Sensitivity sets how far a value has to move from the expected value before the monitor raises an anomaly. Monitors use balanced by default. Change it up front only when you already know what to expect from the data, for example a financial metric where small changes matter. Otherwise, change it when a monitor alerts too often or not often enough.
Precise raises an anomaly on small changes. Use it for stable time series and for metrics where a small change is already a problem.
Balanced is the default and fits most monitors.
Relaxed raises an anomaly only on large changes. Use it when small changes are normal for the data and you only want to hear about big ones.
Sensitivity changes the threshold only. At every level, a sharp change raises an anomaly on the first measurement, and a smaller change raises one only after it repeats over several measurements in a row. See How monitors work.
Start with balanced and look at the first predictions. If the monitor misses issues you care about, change it to precise. If you get too many alerts, change it to relaxed, or ask yourself if you need the monitor at all.
Select a monitor, and under Settings, select a sensitivity level. The monitor is updated immediately.Table stats and field stats monitors can also set a sensitivity per metric, for example precise on row count and relaxed on delay.
If many tables have the same expectations, you can change the sensitivity of their monitors in one go.In the , select the monitors you want to change, click Change sensitivity, and pick the new level. All selected monitors are updated immediately.
If you have many monitors, use the filter menu at the top to narrow down your selection. To search for monitors that match a specific string, use the Query option.
