How to learn FinOps: from the framework to your own bill

FinOps is learned twice: once from the FinOps Foundation's framework, which names what the practice does, and once on a real bill, where each name becomes a column, a tag or a setting. This guide does both:

• The map: the framework, the courses and certifications at their prices, the book, the data format.
• The practice: one exercise per capability on your own cloud account, linked to the article on this site where that part of the bill is taken apart.

The exercises use AWS, like the service articles of this blog; the framework and FOCUS, its billing data format, are provider-neutral.

The path at a glance

1. 1

   Introduction to FinOpsFree · 30 min

   The FinOps Foundation's first course, "intended for people who are brand new to the concept of FinOps".

2. 2

   The framework pagesFree

   Principles, domains, capabilities and phases, read on finops.org: what every exam tests.

3. 3

   Introduction to FOCUSFree · 1 h

   What the billing data format is and why it matters, before you export your own data.

4. 4

   The nine exercises on your own accountYour bill

   One capability at a time, on a real bill: read it, export it, allocate it, watch it, reduce it, measure it.

5. 5

   Cloud FinOps, second editionBook

   The FinOps Foundation's book, read alongside the exercises: its parts follow Inform, Optimize and Operate.

6. 6

   FinOps Certified Practitioner$325

   The entry certification, once the exercises hold no surprise: $325 for the exam alone, $500 with the course.

7. 7

   The certification for your role$400 to $500

   Engineer if you build infrastructure, FOCUS Analyst if you own the cost data, Professional to lead a practice.

The order is this guide's recommendation, not the Foundation's. The rest of the page takes each step in turn.

What you are learning: FinOps in the framework's words

The FinOps Foundation defines FinOps as "an operational framework and cultural practice which maximizes the business value of technology, enables timely data-driven decision making, and creates financial accountability through collaboration between engineering, finance, and business teams."

The goal of FinOps isn't to save money, but to maximize revenue or business value through the cloud.

So learning FinOps is not learning a list of discounts. It is learning to see what technology costs, to tie each cost to someone who can act on it, and to decide with that information. For the longer answer, see What is FinOps?. Courses, certifications and job descriptions all use the framework's words, so learn them by name.

Six principles

The principles page lists them "in no particular order". The right-hand column is this guide's reading, not the Foundation's text.

Four domains, 22 capabilities

• Understand Usage & Cost

  • Data Ingestion
  • Allocation
  • Reporting & Analytics
  • Anomaly Management

• Quantify Business Value

  • Planning & Estimating
  • Forecasting
  • Budgeting
  • KPIs & Benchmarking
  • Unit Economics

• Optimize Usage & Cost

  • Architecting & Workload Placement
  • Usage Optimization
  • Rate Optimization
  • Licensing & SaaS
  • Sustainability

• Manage the FinOps Practice

  • Executive Strategy Alignment
  • FinOps Practice Operations
  • Governance, Policy & Risk
  • FinOps Education & Enablement
  • Invoicing & Chargeback
  • FinOps Assessment
  • Automation, Tools & Services
  • Intersecting Disciplines

The framework overview groups the work this way. The exercises below are the capabilities you can practise alone; the ones that need a company around you, such as chargeback or executive alignment, are left for later.

Three phases, used as a way to learn

"The FinOps journey consists of Inform, Optimize and Operate." The phases page insists they are a loop: "FinOps practitioners cycle through the Inform, Optimize, and Operate phases rapidly as they look at the organization's technology use." The nine exercises follow the same loop:

1. 1

   Inform: learn to see

   Read the bill, export it, give each cost an owner, get warned when it moves (exercises 1 to 4).

2. 2

   Optimize: learn to act

   Remove what bills for nothing, rightsize, then commit and price designs before building (exercises 5 to 7).

3. 3

   Operate: learn to keep it going

   Measure unit cost and run a monthly review with owners and one automation (exercises 8 and 9).

Personas and scopes

• Personas: the core ones are FinOps Practitioner, Engineering, Finance, Leadership, Procurement and Product; the allied ones ITAM, ITFM, ITSM, Security and Sustainability.
• Scopes: "a defined segment of spending across Technology Categories, aligned to business constructs–such as products, cost centers, or environment" (framework overview).
• Technology categories: public cloud is one among others; the Foundation's courses also cover SaaS, data centers, data cloud platforms and AI (FinOps Foundation catalog).

