Definition and Operational Context: Beyond a Simple Stop
Distinguish planned and unplanned delays and keep consistent denominators when calculating indicators.
Updated:
Planned means anticipated in the schedule, not necessarily avoidable or cheaper. Unplanned delay warrants investigation but is not invariably more expensive or proof of poor management. Separate observed state, confirmed cause and schedule status.
Operational categories
Distinguish changeover, planned maintenance, material waiting, breakdown, breaks and training. Breaks and safety tasks are not waste merely because they do not transform material. Use classification priorities and an unconfirmed-cause category.
A material shortage does not become planned just because its cause is organisational. Record expectedness separately from cause. Clear definitions reduce disagreement rather than automatically removing judgment.
Design and sample
Define unit and period, cover eligible times and retain omissions. Under a simple independent model, n₀=ceiling(z²p₀(1−p₀)/e²). For p₀=0.20, z=1.96 and e=0.03, the result is 683 (682.951 before rounding). It neither guarantees realised error within three points nor represents unobserved periods.
A category indicator is Bernoulli; its count may be binomial under independence and common probability. Large n does not make states continuous or repair round clustering. Use suitable intervals and assess dependence.
Teaching example
Assume planned delays of 12.5% and unplanned delays of 18.7%. These are scenario inputs, not sector means from 8,500 actual observations. No daily distribution is supplied to justify standard deviations of ±2.1 and ±4.3 points.
On 1,000 machine-hours, the equivalents are 125 and 187 hours. They are not yet costs: establish incremental rates, affected scope and no double counting. This does not show one category is universally two to four times more expensive.
Intervention and validation
Investigate events using records and interviews. Trial supply, sequencing, tool preparation or maintenance without reducing required controls or permits. Safety stock needs a demand-and-supply model, not an automatic conversion from binomial activity states.
Compare compatible post-change periods. Difference alone does not prove causation. Do not require covert observation or a fixed familiarisation period that “eliminates” Hawthorne effects; explain the purpose and predefine procedures.
OEE, health and software
Delay within planned production may affect availability. OEE also needs ideal cycle time and total and good counts. There is no established “unplanned delay>15% implies OEE<75%” rule or universal 65–75% wrench-time benchmark.
Risk assessment and labour compliance need their own methods; sampling does not certify them. Worksamp supports planning, capture and distribution, not guaranteed ROI or predictive maintenance. See Worksamp.
Sources: NIST, OEE, Spanish occupational-risk law.