SkieWhat do the confidence labels on Insights patterns mean?
A confidence label tells you how much weight to give a detected pattern. It reflects how clear the change is and how much data lies behind it. A pattern with a higher label has a clear, well-supported change; a lower one may be a small group or a short blip. It is a guide to attention, not a probability.
Statistics on small groups can mislead. A class of eight devices swings a great deal when one tablet goes flat. To stop that being presented as a trend, each pattern carries a confidence label.
The label is based on two things that you can reason about without any special knowledge: how large the change is compared with the usual variation, and how much data stands behind it, meaning how many devices and days. A big move across a large group scores higher. A small move in a small group scores lower, however dramatic the percentage looks.
How to use it: start with the higher-confidence patterns, because they are the ones most likely to be real and worth a visit. Read the lower-confidence ones with a lighter hand, and watch whether they persist into the next week before acting. A pattern that fades by itself was probably a blip.
The label does not say how serious a pattern is, and it does not say whether the cause is good or bad. A highly confident drop in silence might simply mean the Wi-Fi was fixed. It only says how sure Skie is that something changed. Reading the pattern's own explanation, as covered in pattern detection, tells you the rest.
Skie does not publish a numeric scale for the labels, and the labels are not a promise about the future. Practical tip: if you are explaining a pattern to a principal or a board, describe the label in plain words, as how clear the evidence is.