Audit → quick wins → structural fixes. No feature loss. Same engineer who cut Imohub infra 70% and Cuez 40%.
- Audit
- Architect
- Scale
monthly retainer
Who this is for
CTO or CFO watching an AWS bill climb faster than revenue — $20k to $200k a month with no clear owner and feature work that cannot stop for cleanup. AWS cost reduction consulting is the fastest path to a number you can defend to the board.
The pain today
- AWS bill up 30%+ year-over-year while actual usage stayed flat
- Reserved Instances and Savings Plans permanently on someone's 'to-do' list
- Data transfer line items mysterious, growing, and unexplained
- Cost Explorer open in a tab for months with no clear next action
- FinOps platform evaluated, quoted, and deferred because it adds spend before cutting it
The outcome you get
- 30–60% AWS bill reduction within 4–6 weeks, zero feature loss
- Reserved Instances and Savings Plans correctly sized, purchased, and tracked
- Data transfer audit with ranked fixes for your top three cost centers
- Auto-scaling, right-sizing, and Graviton migration applied where safe
- Cost monitoring in place so the bill does not quietly creep back up
Where AWS bills actually leak
AWS cost reduction consulting starts with pattern recognition. Most mid-sized bills I audit have the same five leak categories, in roughly the same order of magnitude.
Compute: EC2 and RDS instances provisioned for peak traffic that never arrived, running 24/7 at 10–15% utilization. Right-sizing and auto-scaling alone typically cut 20–35% here. Graviton3 instances (ARM-based) run 20–40% cheaper than equivalent x86 for most workloads and are drop-in replacements for containerized applications — this lever gets skipped because it sounds risky and isn't.
Storage: S3 buckets with no lifecycle rules, EBS volumes attached to terminated instances, and snapshots from two AWS accounts ago. Transitioning cold data to Glacier costs roughly 70% less per GB. The buckets are usually large; the fix takes an afternoon.
Data transfer: Inter-AZ traffic is the quiet one. Services talking to each other across availability zones pay $0.01/GB each way. At moderate request volume that compounds fast. Moving chatty services into the same AZ — or behind a VPC endpoint — often cuts 15–25% of the data transfer line.
RDS: Multi-AZ on non-production databases, IOPS set to the AWS default maximum because lowering it felt risky, read replicas that exist from an experiment nobody cleaned up. These are pure cost with no current value.
Lambda: Warm Lambda kept at high concurrency that costs more than equivalent EC2 would. The math is straightforward once the numbers are on a spreadsheet. The pattern is consistent: 30–60% of most mid-sized AWS bills is leak, not value.
30–60% AWS bill reduction within 4–6 weeks, zero feature loss
Quick-win levers vs structural fixes
I split every AWS cost reduction engagement into two phases because the timeline and risk profile are completely different.
Quick wins land in weeks one and two and typically cut 15–30% of the bill. Right-sizing: audit instance utilization via CloudWatch and Compute Optimizer, resize where usage consistently sits below 40%. Scheduling: stop non-production environments outside business hours — 40–60% savings on dev and staging. S3 lifecycle policies: transition cold data to Glacier automatically. Unused resources: EBS volumes on terminated instances, orphaned Elastic IPs, snapshots older than retention policy — these are pure waste. Graviton migration: for containerized workloads with no OS-level dependencies, switching instance family takes a deployment and saves 20–40%.
Structural fixes run from weeks three to six and add another 15–30%. Reserved Instances or Savings Plans for committed workloads deliver 30–50% compute savings on stable traffic — Savings Plans are the modern default because they flex across instance family and region. Database right-sizing and Multi-AZ review. Application-layer caching where query volume is high (this overlaps with API performance work). Data transfer architecture review.
Quick wins alone would save maybe 25%. The structural phase is what gets the bill to the 40–60% range.
120k+: Properties indexed and searchable.
Tagging, cost allocation, and stopping the creep
The most common failure mode after a successful cost reduction engagement is drift. Bills cut by 40% in month one are back within 10% of the original figure by month six. Not because the fixes broke — because new resources were provisioned with the same habits.
Tagging strategy is the fix. Every AWS resource gets a cost-center tag before it is provisioned, enforced via Service Control Policies in AWS Organizations. Cost Explorer then surfaces spend by team, environment, and product in a way that makes every engineer accountable for what they deploy. This sounds like overhead. In practice, it changes behavior faster than any policy memo.
Cost Anomaly Detection catches regressions within hours rather than at the next billing cycle. Budget alerts notify the right team, not just the AWS account owner. Quarterly recalculation of Savings Plans coverage keeps commitment-to-usage ratios from drifting as workloads evolve.
These are not complicated setups. They take a few hours to configure properly. Most AWS accounts I audit have none of them in place.
Savings Plans vs Reserved Instances: the real tradeoffs
The choice between Savings Plans and Reserved Instances is the one buyers get wrong most often, and it costs real money in both directions.
Savings Plans are more flexible — they apply across instance family, operating system, and region for Compute Savings Plans, which makes them the better default for most growing workloads. The commitment is to a dollar-per-hour spend rate, not a specific instance type. One-year partial upfront is usually the right tier: roughly 40% savings with only half the cash committed up front.
