits returns completion to 10 without changing B’s five-hour rate. If the required first-result deadline is 10 and the supplied coordination/cover burden is half a staff-hour outside the service window, this can be preferable to the two-hour fixture preparation. If a wait is instead required for a qualified acceptance observation, it is not removable by dispatch policy; retain it and redo the comparison. The useful intervention follows the actual path and permissible waits, not the label “nonconstraint.”

OPS.9:5.5 - Choose diagnostic effort for a changing production regime

The fixed five-job case in :5.2 needs durations, readiness and reserved calendars. Those inputs already prove that eight verifier-hours cannot supply ten required hours. A new continuous machine-state feed would not improve that conclusion. Reuse the bound and its feasible extension.

A recurring production service instead alternates its limiting behavior. The question is where movable qualified assistance could help accepted output in each regime. A competent event-history account identifies jobs, accepted outputs and losses, but its weekly totals do not distinguish the shifting intervals. The supplied instrumentation can recover complete active intervals for machines A and B using a specialist-confirmed active/passive definition, including how blocking, starvation and unavailable states are classified. That definition, synchronized times and observed interval ends are prerequisites; a generic “busy” field is insufficient.

In a small constructed diagnostic return, A’s complete active interval is 0–6 and B’s is 4–10, with no other active intervals in that ten-hour window. Both have six-hour durations. The active-period analysis returns A alone over 0–4, shared/shifting involvement over 4–6 and B alone over 6–10. Each is involved for six of ten hours; the two 60% figures include the same overlap and must not be normalized into a single winner. These are descriptive shares of this trace, not confidence levels or a forecast for next week.

The analyst needs two hours to qualify and align the existing events, three to implement and check the duration-based analysis, and fifteen minutes per shift to examine missing states and the operating return. These are supplied local effort estimates. They are worth comparing with the cost of a permanent one-machine intervention and with a simpler manual replay if shifts are rare. If the treatment decision calls for a probe, use the identified intervals to design a permitted test of assistance in the relevant regime. Inspect accepted output, quality and interference. Do not buy permanent extra A capacity from an equal weekly total or from the diagnostic label alone. A model-based sensitivity study becomes the stronger supplier when a proposed finite change alters both machines and their buffers; it must represent that coupling and the relevant regime, rather than extrapolate a local derivative into the whole investment effect.

OPS.9:6 - Bias-Annotation

BiasHow it misdirects treatmentCountermeasure
busy-resource attributionHigh utilization becomes the causal explanation.Relate usable capacity and timing to the accepted-result shortfall.
queue-length attributionVisible waiting is treated as sufficient proof of the constraint.Check eligibility, arrivals and returns. Examine a rival explanation when it fits the evidence and matters to the diagnosis.
preferred-solution biasAutomation, hiring or acquisition is selected before diagnosis.Identify what would defeat the proposed explanation. If a different mechanism remains plausible and material, compare the treatment it would require.
before/after confidenceCoincident changes are attributed to the intervention.Preserve comparison conditions and account for interference and confounding.
permanent-bottleneck identityAttention remains on a resource after conditions change.Remeasure the current limiting relation and accepted result.

OPS.9:7 - Conformance Checklist

  1. Are the accepted result and shortfall defined for the relevant population and horizon?
  2. Which limiting mechanisms do the observations distinguish, and which remain open?
  3. Where rival explanations matter to the required conclusion or treatment, are they resolved or is the remaining question identifiable?
  4. Does the selected bound, replay, probe or specialist analysis support the claim actually made?
  5. For a selected treatment, does it act on that mechanism and include consequential interactions and transition costs?
  6. Can participants act on the treatment, hold or exact probe request and recognize when it must reopen?

These questions help readers interpret the diagnosis and any decision based on it. For reliance, check the quality of the observations, the validity of any comparison, the professional premises and the relevant authority.

OPS.9:8 - Common Anti-Patterns and How to Avoid Them

Anti-patternFailureRepair
Every operation has one permanently named bottleneck.Distributed delay and changing regimes are excluded before observation.State the result-relative limiting relation and allow unresolved or multiple mechanisms.
A fast local step proves whole-operation improvement.Acceptance, joins and rework may dominate the result.Preserve the accepted unit and inspect the whole relevant path.
A capacity ratio is a feasible schedule.Precedence, access and simultaneous-resource conditions can defeat it.Supply the timed witness or retain an infeasibility bound only.
A generic causal checklist replaces the field probe.No actual admissible change or observation is specified.Name the professional Method, changed action, comparison and stop.
A successful treatment remains the answer forever.The arrangement and limiting relation have changed.Reopen the account after the material change.

