TOC course · Lesson 6 of 6

Run a four-week constraint experiment

Complete the TOC cycle by making a prediction, changing the operating system, protecting necessary conditions, and deciding from observed flow.

Time: 15 minutes to design · 4 weeks to run Previous: connect flow to money Reference: field guide

The capstone hypothesis

Candidate: The decision process that moves RFCs from clarification to an approved, scheduled plan currently limits invoiceable RFC throughput. Two weekly 45-minute decision clinics, a complete-input rule, active-WIP cleanup, and a small protected execution block for plan-ready work will increase plan-ready transitions and invoiceable work without harming recurring-service SLA.

This hypothesis is deliberately falsifiable. It predicts both improvement and a guardrail. It also predicts that execution may become the next constraint.

The four-week protocol

WhenActionEvidence
Before Week 1Confirm system boundary, active RFC list, phase definitions, and ownership.Baseline WIP, age, transition rate, delivered/invoiceable RFCs, SLA.
Twice weeklyRun a 45-minute clinic in age order. Every RFC leaves with a customer decision, proposal, plan-ready move, or explicit blocker/closure decision.Blocked days by reason and transitions to plan-ready.
ContinuouslyAdmit only complete decision packets; do not create clarification work without an owner and next decision date; protect execution only for plan-ready work.Ready backlog, execution age, interruptions, SLA.
End of Week 2Check adherence; fix the experiment, not the hypothesis.Was the intervention actually applied?
End of Week 4Compare observed signature with prediction and choose the next focusing step.Throughput, leading measures, guardrails, new accumulation point.

Use one scorecard, not a dashboard zoo

RoleMeasureDecision it supports
LeadingMedian age in clarification; plan-ready transitions; touch-time share by phase.Is the limiting mechanism changing?
OutcomeDelivered and invoiceable RFCs; project throughput when classified.Did system output improve?
GuardrailRecurring-service SLA and incident/service-request age.Did improvement damage a necessary condition?
MovementWIP and age in execution, verification, invoicing, and collection; incident/SR capacity mix.Where should identification restart?

Interpret the result

Predicted improvement

Plan-ready and invoiceable flow rise; SLA holds. Continue until another point limits throughput.

Queue moves downstream

Clarification falls; execution or verification age rises. Return to Step 1 there.

No response, good adherence

The hypothesis is weakened. Re-examine blocker causes, system boundary, or another constraint.

No response, poor adherence

The experiment is inconclusive. Repair the operating rules before judging the hypothesis.

Throughput rises, SLA falls

The change violated a necessary condition. Redesign shared-capacity subordination.

Flow rises, cash does not

Expand the boundary: invoicing or collection may now constrain cash throughput.

Final retrieval check

Capstone submission

System and goal:
Constraint hypothesis:
Evidence supporting it:
Exploit decision:
Subordination rules:
Possible elevation:
Leading measure:
Outcome measure:
Guardrail:
Movement signal:
Disconfirming result:
Review date:
Ask follow-up questions: Submit this completed card to the agent. Feedback on the card—and later on observed results—is how the course turns exposure into demonstrated skill.

Primary reading