OEE software: settle the definitions before you argue about the figure

The ideal rate is negotiated, and the system insists on one

Effectiveness is computed against a reference speed, and where that reference comes from changes everything. The rate on the machine plate flatters nobody and is often unreachable with real material; the best rate ever achieved sets a bar the line hits occasionally; the standard in the routing may have been set to make costing work. Whichever is chosen, it should be documented per product and per machine, approved by engineering and production together, and reviewed when the process genuinely changes. Quietly adjusting it after a poor period is the most common way a performance figure loses all meaning.

What counts as planned time decides the headline

Excluding scheduled maintenance, breaks, meetings, trials, planned changeovers and shifts nobody staffed will lift the reported figure considerably without a single improvement on the floor. Including everything produces a low number that management dislikes and teams find demotivating. Neither is wrong, but the boundary must be written down and stable, and comparisons must never cross a boundary defined differently. A useful practice is to report the strict measure against total calendar time alongside the operational one, so the excluded time remains visible rather than disappearing from the conversation entirely. Sites that quietly widen their exclusions each year end up reporting steady improvement while shipping the same volume, which is the point at which leadership stops believing any of it.

A loss taxonomy operators can use at the machine

The classification of losses is where a system either produces improvement targets or produces noise. Long lists organised by theory get used down to the first three entries; short lists framed in the words the team uses get used properly. Build the taxonomy with the people who will select from it, keep it to a size that fits one screen, allow a catch-all with a text field and then watch what accumulates there, since an overflowing other category is a design brief for the next revision. Reclassifying historical data after changing the list is worth the effort, otherwise trends break at the point of change.

Comparing across machines, lines and sites invites nonsense

The measure was designed to track one asset against itself over time. Ranking different equipment, different products or different plants by it produces conclusions driven by definition rather than performance: a line running long stable campaigns will beat one absorbing constant changeovers, regardless of how well either is managed. Where group leadership insists on comparison, the defensible version compares each asset with its own history and shows the direction of travel. Presenting a league table across sites is the fastest way to teach every plant how to configure its exclusions generously.

The figure should point at a loss, not at a person

The moment effectiveness appears in an individual objective, data quality collapses in predictable ways. Stoppages get attributed to categories that are excluded, short losses go unrecorded, and disputes about the number replace work on the process. The systems that survive are used to find the largest loss on the largest constraint and to see whether an intervention moved it, with the headline figure treated as an index rather than a score. Improvement teams that report which loss they attacked and what happened tend to be trusted; those that report only the composite figure rarely are.

Frequently asked questions

Should we calculate effectiveness on every machine?

Only where the answer would change a decision. On a constraint that governs the output of the whole plant, understanding its losses in detail is directly valuable. On a machine with spare capacity that never holds anything up, improving its effectiveness produces inventory rather than throughput and consumes attention that belongs elsewhere. Many plants measure the constraint intensively, apply a lighter measure across the rest, and revisit the choice when the bottleneck moves, which it will as the product mix changes.

Our reported figure improved but output did not. What happened?

Usually a definitional change rather than a process change. Check whether excluded time grew, whether the reference rate was revised downward, whether short stoppages stopped being recorded, or whether the mix shifted towards products with a gentler standard. It can also be genuine improvement on a machine that was never the constraint, which raises a local figure without adding shipments. Comparing the composite against actual good units produced over the same period settles it quickly.

Where does the data come from in practice?

State and counts usually come from the machine or from sensors added for the purpose, reasons come from operators at the point of stoppage, good and rejected quantities come from production reporting or from inspection, and planned time comes from the shift calendar someone maintains. That last source is the one most often neglected, and an out-of-date calendar makes every figure wrong in a way that is difficult to spot, because the arithmetic still looks reasonable.