Plot your data, read the two charts in the right order, and find out which of the three situations you are actually in before you change anything.
Chapter 4: "Is any point genuinely outside the limits, or is this just what normal looks like for this process?"
Use the metric you agreed on in Focus. Last 20 to 25 measurements, in the order they happened. Time order is not optional: a control chart reads the sequence, so sorting the values by size destroys the thing you came to look at.
Comma or full stop as decimal mark, both work. If each line has two fields, the first becomes the point label. A header row is ignored.
No data at hand? Load an example and read the charts with it. The first two are the same process with the same target: the only difference is that in the second one, someone corrects the setting after every result. Compare how wide the limits become.
The customer can be external, the next step in your own process, or a regulator. Control limits tell you what your process does. Specification limits tell you what it has to deliver. You need both to diagnose. Many processes have only an upper limit, and that is fine. Once your data is plotted, changing these updates the diagnosis straight away.
This chart plots the distance from each measurement to the one before it. It comes first because every limit on the next chart is calculated from these distances. If the movement between consecutive points is unstable, the limits you are about to draw are built on a moving foundation, and the individuals chart will mislead you.
Your measurements in time order, with the average and the natural limits of this process. The shaded band is what normal looks like here. Everything inside it is noise, however dramatic it felt on the day.
Three situations, three different treatments. Treating all three the same way is how a team spends a year taking action without moving the metric.
Nothing here is hidden. Check it against your own spreadsheet if you want to.