Production loss accounting: explaining the gap between the plan and the output

Start from a maximum somebody will defend

The top of the tree is the output the equipment would give at design rate for every hour in the calendar, and it has to be a figure engineering will stand behind. Plants get this wrong in both directions. Using the rate the line currently achieves builds today's losses into the baseline permanently, so improvement becomes invisible. Using an unattainable supplier claim makes the whole tree look absurd and gets it dismissed on the floor. Take the demonstrated best sustained rate for each product, agree it, write it down, and revisit it only when the equipment or the product genuinely changes.

One tree, mutually exclusive branches, nothing unexplained

Below the maximum sit the deductions in sequence: time the plant chose not to run, planned stoppages such as maintenance windows and trials, unplanned breakdowns, changeovers, minor stops and idling, running below rate, and finally quality losses including scrap, rework and start-up rejects. Every branch must be defined so a supervisor can allocate an event without ambiguity, and the branches must sum exactly to the gap. The value of the discipline lies in that closure — the moment a residual appears that nobody can explain, the plant has found something it was not measuring at all.

Settling the attribution argument before it starts

Whether a stoppage is a breakdown or a changeover, whether a slow run is an equipment issue or a material issue, whether time with no orders is a commercial loss or a planning one — these arguments consume more management energy than the losses themselves. Write the definitions down, with worked examples from your own plant, and put a single person in charge of adjudicating disputed events. Consistency matters more than theoretical correctness: a category applied the same way every day produces a usable trend, while one relitigated each shift produces noise no matter how elegantly it was defined.

Every branch needs a name against it

Losses without an owner persist indefinitely because everybody assumes somebody else is working on them. Assign each significant branch to a role: unplanned stoppages to engineering, changeover time to production, rate loss to process engineering, no-demand time to the commercial side. Ownership means being asked in the weekly review what the trend is and what is being done, not being blamed for the number. Where a branch straddles functions — material-induced slow running, for instance — name one owner anyway, with the authority to convene the others, because shared accountability reliably produces none.

Converting minutes into money so losses compete for funding

A loss tree in hours does not compete against other calls on capital. Price each branch: at a constrained resource, lost time is lost contribution and the number is large; at an unconstrained one, it is only the marginal cost of the labour and utilities consumed, and chasing it may be pointless. This distinction stops the plant investing to recover hours on machines that already have spare capacity, which is a remarkably common outcome when improvement effort is directed by percentage figures rather than by cash. Efficiency measures such as overall equipment effectiveness belong to improvement practice; the money conversion is what makes the loss tree an operating tool.

Frequently asked questions

Should unmanned time be counted as a loss?

Count it, but in its own branch and label it a business decision rather than an operational failure. A line not staffed on a night shift because demand does not justify it is a commercial choice; treating it as a production loss makes the tree look catastrophic and invites people to dismiss the whole exercise. Keeping it visible and separate matters, though, because when demand rises that unstaffed time is the least costly capacity the plant owns.

How do we capture short stops that operators never record?

Assume you are missing most of them and treat that as a known gap rather than a solved problem. Manual recording will never catch a stoppage shorter than the time it takes to write it down, so short stops appear instead as unexplained rate loss. Two practical routes: measure a single line intensively for a period using timed observation, and use the residual between theoretical rate and actual output as the standing estimate. Both are more honest than a downtime log showing none.

Who should build the loss tree, engineering or finance?

Operations builds it, engineering defines the theoretical maximum, and finance supplies the cost per branch. Building it in finance produces something reconcilable to the accounts that nobody on the floor recognises; building it in engineering alone produces a technically precise model with no financial force. The construction meeting itself is valuable, because the definitions get argued out once with all three parties present instead of being disputed every month afterwards.