Give every factory stop a better question.

Define what counts as a stop

A factory can collect many downtime entries and still struggle to explain why output was lost. Begin with a shared event definition: which machine or line is being observed, when the event begins, when it ends and what planned interruptions are treated separately. Record production that was scheduled, not an imaginary operating day with no breaks or changeovers. A diary built for one bottleneck should not silently become a performance comparison across unrelated machines. Define the purpose first: improving a repeat problem, investigating a major loss, or identifying information that maintenance needs before choosing an intervention.

Record the event before assigning blame

Create an event record close to the time of the stop. Include the asset reference, time, product or order context, observed condition, person recording it and the work order where one exists. Use plain observations such as “feed stopped after a jam was reported” rather than an unverified verdict about operator error or defective material. NIST describes maintenance work orders as human-generated data used to capture equipment condition, faults and repairs. The practical lesson is to preserve a traceable record before memories merge several events into one convenient explanation.

Separate symptoms from confirmed causes

Keep three fields: observed symptom, suspected cause and confirmed cause. A suspicion is useful if it guides investigation, but it should not be reported as an established fact. Let the responsible technical team record the evidence supporting a confirmed cause and identify any unresolved alternatives. If a sensor alarm appears, retain the actual alarm reference and relevant context rather than translating it into a different diagnosis. Operators and maintenance staff may see different parts of the event; the diary should allow both accounts to be captured without turning the form into an argument.

Account for elapsed time carefully

Distinguish the machine’s unavailable interval from maintenance labour hours. Two people working for one hour contribute two labour-hours, while only one elapsed hour has passed. Waiting for a part, obtaining access, diagnosing and carrying out authorised work may occupy different portions of the same interruption. Define how overlapping events are handled so that losses are not counted twice at line level. This is a measurement exercise, not permission to modify maintenance procedures. Safety controls and technical work remain with qualified personnel. A precise clock cannot compensate for an unsafe or unauthorised repair.

Choose a learning priority

Review the diary with the people who own the process. Compare event frequency, elapsed loss and the importance of the affected operation; one large stop and many short stops may require different questions. Do not automatically rank a harmless delay above a more serious risk merely because it lasted longer. Choose one tractable issue and state what evidence would change the decision. An illustrative investigation might compare the same approved material and setup under controlled conditions, with the technical team defining the method. Record the question before buying a sensor or changing a setting.

Close the action and check recurrence

An action is not complete because someone entered “done”. Record the approved change, responsible person, completion date and the evidence used to verify it. Then watch for recurrence using the same event definition and an appropriate observation window. If product mix or operating hours changed, explain why the before-and-after comparison is limited. Retain unsuccessful actions as learning records rather than deleting them from the story. A good diary gradually improves the quality of questions people ask. It supports a maintenance decision with traceable observations, while refusing to claim a root cause or saving that the evidence cannot establish.