| Description |
Returns zero when the value of each sample in the time-series input is not zero. Returns the previous value +1 for each consecutive sample in the time-series input which is zero. |
| Input | Time Series: Multiple (Single or multiple time-series) |
| Parameters | None |
| Output | Transformed Time Series: Multiple |
| Available in | Alerts, Composites, Dashboards |
Example
| Input | 5 | 6 | 7 | 0 | 0 | 0 | 5 | 5 | 0 |
| CountZeros | 0 | 0 | 0 | 1 | 2 | 3 | 0 | 0 | 1 |
Usage Example
Use this composite function when you want to be alerted on anomalous periods without transactions, represented as data gaps.
For certain types of metrics - especially those measuring volumes of transactions (e.g., orders, add to carts, etc), it often happens that at a granular level, the volume of transactions can be very low for a given timescale (e.g., 1-5 orders per hour), and there are many periods of no transactions at all (i.e., 0 transactions).
Learning the baseline of such metrics to detect anomalous drops to 0 is challenging because the metric is often at 0, and the baseline learns it as a normal value.