In short: Learn the framework's words first, six principles, four domains, 22 capabilities and three phases, because every course and every certification is built on them.

Where to learn it: courses, certifications, the book, the docs

Start free

• Introduction to FinOps: "Approximately 30 minutes", for "people who are brand new to the concept of FinOps."
• Introduction to FOCUS: about an hour, on what FOCUS is and why it matters.
• The framework pages, from the overview to each capability: what the exams test.
• The providers' own guides: the AWS Well-Architected Cost Optimization Pillar, whose first principle is "Implement cloud financial management", and Microsoft's FinOps documentation.

The certifications, and what they cost

Every certification in detail, with the bundles and the recertification: FinOps certification.

What the Practitioner pages add (course page, exam page):

• "There are no formal prerequisites required"; "familiarity with a major service provider such as AWS, Azure, or Google Cloud" is recommended.
• "You have three attempts to pass the exam within 12 months of the purchase date", and "You are not proctored while taking the exam".
• The certification is "Valid for 24 months".

The catalog also sells bundles, from $600 for the Practitioner with the FOCUS Analyst to $2,495 for the full catalog, a $250 course on FinOps for containers, and a recertification exam for engineers at $162.50 (FinOps Foundation catalog).

Which one first?

• Most readers: the Practitioner, after the free courses and the exercises. The exam alone is $325; buy the $500 course only if the framework pages were not enough.
• Engineers who build infrastructure: the Engineer next.
• Whoever will own the cost data: the FOCUS Analyst next.

This order is a recommendation, not the Foundation's.

The book: Cloud FinOps, second edition

Cloud FinOps (O'Reilly) is written by J.R. Storment, "Executive Director of the FinOps Foundation", and Mike Fuller, "CTO at the FinOps Foundation" (the FinOps Foundation's book page). Its 27 chapters follow the phases:

• Introduction, with chapter 5, "Anatomy of the Cloud Bill".
• Inform: "Allocation: No Dollar Left Behind", "Tags, Labels and Accounts, Oh My!", "Accurate Forecasting".
• Optimize: "Using Less: Usage Optimization", "Paying Less: Rate Optimization", two chapters on commitments.
• Operate: "Automating Cost Management", "Metric Driven Cost Optimization", "Partnering with Engineering to Enable FinOps".

Each exercise below names its chapter.

FOCUS: the billing data format to learn once

FOCUS, the FinOps Open Cost & Usage Specification, "is an open specification that normalizes billing datasets across AI, cloud, SaaS, data center, and other technology vendors".

• Latest version: 1.4 (specification), ratified by the FOCUS Steering Committee on June 4, 2026 (announcement).
• On AWS: the export available is "FOCUS 1.2 with AWS columns" (AWS Data Exports); providers lag behind the specification.

Learning to read one FOCUS dataset is learning to read the bills of every provider that exports one.

In short: Start with the free introduction and the framework pages, practise on your own bill, and pay for the Practitioner exam once the exercises below hold no surprise.

What the State of FinOps 2026 says to someone learning

The FinOps Foundation's yearly survey of practitioners, the State of FinOps 2026, shows where the work is going. Its figures, with the number of answers (N) the report gives:

• 1,192respondentsThe sixth annual survey, "representing $83bn+ in annual cloud spend".
• 78%of practices report into the CTO or CIOTeams reporting to the CFO declined to 8% (N=523).
• 60%use centralized enablementThe most common team structure, followed by hub-and-spoke at 21% (N=657).
• 8 to 10practitioners at $100M+ of spendPlus 3 to 10 contractors, on average: "Team sizes remain lean".

