What are the five cost levers?

Eliminate (stop paying for things that are not used), rightsize (pay for actual usage rather than worst-case provisioning), restructure (use a cheaper pricing model or service tier for the same workload, like reserved instances or cold storage), negotiate (pay less for the same thing through annual billing, volume discounts, or renewal negotiation), and reframe (change the question entirely, like build versus buy). Every cost opportunity falls into one of the five, so walking each major line item through all five surfaces the options.

Where is the biggest opportunity in a first audit?

The eliminate lever, usually 10 to 30% of spend: idle resources, subscriptions no one logs into, duplicate tools doing the same job, old projects still running, and forgotten domains, backups, and snapshots. Go top-down, because roughly 20% of vendors account for 80% of spend, so focus the audit on the top 80% of cost and clean up the long tail only as time allows, since per-item savings there are small.

How do I avoid cost-cutting that breaks something?

Require real usage data before rightsizing rather than theoretical needs, test changes in a non-production environment (especially database rightsizing, which is medium risk), and prioritize on savings, effort, and risk together. Eliminating something genuinely unused is low risk; rightsizing a database or replacing a critical dependency needs careful planning. The cost of an outage is usually larger than the savings, so aggressive cuts without testing are a false economy.

How do I keep costs from creeping back?

Set up ongoing visibility and a policy, because optimization is not one-time. Visibility means a monthly cost review, a dashboard showing current spend and trend by category, alerts on cost spikes, and tagging so spend can be attributed to a team or project. Policy means cost justification for new vendors, approval thresholds on resource provisioning, required tagging, and a calendared quarterly review. Without the policy, the savings erode within a few cycles.

When should I negotiate a contract?

Start the conversation 60 to 90 days before renewal, with alternatives identified even if you do not intend to switch, because the alternatives are what give you bargaining power. Ask for a multi-year discount, volume tiers, and usage rightsizing, and be willing to walk, since most vendors find a way to keep you. Renewal autopilot, where annual contracts auto-renew at increases without review, is a common and avoidable waste, so calendar every renewal.

Skill · Cost optimization

Cost optimization.

Cut what is not used, rightsize what is oversized, renegotiate the rest.

Audit cloud, SaaS, and infrastructure spend, then cut what is not earning its keep without breaking what works. Every opportunity falls into one of five levers: eliminate, rightsize, restructure, negotiate, and reframe. The audit goes top-down, because roughly 20% of vendors account for 80% of spend.

The discipline is making the math work: optimization that costs more time than it saves, or that breaks something, is a net loss. And costs creep back, so the work ends in governance, not a one-time cut.

Audience: engineering and ops leads running a cost review, facing a flagged budget or a contract renewal, or auditing for waste before scaling.

View the skill on GitHub Browse the full catalog

The framework

Five levers, applied to every line item.

Every cost opportunity falls into one of these. Walk each major line item through all five.

01Eliminate: stop paying for the unused. Idle resources, subscriptions no one logs into, duplicate tools, old projects, and forgotten backups and snapshots. Usually the largest first-audit win, 10 to 30% of spend.
02Rightsize: pay for what you actually use, not worst-case provisioning from three years ago. Oversized instances, over-provisioned databases, over-purchased seats. Requires real usage data.
03Restructure: cheaper structures for the same workload. Reserved or committed-use pricing, spot instances for fault-tolerant work, cold and tiered storage, CDN caching, serverless for spiky loads.
04Negotiate: pay less for the same thing. Annual over monthly, volume discounts, multi-year commitments, and renewal negotiation. Most enterprise vendors negotiate; many consumer-tier services do not.
05Reframe: change the question. Build versus buy, a different architecture's cost profile, a different stack. The longest-lead lever, worth thinking about even when it is not actionable now.

How to cut safely

Top-down, without breaking things, then govern.

Go top-down. Roughly 20% of vendors account for 80% of spend, so audit the top 80% first and resist optimizing the small stuff while the big stuff sits untouched. The eliminate lever is usually the largest first-audit opportunity, often 10 to 30% of spend in idle resources, unused subscriptions, and forgotten snapshots.

