A development-heavy AWS account was burning close to $9,000 a month, a third of it pure waste. A structured FinOps program built on Clouds Analytics brought the bill down, closed the waste gap, and left a savings pipeline worth roughly $28K a year — without touching a single production workload.
By February 2026, the account showed every classic symptom of unmanaged growth in a fast-moving dev environment. Forecast spend had peaked at $8.95K/month, and Clouds Analytics flagged a 32.91% optimization ratio — nearly one in every three dollars was avoidable.
The waste was concentrated in four areas: over-provisioned EC2 instances running 24×7 in a dev environment, unattached EBS volumes still being billed with zero utility, a large backlog of orphaned EBS snapshots, and non-critical RDS databases left running outside business hours. One mid-February snapshot showed a 64.42% waste ratio on its own — a clear signal that cleanup, not just rightsizing, was overdue before any deeper optimization would move the needle.
Rather than attempt a risky, all-at-once change, the team sequenced the work from clearly reversible actions (deleting dead resources) toward changes that needed more utilization confidence (rightsizing and commitments). Four levers were applied in order.
Thirteen-plus volumes in ap-south-1 were unattached to any running instance yet still accruing charges. Once deleted, unused EBS cost dropped from $718.47/mo to $13.45/mo — a 99.97% reduction.
$613.68/mo · $7,364/yr recoveredEleven snapshots totaling roughly 6.1 TiB — most tied to volumes already marked "Not Available" — were removed from ap-south-1, clearing pure storage waste with no downstream impact.
$420.90/mo · $5,050/yr recoveredFour RDS databases were placed on a tag-based start/stop schedule so they stopped running around the clock. Across March–May this returned 5,954 stopped hours — before a single EC2 instance was even added to the policy.
$364.78 saved · ~$1,459/yr run-rateUtilization data surfaced a fleet of dev instances provisioned a tier or two above their actual load. Recommendations moved t2/t3 "small" and "medium" machines down to "micro" and "nano" sizes.
Up to $1,190/mo identified · in progressMonthly forecast fell from $8.95K to $7.13K, and the waste ratio compressed from 32% to under 19% — moving the account from a "High Waste" flag to a "Lean Profile" status on the Clouds Analytics dashboard, all within the February–May window.
| Initiative | Status | Monthly impact | Annualized |
|---|---|---|---|
| Unused EBS volume deletion | Done · Apr 13 | $613.68 | $7,364 |
| EBS snapshot cleanup | Done · Apr 4 | $420.90 | $5,050 |
| RDS instance scheduling | Ongoing | $364.78 (Mar–May) | $1,459 |
| EC2 rightsizing (dev) | In progress | Up to $1,190 | $14,280 |
Deleting dead volumes and snapshots produced immediate, zero-risk savings and shrank the surface area left to optimize.
Reversible, low-judgment actions ran first; rightsizing and commitment purchases — which need utilization confidence — followed.
Bi-monthly forecast-vs-savings snapshots kept the waste ratio visible, turning optimization into an ongoing habit rather than a one-off event.
Targeting the non-production account first captured large savings with zero customer-facing risk before touching anything live.
The same playbook now extends to production and to commitment-based discounts.
| Next step | Est. monthly saving |
|---|---|
| EC2 rightsizing (prod) — move to smaller / Graviton (t4g) types | $788–$1,190 |
| Purchase 1-yr EC2 Compute Savings Plan (~40% on baseline) | $580–$600 |
| RDS rightsizing to smaller instance classes | $45–$180 |
| Expand tag start/stop scheduling to remaining EC2 dev fleet | $200+ |
| Combined, forecast below $5.5K/month | ~39% below Feb peak |