Cloud AWS area
The cloud costs more
than it returns.
Start with the assessment: it is free and takes five working days. We tell you who owns every euro, what stops costing once it is removed, and in what order to act. A mature FinOps programme cuts 20-25% of spend in the first year: it is not an expense to approve, it is money coming back.
As an AWS Partner, the discount starts at 8% of consumption.
It applies to AWS consumption billed through us, and it grows with volume. It is not an optimisation and it does not ask you to touch a line of infrastructure: same account, same architecture, costing less from the first month.
The problem in numbers
The saving is already there, it just has to be found
This isn't a cost to get approved: it's money you're already paying and can stop paying. That's why FinOps doesn't go through the project budget.
- 20-40%
- Cloud spend cut in the first year, between public benchmarks and our own findings
- 40-80%
- Reduction on individual workloads via right-sizing
- 1 month
- To get the first quick wins on the books
- AWS
- Selected Partner, with verified expertise
Sources: FinOps Foundation, State of FinOps 2026 · Oncode findings on live infrastructure. The actual cut depends on how long the inventory has been left alone.
What the assessment looks at
Turning off the machine nobody uses is the starting point, not the work
Anyone who opens the console finds the easy savings. The question that matters is a different one: is this resource right for the load it actually carries? The answer is not in a monthly average, and getting it wrong costs more than it saves.
Peaks versus average
A machine averaging 6% CPU should not be downsized if at two in the morning it runs at 95% for forty minutes. We look at the distribution and the high percentiles: cutting on the average is the fastest way to break a nightly batch.
Fit to the real load
Family and size compared with what the resource actually does: memory, network, I/O, not just CPU. A database struggling on disk inside a CPU-oversized instance costs twice and still performs badly.
Burned credits
Burstable instances living steadily above their baseline pay for credits. We report them separately and never add them to the savings: moving to fixed performance often costs more, and calling it a saving would be a false promise.
Performance paid for twice
Volumes configured above the IOPS and throughput already included in the price, with nobody using them. It is a line left over from a load test, and it is worth hundreds of euros a year per disk.
What stays on by itself
Disks attached to nothing, disks of stopped machines, public addresses allocated and never associated, gateways forgotten in a network no one needs. A stopped machine costs no compute, but its disk is paid for in full.
Your rates, not list price
We derive the hourly rates from your billed cost divided by the quantities actually consumed. Anyone buying at a discount who reads list prices gets promised savings that never show up on the invoice.
Assessment
5 working days
Free
An X-ray of your cloud spend
Spend attributed to projects and teams, savings quantified with the risk stated next to every line, actions put in the order they should happen. On AWS, Azure and Google Cloud.
Pilot Build
2 weeks
First recovery
Quick wins only, on one scope: right-sizing, shutting down non-production environments, storage tiering. Savings on the books within the month.
Build
6-8 weeks
Cost recovery
Commitments (Reserved Instances and Savings Plans), tagging and showback, architecture, budgets and alerts.
Stewardship
Retainer
Continuous FinOps
Agreed savings target, monthly report, anomaly watch.
What you get
Six moves you can see on the invoice
No redesign for its own sake: only things that move the number at the bottom of the bill.
Tagging and showback
Every resource attributed to a team and a product. Without it, every cost discussion ends in an opinion.
Resource right-sizing
Instances, storage and databases sized to real usage, not to what was needed two years ago.
Non-production on a schedule
Dev and staging switched off out of hours and at weekends. The fastest quick win, and almost nobody does it.
Commitments and Savings Plans
Reserved Instances and Savings Plans sized to steady-state usage, without locking you into commitments you can't hold.
Anomaly alerts
A spend spike surfaces the same day, not at month end when it's already been paid.
Architecture, where it matters
Structural work on the few places where the cost model is wrong at the root, with the return estimated before we start.
Questions we get
What people ask before signing
- A read-only role over costs and inventory: machines, disks, addresses, networks, and the usage metrics the provider already collects. No write permissions, no access to databases, buckets or the contents of machines. You can revoke it whenever you like, without telling us.
Contact us
A quote, a particular request, or simply a coffee to meet us? Write to us and we will reply to a "nano-second".