On Azure, AWS or any other cloud, we start by making your costs visible in a dashboard. Then we find the optimizations and get them applied. Finally, we set up FinOps governance that makes FinOps by design: every new application lands in the cloud already optimized.
✓ An engagement that very often pays for itself in the first month.
You pay for the cloud by the minute. That's its strength, but without steering the bill climbs every month: VMs run at night and on weekends, others are sized for a peak that never comes, and disks keep being billed long after a project ends.
Without steering, a few percent of growth per month goes unnoticed. At 5% a month, the bill grows by 80% in a year.
When tagging isn't under control, part of the bill isn't tied to any team. Nobody looks at it, so nobody optimizes it.
A known recommendation that isn't applied keeps costing money, day after day. Across a large estate, that quickly adds up to hundreds of thousands of euros a year.
These three cases are typical of what we find in every audit (illustrative amounts). Multiplied across hundreds of resources, they add up to hundreds of thousands of euros a year. FinOps is about finding them, putting a number on them and getting them fixed.
Estimate my savingsFinOps (Cloud Financial Operations) brings finance, IT and engineering teams together around one goal: spending right. We cover all three phases of the FinOps Foundation framework, and we add a fourth, essential one: staying in control over time.
You can't optimize what you can't see. The first pillar is rigorous tagging linked to the CMDB, and a clear dashboard where application owners and budget holders see exactly what they consume. They are often the first to spot what's wrong.
Illustration. When non-prod costs more than prod, there are almost always environments running for nothing or oversized.
As few mandatory tags as possible. Depending on the client, a single tag is sometimes enough: it links the cloud to the CMDB, and every other piece of information lives in the CMDB.
app-codeCMDBowner · cost center · environment · criticalityWhen we arrived at one client, 35% of resources were untagged. Today, 99.9% are correctly tagged.
Our method wasn't written in an office. It was built over eight years of engagements in several large enterprises, each with its own organization and FinOps maturity. It has been challenged against the approaches of the big consulting firms, then enriched by the experience of other CAC 40 groups.
Recommendations applied in production and savings measured on the bill. We don't deliver a report that ends up in a drawer.
Centralized or decentralized organizations, with or without chargeback, just starting or already advanced: the approach adapts to each client's starting point.
The method has been tested against the approaches of the big consulting firms. We kept what delivers concrete results and dropped the rest.
It keeps improving through exchanges with other large groups, especially within the CRIP, the French community of infrastructure and operations leaders.
Finding a recommendation isn't enough: it has to be applied. A large part of our work is removing what blocks the path between the idea and production.
The orders of magnitude below come from our engagements. They apply to the optimized resource, not to the total bill.
For non-prod VMs and VDIs: automatic shutdown every evening, restart only when someone needs it. Why not a classic Stop/Start?
Commit for 1 or 3 years on the baseline. It's the key lever for prod, the fastest to put in place, and it doesn't touch the applications. See the strategy
VMs whose CPU and RAM stay below 50% (P95 over 30 days) move down one SKU. In the cloud, you size for today's need, since scaling up is easy.
Backup policy reviewed: retention matched to each environment and to the real need, long-term backups moved from the Cool to the Cold access tier. Storage cost is roughly halved.
Filtering so only the log types you need are collected, retention matched to the need, pricing tiers adjusted (Azure Monitor, Log Analytics, Sentinel). Costs can sometimes be divided by 4. See per-service work
Unattached disks, stale snapshots, Windows Server licenses paid twice: remove what's no longer used.
Before optimizing a service, we check it's the right one. A high-end managed service or a premium SKU picked "just in case" often costs far more than the real need. Sometimes a small tool built in two days, even faster today thanks to AI, replaces an expensive service for a fraction of the price. We cost both options, development and maintenance included, before recommending.
Take a non-prod VM costing €30 a day. With a Reserved Instance or a savings plan, it costs about 50% less. With a classic Stop/Start (scheduled shutdown in the evening, scheduled restart in the morning), you save 50 to 60%.
But the two don't stack: a reserved VM that is stopped keeps paying for its reservation. A classic Stop/Start takes a lot of effort for barely more than a simple reservation.
What makes the difference: an automatic Stop every evening and a Start on demand. The VM only restarts when someone needs it, and savings reach about 90%.
In prod, the VM runs continuously: reservations apply, not Stop/Start.
