Overview
Cloud Cost Monitors help you proactively identify cost changes, and understand if you’re projected to go over budget, so you can investigate the cause.
- Instantly view all your cost monitors and filter or search by team, service, tag, provider, or alert status.
- See a summary of how many monitors are set up, which are alerting, and what areas of cloud spend are tracked.
- Create new cost monitors using templates and take action on monitors that need attention.
In order to configure Cloud Cost monitors, you need to have Cloud Cost Management set up.
Choose the setup that matches the cost data you want to alert on:
- Create a monitor for changes, thresholds, forecasts, budgets, and finalized cost anomalies. These monitors use finalized billing data, a 30-minute evaluation frequency, and a 48-hour delayed evaluation window, since billing data may not be available until 48 hours after usage. For example, a 7-day lookback evaluated on January 15 examines cost data from January 6 to January 13.
- Create a real-time AI anomaly monitor to alert within 15 minutes when estimated AI cost from Agent Observability increases unexpectedly.
Create a monitor
This procedure covers Cloud Cost monitors that use finalized billing data: changes, thresholds, forecasts, budgets, and finalized anomaly monitors. To alert on estimated AI cost within 15 minutes, see Create a real-time AI anomaly monitor.
To create a Cloud Cost monitor in Datadog, navigate to Cloud Cost > Analyze > Cost Monitors and click + New Cost Monitor.
Alternatively, you can set one up from Monitors > New Monitor > Cloud Cost, the main navigation, the Cloud Cost Explorer, or through Terraform.
Select a cost monitor type
You can select from the following monitor types:
| Monitor Type | Cost Metric-based | Purpose | Example |
|---|
| Changes | Yes | Detect daily, weekly, or monthly cost changes. | Alert when the difference between today’s cost and the week before is over 5%. |
| Anomalies | Yes | Identify unusual or unexpected cost patterns.
Finalized monitors exclude incomplete days and require at least 1 month of cloud cost data, because historical data is required to train the algorithm. Real-time AI anomaly monitors alert on estimated AI cost within 15 minutes. | Alert if 3 days from the past 30 days show significant cost anomalies compared to historical data, or alert within 15 minutes when AI cost increases unexpectedly. |
| Threshold | Yes | Alert when costs exceed a set value. | Set alerts when today’s total cost exceeds $10,000. |
| Forecast | Yes | Alert if forecasted costs exceeds a threshold. | Alert daily if the forecasted cost for this month is projected to exceed $500. |
| Budget | No | Alert if actual or forecasted costs exceed your budget. | Alert if the forecasted month cost is projected to exceed 90% of the assigned $10,000 budget. |
Specify which cost to track
Any cost type or metric reporting to Datadog is available for monitors. You can use custom metrics or observability metrics alongside a cost metric to monitor unit economics.
| Step | Required | Default | Example |
|---|
| Select the cost metric | Yes | All providers | azure.cost.actual |
Define the filter by | No | Nothing | aws_product:s3 |
| Group by | No | Nothing | aws_availability_zone |
| Add observability metric | No | system.cpu.user | aws.s3.all_requests |
Use the editor to define the cost types or exports.
Select an existing budget to monitor from the dropdown.
For more information, see the Cloud Cost Management documentation.
Set alert conditions
If you are using the Cost Changes monitor type, you can trigger an alert when the cost increases or decreases more than the defined threshold. The threshold can be set to either a Percentage Change or set to Dollar Amount.
If you are using the Percentage Change, you can filter out changes that are below a certain dollar threshold. For example, the monitor alerts when there is a cost change above 5% for any change that is above $500.
These conditions apply when Alert on is finalized. To alert on estimated AI cost within 15 minutes, see Create a real-time AI anomaly monitor.
For the Cost Anomalies monitor type, you can trigger an alert if the observed cost is above, below, or above or below a threshold compared to historical data.
The agile anomaly algorithm is used with two bounds and monthly seasonality.
If you are using the Cost Threshold monitor type, you can trigger an alert when the cloud cost is above, below, above or equal, or below or equal to a threshold.
If you are using the Cost Forecast monitor type, you can trigger an alert when the cloud cost is above, below, above or equal, below or equal to, equal to, or not equal to a threshold.
If you are using the Budget monitor type, you can trigger an alert when the actual or the forecasted cloud cost exceeds a percentage of the budget you selected in the previous step.
| Step | Purpose | Values |
|---|
| Evaluation basis | Whether the monitor compares actual spend or forecasted spend against the budget. | actual, forecasted |
| Granularity | Level of detail at which the cost is evaluated. | overall (total cost), per_row |
| Threshold | Percentage of budget that is utilized to trigger the alert. | Number between 0 and 100 (%) |
| Timeframe | Evaluation window used to assess if the threshold is breached. | all_months, current_month |
When you select is forecasted to reach, the monitor uses the same forecasting model as budget cards and the budget status page.
For detailed instructions on the Configure notifications and automations section, see the Notifications page.
Define permissions and audit notifications
Choose which teams, roles, users, or service accounts are allowed to view or edit the monitor. By default, all members of your organization have access.
You can also turn on Audit Notifications to alert the monitor creator and recipients whenever the monitor is changed.
Create a real-time AI anomaly monitor
Real-time AI anomaly monitors detect unexpected increases in AI cost and alert within 15 minutes. They use estimated cost from Agent Observability, not finalized cloud billing data. Datadog identifies anomalies from a rolling 4-hour evaluation window of estimated cost.
Prerequisites
- Navigate to Cloud Cost > Analyze > Cost Monitors and click + New Cost Monitor.
- Select Anomalies.
- Set Alert on to real time and the cost type to AI cost. The monitor alerts on anomalies detected within the last 15 minutes.
- Optionally, use Filter cost to to scope the cost, and Detect anomalies on to group by up to two tags.
ml_app and model_provider are listed under Preferred Tags. - Set a threshold for the estimated total cost over the next 24 hours. Enter at least 500 in your organization’s currency. The monitor alerts when Datadog detects an anomaly and the estimated total cost exceeds this threshold.
- Configure notifications.
To monitor finalized cloud billing data instead, set Alert on to finalized and follow Create a monitor. Finalized anomaly monitors use the agile anomaly algorithm, exclude incomplete days, and require at least 1 month of cloud cost history.
Other actions you can take
- View in Monitors to see your monitor’s alert history, adjust visualizations, and review how often it has triggered alerts.
- View in Explorer to open the monitor in the Cloud Cost Explorer for deeper analysis.
- Edit a monitor to update the monitor’s settings or configuration.
- Clone a monitor to create a copy of an existing monitor by choosing Actions > Clone.
- Delete a monitor to permanently remove a monitor you no longer need.
Further Reading
Additional helpful documentation, links, and articles:
Further reading
Additional helpful documentation, links, and articles:
