Forecasting, anomalies & capacity planning

Three related but separate screens, all under Intelligence in the sidebar, look ahead rather than at what’s already happened: one projects future cost, one flags unusual spend as it happens, and one projects infrastructure capacity rather than money.

Cost forecast

Intelligence → Cost Forecast (/gravity/analytics/cost-forecast) projects your spend for the next three months, based on up to six months of your organization’s trailing cost history.

Click Generate Forecast to produce a new one. The forecast for each of the next three months comes with:

• A predicted spend figure
• A confidence level - high with six or more months of history behind it, medium with four or five, low with less than that
• A trend - increasing, decreasing, or stable, based on comparing your most recent month’s spend to your trailing average

Every forecast you’ve generated stays in a history table below the chart, so you can see how a projection compared to what actually happened once the month passed.

Note

This is a statistical projection from your own historical cost data - a trend line, not a prediction that accounts for planned changes to your infrastructure. It needs cost data to already be flowing before it has anything to project from.

Cost anomalies

Intelligence → Cost Anomalies (/gravity/analytics/cost-anomalies) lists days or resources where spend deviated sharply from your organization’s own recent baseline. Detection compares your daily spend against a 30-day rolling window using two statistical methods together - an interquartile-range (IQR) outlier check and a z-score check - and flags a day as an anomaly only when both agree.

Each anomaly shows the resource type, expected spend, actual spend, the percentage deviation (highlighted when it’s over 20%), which detection method flagged it, and its status. Open anomalies can be acknowledged, which records that you’ve reviewed it and removes it from the open count.

The detection method and its thresholds are fixed by the platform - there’s no per-organization sensitivity setting to tune today. If your spend is naturally volatile, expect more anomalies to surface; there’s no way to relax the threshold from this screen.

Note

Like forecasting, anomaly detection depends on cost data already flowing, plus a recurring analysis job your account team enables per environment. An empty list can mean either “nothing unusual happened” or “detection isn’t turned on yet” - if you’re not confident which, ask your account team.

Capacity planning

Intelligence → Capacity Planning (/gravity/analytics/capacity-planning) is a different kind of forecast - it projects infrastructure capacity (CPU, memory, disk, network, instance count), not cost, for the next 30 days, using linear regression over recent metric history.

Unlike cost forecasting, this one is driven by fleet agent metric reporting, not by cloud provider billing data - an instance needs a fleet agent actively reporting metrics for it to have any capacity forecast at all. Each metric card shows current usage against a predicted peak, with a marker at the 80% threshold; when the predicted peak crosses that threshold, the card includes a sizing recommendation.

Caution

Capacity planning’s underlying weekly forecast job is not available to turn on in any deployed environment, so there is currently no way to populate this screen with data. If you need infrastructure capacity forecasting, ask your account team about the current status rather than assuming a misconfiguration on your end.

What’s next

• Cloud cost management - where the cost data behind forecasting and anomalies comes from
• Fleet monitoring - the agent metrics capacity planning is built on
• Spend caps - set hard and soft limits rather than just watching a forecast
• Recommendations - proactive alerts that can be triggered by a cost spike