SkieHow does pattern detection in Insights work?
Pattern detection compares the last seven days of a group's figures with the previous twenty-eight and flags changes that stand out. Every flag comes with the reason, so you can see why it was raised. It points you at where to look; it does not tell you the cause.
Most administrators do not have time to read every graph every day. Pattern detection reads them for you and surfaces the few that moved.
The method is deliberately plain. For each group, Skie takes the last seven days of a figure and compares it with the previous twenty-eight days. If the recent week sits clearly apart from the longer baseline, it raises a pattern, such as silent share rising in one hostel block or compliance dipping in one class. Because it compares each group with its own past, a group that is always a little lower is not flagged for being lower.
Each pattern is explainable. It shows the figure, the two periods compared and what changed, so you can judge for yourself. Nothing is hidden inside a score. It also carries a confidence label, described in confidence labels, which tells you how much weight to give it.
A pattern is a pointer, not a diagnosis. It says something changed in this group; it does not say why. You supply the why, from your knowledge of timetables, holidays, a new batch of devices or a Wi-Fi change. After you act, intervention memory lets you record what you did and look at the figures before and after.
Practical tip: treat patterns as a weekly reading list. Start with the ones that have higher confidence, check the group, and ignore the ones that match something you already know, such as a holiday week.