Optimize without breaking things. Aggressive rightsizing without real usage data and a staging test causes outages, and the cost of an outage is usually larger than the savings. The math also has to work on effort: a team that spends a quarter to save 5,000 dollars a year has optimized the wrong thing, so prioritize where savings clearly exceed effort and risk.

Costs creep back, so the work is not one-time. Set up ongoing visibility (a monthly review, a cost dashboard, spike alerts, and tagging so every line has an owner) and a policy that requires cost justification for new vendors and calendars the quarterly review. Without a policy, the savings erode.

Reference files

The reference that goes alongside the SKILL.md.

references/cloud-audit-checklist.md
A practical walkthrough for auditing a cloud account (compute, storage, database, network, monitoring) for waste and rightsizing opportunities.

Browse all reference files on GitHub

Bridges to other skills

The skills around a cost review.

Cost optimization reduces spend on what exists. These cover new purchases, the performance side, and planning the larger changes.

New purchases
vendor-evaluation
Evaluating a vendor for a new purchase is a different exercise. Cost optimization audits what is already being spent; vendor evaluation decides what to buy next.
Performance, not cost
performance-optimization
A performance issue that happens to involve infrastructure is a performance problem. Reach for that skill when the goal is speed, this one when the goal is spend.
Plan the larger work
pm-spec-writing
When an optimization is a real project (a reserved-instance migration, a vendor consolidation), spec it there before executing. Cost optimization finds it; the spec plans it.
Not during an incident
incident-response
Cost work is deliberate and scheduled. An active incident is the wrong moment to be rightsizing; that belongs to incident response.

Open source under MIT

Read the SKILL.md on GitHub.

The skill source lives in the rampstackco/claude-skills repository alongside dozens of other skills covering the full lifecycle of brand and product work. This page is a structured overview; the SKILL.md is the source. MIT licensed.

View SKILL.md Browse the full catalog

Frequently asked questions.

What are the five cost levers?: Eliminate (stop paying for things that are not used), rightsize (pay for actual usage rather than worst-case provisioning), restructure (use a cheaper pricing model or service tier for the same workload, like reserved instances or cold storage), negotiate (pay less for the same thing through annual billing, volume discounts, or renewal negotiation), and reframe (change the question entirely, like build versus buy). Every cost opportunity falls into one of the five, so walking each major line item through all five surfaces the options.
Where is the biggest opportunity in a first audit?: The eliminate lever, usually 10 to 30% of spend: idle resources, subscriptions no one logs into, duplicate tools doing the same job, old projects still running, and forgotten domains, backups, and snapshots. Go top-down, because roughly 20% of vendors account for 80% of spend, so focus the audit on the top 80% of cost and clean up the long tail only as time allows, since per-item savings there are small.
How do I avoid cost-cutting that breaks something?: Require real usage data before rightsizing rather than theoretical needs, test changes in a non-production environment (especially database rightsizing, which is medium risk), and prioritize on savings, effort, and risk together. Eliminating something genuinely unused is low risk; rightsizing a database or replacing a critical dependency needs careful planning. The cost of an outage is usually larger than the savings, so aggressive cuts without testing are a false economy.
How do I keep costs from creeping back?: Set up ongoing visibility and a policy, because optimization is not one-time. Visibility means a monthly cost review, a dashboard showing current spend and trend by category, alerts on cost spikes, and tagging so spend can be attributed to a team or project. Policy means cost justification for new vendors, approval thresholds on resource provisioning, required tagging, and a calendared quarterly review. Without the policy, the savings erode within a few cycles.
When should I negotiate a contract?: Start the conversation 60 to 90 days before renewal, with alternatives identified even if you do not intend to switch, because the alternatives are what give you bargaining power. Ask for a multi-year discount, volume tiers, and usage rightsizing, and be willing to walk, since most vendors find a way to keep you. Renewal autopilot, where annual contracts auto-renew at increases without review, is a common and avoidable waste, so calendar every renewal.