AI	98%
SaaS (managed or planned within a year)	90%
Licensing	64%
Private cloud	57%
Data center	48%

Source: State of FinOps 2026 Report (FinOps Foundation)

What it means for what you learn:

• Optimization is still the core. "Workload optimization and waste reduction remain the single top current priority for FinOps teams" (N=685): exercises 5 and 6.
• AI is the skill teams lack. "AI cost management stands out as the single most desired skillset across organizations of all sizes" (N=695); the report gives no percentage for it. The catalog's course on it is the AI Value certification: "Build the skills to manage and optimize AI costs". What an AI bill counts and the levers on it: FinOps for AI.
• Public cloud is no longer the whole scope. SaaS, licensing and data centers now sit beside it, as the chart shows: the framework's scopes and technology categories are worth learning early.
• Teams are small and sit with technology. Most report into the CTO or CIO, and even the largest spenders run lean teams: learn to speak to engineers, who make most of the changes.

Applying each precept: one exercise per capability

Reading the framework teaches its words; only a bill teaches what they cost. Each exercise takes one capability, gives the steps on an AWS account, and links where that part of the bill is explained line by line. A personal account with a few running resources is enough for most of them.

1. 1

   Read your billReporting & Analytics

   Group last month by service, then by usage type, and compare unblended with amortized cost.

2. 2

   Export your billing dataData Ingestion

   Create a FOCUS export, query one month in SQL and make its total match the bill.

3. 3

   Tag and allocateAllocation

   Three tag keys, activated, and a rule for what cannot be tagged; track the untagged share.

4. 4

   Budgets and alertsBudgeting · Anomaly Management

   One monthly budget with an actual and a forecast alert, and one anomaly monitor.

5. 5

   Rightsize and clean upUsage Optimization

   Find what bills for nothing, then what is too big, before buying any discount.

6. 6

   Commit to what runsRate Optimization

   Check the Savings Plans recommendation against the usage that runs every hour.

7. 7

   Price before you buildPlanning & Estimating

   Price two designs of the same feature before writing either.

8. 8

   Measure unit costUnit Economics

   Divide an allocated cost by a unit the business counts, and track it monthly.

9. 9

   Run the practiceFinOps Practice Operations

   A monthly review, a decision log and one automation: the first eight, as a routine.

1. Read your bill (Reporting & Analytics)

The Inform phase: "examine technology cost, usage and efficiency data". Cost Explorer's console is free: "You can view your costs and usage using the Cost Explorer user interface free of charge."

1. Open Cost Explorer, set the last full month, monthly granularity, grouped by service. Write down the three largest services and their share.
2. Group the same month by usage type. The surprising lines are the exercise: data transfer, NAT gateway processing, the volumes and addresses filed under "EC2-Other".
3. Switch from unblended to amortized cost. If the totals differ, an upfront commitment is being spread over the months.
4. Remember the data is a day behind: Cost Explorer "refreshes your cost data at least once every 24 hours".

Go deeper: the cost types in Cost Explorer, and the bill-reading section of each service article, from Amazon EC2 to Amazon S3. Book: chapter 5. Series: How to learn FinOps – Read your bill.

2. Export your billing data (Data Ingestion)

The fourth principle: "FinOps data should be accessible, timely, and accurate". An export lets you ask your own questions, in SQL.

1. In AWS Data Exports, export the table "FOCUS 1.2 with AWS columns" to an S3 bucket you own, at daily granularity: the table accepts "HOURLY, DAILY, MONTHLY" (AWS Data Exports).
2. Once a month is delivered, query it, for example with Amazon Athena: total BilledCost by ServiceName, then by ChargeCategory, columns the FOCUS specification defines.
3. Compare the total with the Bills page. Finding why they differ (taxes, credits, the day of delivery) teaches more than any course.

What it costs: the files are billed by Amazon S3 and Amazon Athena; in a management account the table "includes cost and usage data for the management account and all member accounts in your organization" (AWS Data Exports).

Go deeper: the export compared with the console, and the free Introduction to FOCUS. Series: How to learn FinOps – Export your billing data.

3. Tag and allocate (Allocation)

Ownership made real: "Define strategies to assign and share cost and usage using accounts, tags, labels, and other metadata, creating accountability among teams and projects within an organization". And: "Most organizations will also have shared costs to be allocated."

1. Choose three tag keys and their allowed values, for example an owner, an application and an environment.
2. Apply them, then activate them as cost allocation tags. "You must activate both types of tags separately before they can appear in Cost Explorer or on a cost allocation report", and they "can take up to 24 hours to appear" (cost allocation tags).
3. A few days later, group Cost Explorer by each tag. The untagged group is your unallocated spend: write down its share and bring it down month after month.
4. For what cannot be tagged, such as support or shared networking, write the split rule: by share of spend, evenly, or to one owner.

