Shop floor time capture: attendance and job booking are not the same data

Two clocks, two purposes, one reconciliation

Attended time is a contractual and legal record. Booked time is a costing and progress record. They will never match exactly, because meetings, cleaning, waiting for material, training and moving between machines all consume attended time that belongs to no job. The mistake is designing the system to force them into agreement. A better design records indirect categories explicitly, so the unbooked portion is described rather than hidden, and the size of that portion becomes a genuinely useful figure about how the plant spends its day.

Capture methods and what each costs the operator

A terminal login at the start and end of every operation gives the richest data and interrupts work most, which is why it degrades into batch entry at the end of the shift. Badge scanning at fixed stations is faster but attributes time to a location rather than to an activity. Deriving time from machine signals costs the operator nothing and captures machine state rather than labour. Supervisor entry is cheap to run and produces the least reliable data. Most plants need a mix, chosen per work centre by whether the resulting number will actually change a decision.

One operator, several machines, and a fiction

The one-operator-one-job model that most systems assume breaks immediately in a plant where somebody tends several machines, or where a team works a cell together, or where a long automatic cycle leaves an operator free for other work. Forcing that reality into single bookings produces either double-counted labour or arbitrary splits. Systems that support concurrent bookings with proportional allocation handle it honestly. Where the system cannot, the practical response is to cost those work centres by machine time and treat labour as an overhead on the cell rather than pretending to attribute it.

The gap between attended and booked time

That gap is data, not error. A widening one usually means material shortages, machine availability problems or a growing volume of unrecorded rework, and it is often the earliest visible signal that something upstream has deteriorated. The damaging response is a rule requiring every attended hour to be booked to a job, because operators will comply by attributing waiting time to whatever job is open. That single policy corrupts job costing, invalidates the comparison against estimates, and hides the very problems the gap was pointing at.

Monitoring limits, consultation and what you may use it for

Time capture is personal data about identifiable workers, and in many jurisdictions its introduction requires consultation with employee representatives before installation rather than after. Agree in advance what the data will be used for, what it will not be used for, who can see individual records, and how long it is retained. Systems introduced without that agreement tend to be resisted in ways that destroy data quality, and a workforce that believes booking terminals exist to catch them out will produce numbers that are technically complete and operationally meaningless.

Frequently asked questions

Do we still need job booking if we have machine monitoring?

They answer different questions. Machine monitoring tells you what the equipment was doing, which is enough to understand utilisation and downtime on automated processes. Job booking tells you which order consumed the time, which is what costing and progress tracking need. Where a machine runs one job at a time and the order is known to the controller, monitoring can supply booking automatically. Where operators move between jobs and machines, it cannot, and manual booking remains necessary at those work centres.

Why do operators book time to the wrong job?

Almost always because the right option was not available or was slower than the wrong one. The job was not released, the operation was not on the terminal, the previous booking was never closed, or the correct indirect category does not exist so anything open gets used. Watch what people actually do before changing the rules: the pattern of miscoded time usually maps precisely to gaps in the available options rather than to carelessness.

Can we use booking data to manage individual performance?

Legally it varies by jurisdiction and by what was agreed when the system was introduced, and practically it is usually a poor idea. The moment people believe their bookings are being scored, the data starts describing what is safe to record rather than what happened. Using aggregated data to find process problems keeps the numbers usable. Using it to rank individuals buys a short-lived improvement and a permanent loss of the information you built the system to obtain.