A Log Analytics Workspace enabled by the security team or a default backup policy costs little on day one. Two years later, as volumes grow, they weigh tens of thousands of euros a month. The services shown below are only examples: the same approach applies to every cloud service. We have led major optimization projects on these services with the experts in charge (SOC, data teams, backup Ops). The goal is to reduce volume without losing anything in security, compliance or performance.
Every new resource and every diagnostic setting enabled by default adds volume, often without anyone using those logs.
Costs sometimes divided by 4, without losing the logs that are actually used
Deployed by security for a narrow scope, then extended source after source: firewalls, proxies, Entra ID, endpoints… Ingestion explodes.
Savings achieved without weakening detection, with every filter validated by the security team
10-year retention, non-prod backed up like prod, everything stored in Cool: the bill follows the volume, month after month.
A cost per GB that goes down, even as volume grows
Auto-scale without a sensible cap, auto-pause too long, UAT refreshed like prod: the data platform consumes well beyond budget.
Every setting is tested with the data team, with performance impact monitored
Datasets and refreshes pile up, the capacity moves up a SKU, and nobody knows what is still in use.
The right SKU at the right time, without slowing data teams down
On these services, a poorly designed saving can cost you a security alert or a backup you needed. So every project is run with the team that owns the service:
Commitments (Reserved Instances, savings plans, MACC) are the most powerful FinOps lever: up to 50% savings on a reserved VM, without changing a single application. Poorly calibrated, though, they lock in waste for three years. We build and run the whole strategy, from workload analysis to commitment renewal.
Rightsizing and Stop/Start first. You don't commit for three years to oversized or useless resources.
Usage history by VM family (SKU), region and service, to reserve only what runs all the time.
1- or 3-year Reserved Instance or savings plan, shared or single scope. Beyond VMs: storage, databases, Synapse, Fabric, and an overall commitment with the cloud provider (MACC, EDP).
Coverage and utilization tracked daily. Exchanges and refunds when the architecture changes, and renewals prepared ahead of time.
No generic template: we build every dashboard around your organization, your KPIs and your tools, including your CMDB. Leadership sees the overall trajectory, each application owner sees their scope, and everyone reads the same numbers.
| Application | App owner (CMDB) | Cost / month | Potential savings | CO₂ |
|---|---|---|---|---|
| Data platform | Data team | €77k | €14.5k | 1.9 t |
| Finance ERP | Finance IT | €48k | €9.2k | 1.1 t |
| Customer portal | Digital | €31k | €6.8k | 0.7 t |
| VDI | Workplace | €22k | €8.1k | 0.5 t |
Billing data from your cloud providers is reloaded every day. A drift shows up the next day, not when the invoice arrives.
KPIs, breakdowns and views are defined with you. Every cost is tied to its application and its owner through your CMDB, whatever the tool (ServiceNow or other).
From the executive committee to the application owner, everyone gets the level of detail they need: group, business unit, application, environment.
Every recommendation is tracked until it is applied, and the savings are measured. The freed-up budget can be reallocated, per application or globally, to new projects.
CO₂ emissions are shown next to costs, by application and business unit, with their trend over time.
Coverage, utilization and expiry dates of RIs and savings plans: nothing expires by surprise and no commitment stays under-used.
Our dashboards follow the same principle as Azure Advisor or AWS Trusted Advisor, but go much further. We collect usage metrics for every resource and run them against concrete rules, defined with your IT department and adapted to your context: environment, criticality, expected load peaks. Every recommendation is displayed, quantified, then tracked until it is implemented.
With every piece of feedback, the rules get sharper. As a result, applied recommendations deliver what they promise, and real savings match estimated savings.
| Case | Condition (last 30 days) | Recommendation |
|---|---|---|
| Oversized VM | CPU and RAM P95 < 50% | Move down one SKU |
| Unused VM | CPU P95 < 3%, average CPU ≤ 2%, network out < 2% | Stop, then delete once confirmed |
| Oversized disk | % used disk < 50% | Reduce the disk size |
| Unattached disk | Disk attached to no VM | Delete after validation |
Thresholds are starting points: they are adjusted with each client, then refined from field feedback.
A dashboard shows what happened. Monitoring watches consumption continuously, with every billing data update, and raises an alert as soon as a trajectory goes off track. You react to an overrun or a FinOps incident before it gets expensive.
An overrun can cost tens of thousands of euros per day.