OPS.9:9 - Consequences

Effort can be directed to a mechanism that affects accepted results. An exact “not yet justified” treatment decision can prevent waste while leaving a cheap useful policy or probe available.

The cost is some additional measurement and comparison discipline. The appropriate depth depends on the consequence of error: a reversible readiness trial and a major capacity acquisition need not carry the same evidence burden.

OPS.9:10 - Rationale

A constraint claim connects an arrangement to an operating result. Its meaning therefore depends on demand, acceptance, time and complementary resources. A resource label cannot retain that meaning when those conditions change.

Identifying what limits completion and deciding whether a particular treatment will help require different grounds. Retaining a live alternative and the evidence limit permits action at the smallest justified scope; a sufficient bound or established condition can finish the current decision.

OPS.9:11 - SoTA-Echoing

Practice question and selected lineSerious alternative, trade-off and pattern changeSource roles, limits and reopen condition
What should be treated when local speed does not improve accepted output? Adapt result-relative diagnosis followed by mechanism-specific treatment.Constraint-focused improvement is useful, but a forced single-resource answer can miss many waits or incomplete inputs. Sections 4.1–4.3 and 5.1/5.3 preserve the accepted result and compare a serious rival when the available grounds leave one live. Further examination of events is worthwhile if it can change the choice or substantiate a consequential diagnosis at a justified cost.Tendon, The Book of TameFlow (2022), constraint and complete-kit reasoning, is a mechanism candidate. Reinertsen, The Principles of Product Development Flow (2009), E8, explains how many small waits can accumulate. Use evidence from the operation to establish which mechanisms contribute and how. Reopen for a different waiting mechanism or demand regime.
How much evidence is enough before acting? Adapt the least burdensome bound, replay or admissible probe sufficient for the operating question. Compare live alternatives when their differences matter to the conclusion.Pure utilization ranking is cheaper but cannot support the causal move; a complete optimizer or experiment is unnecessarily costly for some bounded decisions. Sections 4.4–4.6 and 5.2 separate infeasibility, feasible scenario and actual effect.Current FPF C.16 and C.28 supply measurement and causal-use discipline; C.11.CRC supplies finite comparison. The professional operational data, probe and authority remain case inputs. Reopen when confounding, interference or consequence makes the selected basis insufficient.
Should the former constraint keep receiving improvement effort? Adopt remeasurement of the accepted result and current arrangement after change.A fixed improvement programme can keep optimizing a relation that no longer limits completion. Section 4.6 and the changed certification case make migration observable at modest extra comparison effort.The constraint-refocusing line in Tendon and the changing-information reasoning in Reinertsen suggest mechanisms to examine against the accepted result after a change. Reopen if the result, acceptance conditions or connected resources change.

The first-result/steady-pace distinction in :5.4 adapts the deterministic timing comparison in TameFlow (2022), printed p.92, and Reinertsen’s E8 accumulated-wait argument. It rejects the stronger claim that only the sustained-throughput constraint can be worth improving: a finite deadline is a different receiving result, and removing waits has a real coordination cost.

For the diagnostic choice in :4.4/:5.5, Roser et al. (2021), §§2–3, supplies a capable active-period method and duration-aware matrix construction. Its author manuscript makes state preparation and irregular timestamp handling explicit. When the timing of the changing constraint matters, use active-period analysis if the state meanings and observation effort support it. The intervals can identify involvement in that constraint. A claim about the effect of a treatment requires separate evidence. Calvo (2026), §5, compares throughput-oriented state, queue and sensitivity criteria under a serial-line model. It is a serious modeled alternative, not evidence that its steady-regime selection solves a finite deadline or arbitrary changing service. Reopen when the performance criterion, state coverage, coupling or regime makes the cheaper diagnosis insufficient.

OPS.9:12 - Relations

OPS.1–OPS.4 supply matching operating focus, subject identities and current observations. OPS.8 supplies relevant queue, readiness and buffer policy; OPS.10 supplies usable-capacity and service comparisons; OPS.11 exposes couplings that can sustain or move a limiting relation.

OPS.5–OPS.7 consume supported operating-policy consequences under their own admission, continuation, priority and commitment authority. A domain Method change returns to its owner. OPS.16 is available for the separate local operating-Method-improvement question but is not a prerequisite for diagnosing the current constraint.

Current FPF C.16, C.27.TA/C.27, C.28, C.11.CRC and C.11 govern measurement, temporal and causal claims, finite contribution and choice. Use a qualified clinical, engineering, safety, labor or financial source when the operational conclusion depends on such a diagnosis.

OPS.9:End

Referenced in the corpus

37 literal mentions in other sections. Read their context to establish the relation.