- \n
- Two lines can share the same OEE and have opposite problems.<\/li>\n
- Rolling up three ratios can mask real changes.<\/li>\n
- One control chart per component, per shift, shows which loss moved.<\/li>\n
- Limits come from process data, not from targets.<\/li>\n<\/ul>\n<\/div>\n
One OEE number, three different stories<\/h2>\n
OEE is the product of three ratios: availability, performance and quality. That makes it a convenient headline figure, and a poor diagnostic tool. Take an illustrative case. For example, one line runs at 90% availability, 80% performance and 95% quality; another runs at 80% availability, 90% performance and 95% quality. Both show an OEE of about 68%. Same number, opposite problems: the first line loses speed, the second loses time to stops. A single OEE chart puts them side by side as if they were equivalent.<\/p>\n
The multiplication also hides movements that cancel each other out. If availability drops while performance rises over the same period, the product can stay flat. The OEE chart looks calm while two real changes are under way on the line.<\/p>\n
Dr. Bill McNeese: why components must be charted separately<\/h2>\n
Dr. Bill McNeese, of BPI Consulting, publisher of SPC for Excel, reviews common misconceptions about control charts in 5 SPC Myths<\/a>, from the idea that one chart type fits every situation to the habit of removing out-of-control points. His starting definition is a useful one for any production team: “a control chart is a movie of the variation in your process over time.”<\/p><\/blockquote>\n
In another article on the same site, Is Overall Equipment Effectiveness (OEE) an Effective Metric?<\/a>, he applies this view to OEE directly. His warning: “When you begin to rollup variables, you run the risk of masking signals.” His worked example shows availability falling and performance efficiency rising while the OEE control chart stays in control and shows neither change. His conclusion is to look at the individual components in control chart format. The method and the statistics come from his work; this article turns them into a routine for a production meeting, with shift-level data.<\/p>\n
![\"Four-step](\"https:\/\/teeptrak.com\/wp-content\/uploads\/2026\/11\/teeptrak-oee-control-chart-en.png\")
<\/figure>\nShift variation: what is normal and what is a signal<\/h2>\n
OEE never stays still from one shift to the next, and most of that movement is ordinary: a different product mix, one more changeover, a start-up after a weekend, a smaller crew. These are common causes. They are part of how the process currently works, and reacting to each of them one shift at a time leads to unnecessary adjustments.<\/p>\n
A real signal is different: a component moves outside the range the process normally produces, or drifts consistently in one direction. To see it, the control limits must come from the process’s own data, not from a target. A target says where you want to be; the limits say where the process naturally goes.<\/p>\n
How you group the data also shapes what the chart can show. McNeese stresses that the way a process is sampled determines the variation the chart examines. In practice, one point per shift and per machine is usually the right grain for OEE components. If one shift team runs systematically lower (night shifts often do), give it its own chart or stratify the data rather than mixing all shifts into one set of limits, which would widen them and hide signals.<\/p>\n
Step by step: setting up three component charts<\/h2>\n
- \n
- Freeze the definitions.<\/strong> Write down what counts as planned production time, which ideal cycle time is used for performance, and what counts as a good part. Changing a definition later creates false signals.<\/li>\n
- Collect one value per shift and per machine.<\/strong> Gather at least 20 to 25 shifts of availability, performance and quality before drawing limits.<\/li>\n
- Plot three individuals and moving range charts.<\/strong> One for availability, one for performance, one for quality. Where output varies a lot, a p chart on the scrap proportion can replace the quality chart.<\/li>\n
- Calculate the limits from the data, then hold them.<\/strong> Use these limits to judge the following weeks instead of recalculating them every time a point is added.<\/li>\n
- Apply the same detection rules to each chart.<\/strong> A point beyond a limit, a run of points on one side of the center line, a steady trend.<\/li>\n
- Keep the OEE chart as a summary.<\/strong> Report it, but decide from the component charts.<\/li>\n
- Recalculate the limits only after a confirmed change.<\/strong> A new tool, a new product family or a completed improvement project justifies new limits; a bad week does not.<\/li>\n<\/ol>\n
Reading the three charts together<\/h2>\n
Once the three charts sit on the same page, typical patterns appear. Each one points to a different kind of loss and a different owner.<\/p>\n
- Collect one value per shift and per machine.<\/strong> Gather at least 20 to 25 shifts of availability, performance and quality before drawing limits.<\/li>\n
- Freeze the definitions.<\/strong> Write down what counts as planned production time, which ideal cycle time is used for performance, and what counts as a good part. Changing a definition later creates false signals.<\/li>\n