Detection Rules

Updated on 2024-11-26 GMT+08:00

View PDF

Cost Anomaly Detection helps you monitor the actual payments of both pay-per-use and yearly/monthly resources.

Pay-per-use expenditures: AI algorithms are used to intelligently identify unexpected expenditure spikes based on machine learning. If the actual cost in a day exceeds the maximum forecasted cost of that day and the difference is greater than $1 USD, a cost anomaly will be identified.

Percentage of pay-per-use costs that are impacted = (Actual cost – Maximum forecasted cost)/Maximum forecasted cost

For example, if the actual cost on July 23 was $105 USD, but the maximum forecasted cost was $100 USD, that will be identified as a cost anomaly.
Yearly/monthly expenditures: If the actual period-over-period (PoP) growth rate of MTD costs (excluding the cost of the current day) exceeds the threshold you set over the previous billing cycle and the difference is greater than $1 USD, a cost anomaly will be identified.

PoP growth rate = (Actual cost for the current month – Cost for the previous month)/Cost for the previous month

For example, if your expenditures from June 1 to 23 were $100 USD and the expenditures from July 1 to 23 (the current day is July 24) were $121 USD, and the threshold was set to 20%, then the actual growth rate (21%) exceeds the threshold, and that will be identified as a cost anomaly.
NOTE:
There are three severity levels for cost anomalies:
- Minor: > 0% and < 20%
- Major: ≥ 20% and < 50%
- Critical: ≥ 50%

Delay of Cost Anomalies

Cost anomalies are not updated in real time. You can view cost anomalies of the previous day in the afternoon of the current day. The anomalies were identified based on data collected the day before yesterday. If you have subscribed to email notifications from Cost Center, you will be notified of all cost anomalies for the previous day after 09:00 a.m. every day.

Parent topic: Cost Anomaly Detection

Previous topic: Overview

Next topic: Creating a Cost Monitor