And a cloud budget sends alerts, it doesn't stop anything: without monitoring, the incident is discovered on the invoice a month later. With an alert, it is stopped the same day.
A pod in a crash loop, with logs in debug mode, can generate several TB a day. At around €5 per GB ingested into Log Analytics and Sentinel (pay-as-you-go), 5 TB a day is roughly €25,000 a day.
A new source (firewall, proxy, EDR) plugged into the SIEM without filtering, and ingestion doubles overnight.
Replication or sync looping between regions or out to the Internet: data transfer is billed per GB, and flow logs follow the same curve.
A serverless function triggered by its own output, endless retries. Public case: a startup got a $72,000 bill in about two hours for a test budgeted at $7 (The Register, 2020).
The year-end forecast is recalculated every day. As soon as a scope is on track to exceed its budget, its owner gets an alert, months before the end of the fiscal year. The action plan can start in the first weeks of the year.
A sudden jump for a service or an application, day over day or month over month: a new log source, the wrong SKU deployed, a forgotten resource.
Generative AI services can run away within hours. Guardrails (budgets, quotas, daily thresholds) prevent nasty surprises.
Under-used reservation, falling coverage or upcoming expiry: every commitment stays profitable.
Data platform: 2026 forecast at €412k against a €350k budget (+18%)
Owner notified by email and Teams · 07:42Log Analytics: ingestion ×3 since yesterday, new diagnostic source
Analysis under way with the security team · 09:15Azure OpenAI: daily cost ×4 versus the 30-day average
Quota applied · 10:03RI / savings plans coverage: 85% (target 80%)
No action requiredBuilding next year's cloud budget is hard: consumption varies, projects come in, prices change. We help you estimate your future costs from your real consumption, your environments and the projects you tell us about.
Real consumption by application and environment, its trend and its seasonality.
New applications, cloud migrations, scale-ups, decommissioning: every announcement is costed (SKUs, volumes, prod and non-prod environments).
Planned savings (remediation plan, reservations), commitment expiry dates, price changes and the USD/EUR exchange rate.
A next-year budget by application and business unit, which IT can defend in front of finance, assumption by assumption.
Cautious, likely, with or without optimizations: you see the impact of each decision before you make it.
The forecast is recalculated continuously and compared with the budget. Any drift triggers an alert, and the variance is explained at the SteerCo.
Optimizing what exists isn't enough: without guardrails, every new project recreates the waste you just removed. Our promise is to reduce your costs and, above all, to keep them under control. To do that, we set up FinOps governance that makes FinOps by design: standards, building blocks and checks built into projects. From then on, everything that lands in the cloud is optimized from the start.
The result: every new application complies from day one with FinOps recommendations, and the savings you achieved aren't lost on the next project.
Let's talk about your standardsOur engagements take place in large international enterprises, on Azure and AWS, often over several years. Confidentiality is part of our job: we never publish client names or consumption data.
GreenOps applies FinOps thinking to the environmental footprint of the cloud. The two go hand in hand: a deleted VM costs nothing and emits nothing. We put carbon emissions in the same dashboard as costs, so your sustainability and IT teams look at the same numbers.
Emissions data from the cloud providers (Azure Emissions Impact Dashboard, AWS Customer Carbon Footprint Tool) is matched with billing, by application and business unit.
A hundred tonnes of CO₂ a year is roughly 50 Paris–New York round trips by plane, or more than 10 times around the Earth by car. A telling number does more for awareness than ten dashboards.
Cloud emissions are formatted, with their history, to feed your carbon accounting and your discussions with the sustainability team.
Typical breakdown of an Azure customer's emissions, according to Microsoft data:
The takeaway: the best GreenOps lever is sobriety. Fewer VMs running, less data retained and right-sized resources mean less hardware to manufacture.
| Lever | € | Real CO₂ |
|---|---|---|
| Cleaning unused resources | Yes | Yes |
| Stop/Start for non-prod and VDI | Yes | Yes |
| Rightsizing | Yes | Yes |
| Data retention policy | Yes | Yes |
| Reserved Instances / savings plans | Yes | No |
A reservation cuts the bill sharply, but it changes nothing in physical consumption. Yet some cloud provider dashboards still show lower emissions. So we separate financial savings from real footprint reductions, and we always state the source and date of carbon figures (they are published with a one- to two-month delay).
Enter your annual cloud spend to get an order of magnitude of the achievable savings.