Reserved Instances still win for very stable, predictable workloads where the instance type and size are locked for the foreseeable future — large RDS databases are the clearest example. Three-year all-upfront delivers 55%+ savings when cash is available and the workload is genuinely stable.
The mistake I see most often: buying three-year all-upfront RIs for EC2 instances six months before a planned architecture change. The savings evaporate. Calculating the right commitment tier requires honest forecasting, not wishful thinking.
Savings Plans also need quarterly recalculation. As workloads grow or change, the optimal coverage rate shifts. A plan that was right-sized in January may be under-covering by April.
Case: Imohub 70% infrastructure reduction
The Imohub engagement is the clearest example of what architecture-level AWS cost reduction looks like versus incremental patching.
The project was a full rebuild of Imóveis SC into Imohub on Next.js, Laravel, Meilisearch, MongoDB, and AWS. The 70% infrastructure cost reduction was not the result of running Compute Optimizer and buying some Reserved Instances. It came from choosing the right tools for each job: Meilisearch instead of Elasticsearch for property search (a fraction of the compute cost at the same query speed), right-sized instances from day one rather than inheriting whatever the previous team provisioned, and efficient data patterns that reduced read load on the database.
The Cuez engagement shows the other direction. The primary goal was API performance — 3 seconds down to 300 milliseconds. The roughly 40% infrastructure cost reduction arrived as a secondary outcome: faster queries meant fewer resources serving the same traffic. Application-layer optimization drove infra savings, not the other way around.
Both cases combined quick wins with structural changes. Quick wins alone would have saved 25% at most. The structural work did the rest.
ROI frame and honest scoping
Before the first call, I run a simple calculation. If your AWS spend is $30k per month and a realistic optimization cuts 40%, that is $12k per month recovered. At Advisory tier ($5,499/mo), the engagement pays for itself in the first month and generates positive ROI from month two onward.
For sub-$10k per month AWS bills, the math often does not work. A 40% cut on $8k spend is $3,200 per month recovered — not enough margin to justify the engagement cost and the time it takes from your team. I will say this on the first call rather than after three weeks of work.
For bills above $50k per month, the structural phase often warrants deeper involvement implementing changes alongside your team at Fractional tier ($9,499/mo) rather than advisory-only. The recommendation is a function of your team's capacity, not of what the higher tier earns.
I don't recommend spot instances for production workloads without careful workload analysis. I don't migrate off AWS for cost alone unless the business case is very clear. I don't engage with third-party vendor savings plans that promise large discounts for unclear operational risk. Honest scoping keeps the engagement focused on durable savings.
Recent proof
A comparable engagement, delivered and documented.
Rebuilt a real estate portal at a fraction of the cost
Rebuilt Imóveis SC's real estate portal as ImoHub, a faster, more scalable successor, handling 120k+ properties with sub-second search and drastically reduced AWS costs.
Read the case studyKeep reading
Frequently asked questions
The questions prospects ask before they book.
On a first engagement, 30–60% of monthly spend is typical. Quick wins alone — right-sizing, scheduling, unused resource cleanup — usually deliver 15–30%. Structural changes like Savings Plans and data transfer architecture add another 15–30%. Very well-optimized accounts see less, maybe 10–20%. Very leaky ones see more. I give a realistic target range after the week-one audit, not before.
Quick wins land in weeks one and two. Structural changes — Savings Plans, architecture adjustments, tagging strategy — run through weeks three to six. Initial cost reduction is visible on the next billing cycle. Ongoing monitoring and quarterly Savings Plans recalculation continue after that. The 4–6 week frame is for the primary reduction, not for ongoing discipline.
Partially. Infrastructure-level wins — right-sizing, scheduling, Reserved Instances, lifecycle policies — require no code changes. Application-level wins like caching and query optimization do require code. For a code-free engagement, savings typically cap around 30–40%. Getting to 60% usually requires some application-layer work, though the scope is targeted rather than a full refactor.
Savings Plans commit to a dollar-per-hour spend rate and apply flexibly across instance family, OS, and region. Reserved Instances commit to a specific instance type and are less flexible but can win for very stable workloads like large RDS databases. For most growing AWS environments, Compute Savings Plans are the default. I size the recommendation to your actual workload stability rather than defaulting to one model.
For bills between $10k and $100k per month, AWS Cost Explorer, Compute Optimizer, and Cost Anomaly Detection are sufficient once properly configured. Purpose-built FinOps platforms add real value above $100k per month, where multi-team cost allocation and automated commitment recommendations compound. Below that threshold, the platform cost and onboarding time rarely justify it over native tooling.
Yes, with least-privilege IAM access scoped to Cost Explorer, read-only on resources, and write access only for agreed specific changes. Alternatively, I can work advisory-only with screenshare sessions where your team applies the changes. Many customers prefer the advisory model for security posture reasons and get the same savings outcome.
A rough monthly spend figure and a sense of which AWS services drive the largest line items — both visible in Cost Explorer in about two minutes. Knowing your top two or three applications and whether you have Reserved Instance or Savings Plans coverage already helps me scope the engagement before we meet. Nothing elaborate is required; a screenshot of your cost breakdown by service is enough to start.