Go deeper: cost allocation tags and Cost Categories. Book: chapters 11 and 12. Series: How to learn FinOps – Tag and allocate.

4. Budgets and alerts (Budgeting, Anomaly Management)

AWS Budgets alerts on "both actual (after accruing) and forecasted (before accruing) spends"; Cost Anomaly Detection "uses machine learning models to detect and alert on anomalous spend patterns in your deployed AWS services".

1. Create a monthly cost budget at last month's total, with an actual alert and a forecast alert, both at 100%.
2. Create an anomaly monitor on your AWS services, with an email subscription.
3. Learn how late each can be: Budgets data "is updated up to three times a day", and "You might incur additional costs or usage that exceed your budget notification threshold before AWS Budgets can notify you" (AWS Budgets); anomaly detection "runs approximately three times a day" (Cost Anomaly Detection). A budget is a warning, not a cap.

Go deeper: Cost Anomaly Detection, and what AWS Budgets charges. Book: chapter 13. Series: How to learn FinOps – Budgets and alerts.

5. Rightsize and clean up (Usage Optimization)

Usage Optimization "is a set of practices that ensure resources across all FinOps Scopes are properly selected", then right-sized, run only when needed and kept busy. It comes before any discount: a commitment bought on waste locks the waste in.

1. Find what bills for nothing:
   • a stopped EC2 instance's EBS volumes and Elastic IP addresses;
   • a stopped RDS instance, which keeps billing storage and restarts after seven days;
   • S3 unfinished multipart uploads and old versions;
   • a NAT gateway's hours and every GB it processes.
2. Opt in to AWS Compute Optimizer: it "analyzes your AWS resources' configuration and utilization metrics to provide you with rightsizing recommendations and identify idle resources", but "You must opt in", and it reads metrics "for the last 14 days" (Compute Optimizer). Read its findings two weeks later.
3. Stop one development resource outside working hours: environments "typically only used for eight hours a day during the work week" allow "a potential cost savings of 75% (40 hours versus 168 hours)" (Cost Optimization Pillar). Count what stays billed while it is stopped.
4. Check whether an Arm instance would do the same work for less: AWS Graviton compares prices family by family.

What the two levers do to one general-purpose server in Paris, if Compute Optimizer finds it four times too big and the code runs on Arm:

Each size down halves the price, and the m7g.large costs 19.0% less than the m7i.large, computed. Rightsizing first matters: the discount of exercise 6 would otherwise apply to the 2xlarge.

Go deeper: the "How to reduce" section of each service article, such as Amazon EC2. Book: chapter 15. Series: How to learn FinOps – Rightsize and clean up.

6. Commit to what runs (Rate Optimization)

Rate Optimization "is the Capability used to manage the rates that you pay for the resources you use in each technology category". Savings Plans trade "a commitment of using a specified amount of compute power (measured per hour) for a one or three year period" for lower rates: "You can save up to 72% on your AWS compute workloads" (Savings Plans).

1. In Cost Explorer, open the Savings Plans recommendation for one year, no upfront. Do not buy it.
2. Compare its hourly commitment with your lowest hourly compute spend last month: above that floor, you pay for hours you do not use.
3. If you commit, add a utilization budget and a coverage budget, two types AWS Budgets offers (AWS Budgets).

What the commitment buys, on the m7g.large that exercise 5 ended with: each rung trades flexibility for rate.

The Compute Savings Plan follows the workload to another family, Region, Fargate or Lambda; the EC2 Instance Savings Plan stays in one family in one Region (Savings Plans types). Commit only to the hours that run every hour: a committed hour that nothing uses is paid all the same.

Go deeper: Savings Plans vs Reserved Instances, Savings Plans and Reserved Instances for EC2, reserved DB instances and Database Savings Plans, Compute Savings Plans for Lambda. Book: chapters 16 to 18. Series: How to learn FinOps – Commit to what runs.

In short: Use less before you pay less: clean up and rightsize first, then commit only to the usage that runs every hour.