Total bill over 12 months, in euros.
Current FinOps maturity
Indicative estimate, based on rates observed in our engagements. Our analysis gives a precise figure, resource by resource. First-month savings very often cover the cost of the engagement.
Have an Azure cost export? Drop it into our analysis tool: it puts a €/month figure on the savings already visible in the bill, resource by resource, and the file never leaves your browser.
Analyse my Azure exportFor SMEs, an engagement paid on results. For large enterprises, a day-rate engagement, working closely with your teams. Either way, we combine FinOps expertise with AI-powered analysis and automation tools.
An engagement that pays for itself in the first month.
The savings achieved in the first month are very often higher than the cost of the engagement. In reality, FinOps costs you nothing: it is funded by the money it saves you, and everything after that is pure gain.
From commitment strategy to tuning a Spark pool, we cover every FinOps lever, especially on the services that weigh most on the bill.
Pay less for the same consumption.
Consume only what is useful.
The services that drift with volume.
Accurate numbers, refreshed daily.
Make savings last.
Carbon and the new cost lines.
FinOps consultant and founder of FinOps Pilote (AB Conseil), Aymeric comes from infrastructure: servers, networks, datacenter migrations. He knows what a recommendation means for the people who have to apply it. For eight years, he has been helping large enterprises take back control of their cloud spend.
One conviction drives our whole approach: a saving only counts once it shows up on the bill.
FinOps (Cloud Financial Operations) is a discipline that brings finance, IT and engineering teams together to manage cloud spend. It rests on three phases defined by the FinOps Foundation: making costs visible (Inform), optimizing them (Optimize) and setting up lasting governance (Operate). The goal isn't to spend less at any cost, but to spend right.
GreenOps applies the FinOps approach to the cloud's environmental footprint: measuring CO₂ emissions by application and business unit, then reducing those that come from unused or oversized resources. A deleted or stopped VM costs nothing and emits nothing. Reservations, on the other hand, lower the bill without reducing the physical footprint.
To understand and control your cloud costs, know your carbon footprint and make savings: about 35% of the bill on average when nothing is managed yet. For a large enterprise, that means millions of euros a year, and a FinOps engagement pays for itself very quickly.
Typically 15 to 35%, depending on the starting FinOps maturity. The main levers are Reserved Instances and savings plans (about 50% on prod VMs), automatic Stop with Start on demand (about 90% on non-prod and VDIs), rightsizing, storage tiering and log optimization (Log Analytics, Sentinel).
Yes. We estimate your future costs from your real consumption, your environments and the projects you tell us about (new applications, migrations, decommissioning), factoring in planned savings and price changes. You get a well-argued next-year budget by application, with scenarios, followed by a forecast tracked continuously.
We ask that question every time. An expensive managed service is sometimes oversized for actual usage: a lower SKU, another service, or a small tool built in a few days may be enough. We compare the options, development and maintenance costs included, before recommending.
For SMEs, we are paid on results: 20% of the annual savings achieved, measured on the bill. No savings, no invoice. For large enterprises, the engagement is billed at a day rate, depending on scope. In practice, the savings achieved in the first month are very often higher than the cost of the engagement: FinOps costs you nothing.
Not to get started. Read-only (Reader) access to billing data and usage metrics is enough. Changes are then applied by your teams or your managed service provider, through the approval process you choose.
Every recommendation relies on cautious thresholds, for example CPU and RAM P95 below 50% over 30 days. It is first tested on a POC, and the impact is monitored right after implementation. Critical applications expecting a load peak are handled separately. Every piece of field feedback is then used to refine our recommendation rules.
In the first month for Reserved Instances, cleaning and storage tiering, which require no application change. Rightsizing and Stop/Start follow in the months after.
It's the most common blocker, especially when teams don't pay for what they consume. Two levers work: making costs visible and charged back to each application, and getting a clear mandate from leadership. We help you put both in place.
Azure and AWS are our core platforms, but we adapt to any other cloud provider: Google Cloud, Oracle Cloud, sovereign clouds or multi-cloud estates. We have already transposed our approach from Azure to AWS. The method (visibility, optimization, governance) stays the same; only the tools and pricing levers change.
Generative AI services can drive a bill up very quickly. We set up dedicated tracking and guardrails (budgets, alerting, quotas) to avoid nasty surprises.
A first 30-minute conversation to understand your context and see how we can help. No commitment.