The Domino Effect of Interferences in Production
Investigate how waiting spreads between stations without confusing observed associations with causal relationships.
Updated:
Interference and cascading stops
An interruption can starve a downstream stage or block an upstream one. A buffer or spare capacity may also absorb it. Propagation depends on sequence, duration, inventory, and constraints; there is no universal OEE reduction per interruption.
The study should locate recurring states and support investigation of the chain, rather than automatically blaming the station where waiting is observed.
Two complementary records
Sampling estimates time proportions in defined states. Event records provide stop start times, finishes, and sequence. Investigating propagation often requires both: finding two stopped stations in one round does not prove that one stopped the other.
Distinguish material starvation, output blocking, confirmed breakdown, adjustment, and an unresolved state. Record a cause only where evidence exists; keep state and cause in separate fields.
Simulated example
Assume A feeds B through a buffer. If A stops, B may continue until it consumes the available material, then become starved. An initially empty buffer can make the effect immediate. If B was already stopped for another reason, adding another complete loss would double count.
Investigate timestamps, buffer occupancy, and good output. Do not add all workstation stopped minutes as line stopped minutes: they can overlap. This illustrates mechanisms, not a customer outcome or sector benchmark.
Distinguish event frequency from time
Ten short stops can occupy less time than one long stop. Pareto charts of event counts and minutes answer different questions. Random snapshots are more likely to encounter long events; they do not directly produce the duration distribution per event.
Plan the sample for the proportion of interest. In n=z²p(1−p)/e², p is the expected proportion and e an absolute margin. Coverage and dependence require separate checks. NIST explains the proportion model.
Connect the diagnosis with OEE and cost
Calculate OEE using equipment data: A=run time/planned time; P=ideal cycle time×total count/run time; Q=good count/total count. See the formulation. Do not replace quality with inspection time or allocate the same loss twice to availability and performance.
Value only effects that can be justified. Potential capacity is not sales, and released hours are not necessarily avoided payroll. Separate additional cost, realized savings, and contribution from additional fulfilled sales, without double counting.
Act and check again
Prioritize by impact, safety, and feasibility, not a mandatory 80/20 split. Form a specific hypothesis—such as staging material before releasing an order—and assess the result against a comparable baseline. Demand or product-mix changes can explain part of an apparent improvement.