Pricing And BillingSpend analysis and forecasting

42. Spend analysis and forecasting

You now have all the pieces — cloud, Daalu, LLM. This chapter is about understanding the spend, finding the surprises early, and forecasting reasonably.

The single page that matters

Billing (Chapter 22). Spend on this page once a week. Most of what follows is just structured ways to use that page.

A weekly habit

Spend 10 minutes every Monday morning on the Billing page. Three questions:

  1. Are we on track? Compare this month’s run-rate against last month’s same-date.
  2. Anything weird? Look at the per-service breakdown for any line item that’s >2× last month.
  3. Are we forecasting OK? Glance at the projected month-end.

The cost-anomaly agent (Chapter 19) flags the obvious bumps in real time, but the weekly skim catches slower drifts the agent’s thresholds don’t.

Per-account drill

When something looks off, drill into the cloud account that’s the culprit:

  • Cloud spend → [account] → Top services — which AWS/GCP service spiked.
  • Top regions — sometimes a developer accidentally spun up in a different region.
  • Top tags — which team/project is responsible.

The Assistant can help: “Show me the spend deltas for AWS account X between this week and last week, grouped by service.”

Forecasting

The Billing page’s projected month-end is a straight-line extrapolation from the month-to-date. That’s right for stable spend; wrong for any of:

  • Newly-deployed services. Their first month under-counts; next month will be higher.
  • Seasonal patterns. Holiday weeks, end-of-quarter, etc.
  • One-off events. A backup job that ran heavily once.

For tighter forecasting, use the Reports → Query tab (Chapter 23) to build a saved query that aggregates by week-of-month or day-of-week, then layer in your own adjustments.

Tag-based attribution

If your cloud resources are tagged consistently with team/project, Daalu surfaces tag-based breakdowns. If not, the breakdown shows blank-tagged spend as “untagged,” and you’ll want to fix the underlying tagging discipline first.

We see a couple of patterns:

  • Mandatory tagging at create-time. Enforced by IaC tools (Terraform sentinel, AWS Config rules). Best.
  • Periodic untag sweep. A workflow in Daalu that lists untagged resources and opens proposals to fix.

Either way, the goal is to get untagged spend below 5 %.

When the cost-anomaly agent fires

The agent threshold is configurable in Settings → Briefings → Cost anomaly. Defaults:

  • Daily total: alert if today’s spend > 1.5× trailing-7-day average.
  • Per-service: alert if a single service’s spend > 2× its trailing average.

When the agent fires, the alert payload includes:

  • The service/region/account responsible.
  • The size of the anomaly in dollars.
  • The Assistant’s inferred cause (if obvious — e.g., a new resource showed up).

Investigate by clicking through to the Billing page filtered to the relevant slice.

Inference spend specifically

Cloud LLM is metered separately and shown in the LLM router panel. Patterns to watch for:

  • Cloud usage growing faster than seat count. Usually means the local pod is offline more than you’d think. Check Managed Infra → Clusters for the GPU node’s health.
  • Suddenly heavy chat-large traffic. Usually means a workflow you authored is summarizing too much. Inspect the workflow’s run history.
  • Embedding traffic spike. Usually means a re-index job triggered. Fine if expected.

Cost-saving recommendations

The recommendations agent surfaces savings opportunities:

  • Idle EC2 / GCE instances. CPU <5 % for 14 days.
  • Unattached EBS / disks. Volumes not connected to any instance.
  • Old snapshots. RDS / disk snapshots older than your policy retention.
  • Oversized RDS / Cloud SQL instances. Instance class far larger than typical utilization warrants.
  • Reserved instance / committed-use savings. When your steady-state spend exceeds a threshold.

Each shows up on the Operations → Recommendations page (or similar; the page evolves) with an estimated dollar saving. Click to open a change proposal to apply, or dismiss with a reason.

Forecasting tools beyond Daalu

Daalu’s forecasting is “enough for the Monday glance.” For real financial planning:

  • Export the CSV from Billing → [account] → Export.
  • Pipe into your finance team’s spreadsheet, FinOps tool, or warehouse.
  • Forecast with proper models there.

We don’t try to replace Apptio, CloudHealth, or Vantage. The goal is operational awareness, not enterprise FinOps analytics.

What’s next

Chapter 43 covers tenant billing and invoicing — the mechanics of getting Daalu paid.

SKU breakdownTenant billing and invoicing