A maintenance job card is not a downtime record. Treating them as the same thing creates misleading performance reports, incorrect maintenance priorities and potentially false estimates of lost production.
The confusion is understandable: breakdowns often create job cards, and repairs often happen while equipment is stopped. But the records describe different concepts. A reliable industrial information system must model their relationships without assuming they are identical.
1. Define each event separately
Job card or work order: authorization or documentation of work to be done. It may have request, assignment, start, completion and closure times. It can cover preventive maintenance, inspection, adjustment, calibration or corrective action.
Breakdown or failure event: a reported condition in which equipment or a component cannot fulfill a required function, under the organization’s defined criteria. Not every abnormal observation represents a confirmed failure.
Downtime interval: a defined time interval during which a machine or line is unavailable for the production activity that the chosen performance metric considers relevant.
Production loss: the difference between an evidenced or appropriately modeled production expectation and actual production. It cannot always be calculated from downtime alone.
2. Work through a realistic timeline
Suppose a job card is opened at 08:00 after a minor fault is reported. A technician begins inspection at 09:00. The machine continues producing until 10:30, when it is safely stopped for repair. Repairs end at 11:15, the machine restarts at 11:25, and the job card is closed at 14:00 after documentation.
Scroll horizontally if needed to read all columns.
| Event | Time | Meaning |
|---|---|---|
| Job card created | 08:00 | Work request exists |
| Inspection begins | 09:00 | Technical intervention started |
| Production stopped | 10:30 | Potential confirmed downtime begins |
| Repair completed | 11:15 | Physical work ends |
| Production restarted | 11:25 | Potential downtime ends |
| Job card closed | 14:00 | Administrative record finalized |
In this hypothetical example, job-card open time is six hours, technician work spans different intervals and the observed stop is 55 minutes. Even that 55-minute interval still needs classification: was it scheduled, within planned production time, attributable to the fault, and eligible for the metric being reported?
3. Build separate but linked records
A sensible data model can represent:
WorkOrder: requested work, assignees, status and work history.FailureEvent: symptom, affected asset, reported and confirmed times, classification.AvailabilityInterval: observed operational state with start and end times.ProductionRun: scheduled context, product, shift and output.EvidenceLink: why a work order is associated with an incident or interval, with confidence and source references.
One failure may create several job cards; one job card may address several issues; and some downtime has no maintenance job card at all. Therefore, a mandatory one-to-one relationship would be structurally wrong.
4. Beware of false KPI precision
Mean time to repair (MTTR) depends on which repair-time definition is used and the population of eligible incidents. Mean time between failures (MTBF) depends on an appropriate operating-time or exposure basis, confirmed failure events and a consistent population. Different organizations use different conventions; the report must state its chosen definition.
For overall equipment effectiveness (OEE), availability, performance and quality each require adequate evidence. Job-card duration alone does not supply availability, and a shift count without design speed, planned production time and quality split cannot establish OEE.
When evidence is missing, use a clear eligibility result such as “not calculable from current records,” and show the fields required.
5. Use operational evidence sources in order of authority
Potential stop evidence includes line state logs, approved PLC or historian data, shift books, supervisor incident reports, production run boundaries and other authorized records. Their authority depends on the process and data quality.
A manual shift note may record a stop accurately but with approximate timing. A machine signal may show motor inactivity but not whether production was expected. Neither source necessarily answers the full question on its own. The system should preserve both and avoid false reconciliation.
6. Handle corrections, overlaps and unknown periods
A later correction can alter the start time of a breakdown. Keep the previous recorded value, the correction author and the source that justified it. Overlapping stops need classification so they are not blindly double-counted. Unknown periods should not be silently treated as running time or downtime.
Practical metric rules
- Define exactly what counts as a failure.
- Define which production windows are planned.
- State whether setup, changeover and planned maintenance are included or excluded.
- Treat missing timestamps as unknown, not zero duration.
- Avoid counting overlapping intervals twice.
- Distinguish observed durations from estimates.
- Show the formula and input records behind every KPI.
Frequently asked questions
If a job card says “machine stopped,” is that downtime evidence? It is relevant evidence but may be insufficient for exact timing or production eligibility. Check the underlying recorded interval and its authority.
Can preventive maintenance reduce availability? It may occur in a scheduled stop or outside planned production; treatment depends on the metric definition and schedule.
Can downtime be measured manually? Yes, if the process captures credible start/end times, classifications, asset identities and corrections. Accuracy and completeness should be measured.
Should a dashboard show OEE when some fields are missing? Not as a verified KPI. Show the missing prerequisites instead.
Related reading
See Industrial Performance Intelligence, Factory Data Before AI and Digitizing Paper Records.