7. Price before you build (Planning & Estimating, Architecting & Workload Placement)

The cheapest change to a bill is made before the resource exists, when "Business value drives technology decisions" (principles) costs a sentence in a document rather than a migration.

1. Write two designs for a feature you are about to build, for example a container that runs all the time and a function that runs per request.
2. Price both in the calculator, which turns an architecture described in words into its monthly cost on AWS, Google Cloud and Azure.
3. Keep the cheaper one only if it meets the same need; write down what the other would have bought.

Example: serverless or servers, at three levels of traffic

Take the service of pattern 3 in the AWS Lambda article: each request is 100 ms of CPU-bound work. Serverless, it runs on Lambda on Arm at 1,769 MB, the memory of one vCPU, billed per request and per GB-second. On servers, it runs on Fargate Arm tasks of 1 vCPU and 2 GB, or on EC2 c7g.large instances, billed every hour whether busy or not, sized for 70% average CPU and never fewer than two, one per Availability Zone. The same design, priced in Paris at three levels of traffic:

The gap is not a fixed percentage: it swings with the traffic. The cheaper server option costs:

• Low and irregular: 27.7 times Lambda's bill, for tasks busy 1.9% of the time.
• Medium: 5.1% more than Lambda, for tasks busy 50.0% of the time.
• Sustained: 27.7% less than Lambda, for tasks busy 69.0% of the time.

The two minimum tasks bill the same $69.14 in the low and the medium case: only their busy share changes.

A · Two Fargate Arm tasks

Running all month, mostly waiting Illustrative

Line	A month
Compute	$69.14
Fargate Arm vCPU: 2 tasks × 1 vCPU × 730 h1,460 vCPU-h × $0.03888 /vCPU-h	$56.76
Fargate Arm memory: 2 tasks × 2 GB × 730 h2,920 GB-h × $0.00424 /GB-h	$12.38
A month, computed	$69.14

B · Lambda on Arm

1 million requests a month, in bursts during business hours Illustrative

Line	A month
Compute	$2.50
Lambda requests: 1 million1 M requests × $0.20 /M req	$0.20
Lambda duration, Arm: 1 million × 0.1 s × 1,769 MB172,753.91 GB-s × $0.0000133334 /GB-s	$2.30
A month, computed	$2.50

Lambda on Arm costs $66.64 a month less than two Fargate Arm tasks (96.4% less), computed.

USD before tax, 730 h a month. AWS Price List, eu-west-3, offer files published September 11, 2026 and October 14, 2026, fetched October 15, 2026. Illustrative configuration, not a client case.

The rule to keep: price the busy share, not the service. Below the tipping point, a server bills for waiting and Lambda wins; above it, Lambda's price per busy vCPU, 1.75 times a Fargate Arm task's, makes servers cheaper, all the more with a commitment from exercise 6. Near it, as in the medium case, the gap is small enough to decide on something else: the scaling, the load balancer and the patching that Lambda runs for you. The break-even table of the Lambda article gives the tipping point for each option and commitment.

Go deeper: the book on this site is about this habit, outside FinOps; its chapter 6 is the case told at the end of this guide. In Cloud FinOps: chapter 24. Series: How to learn FinOps – Price before you build.

8. Measure unit cost (Unit Economics, KPIs & Benchmarking)

A growing total says nothing alone: the business may have grown faster. Well-Architected asks you to "Measure the business output of the workload and the costs associated with delivery".

1. Pick a unit your business counts: a request, an active customer, an order, a simulation run.
2. Divide the allocated cost of the workload that serves it (exercise 3) by the units of the same month.
3. Track the ratio monthly, next to the total. A total that rises while the unit cost falls is growth, not waste.

Go deeper: AWS Lambda bills per request and per GB-second, the easiest service to compute a unit cost on. Book: chapter 22. Series: How to learn FinOps – Measure unit cost.

9. Run the practice (Manage the FinOps Practice)

The phases loop, "rapidly": this exercise turns the first eight into a routine.

1. Book a one-hour monthly review: total and unit cost, untagged share, anomalies, open recommendations.
2. Keep a decision log: what changed, who changed it, what it saved on the next bill.
3. Automate one thing you did by hand, such as stopping development resources at night.
4. Then take the exam that fits your role, from the certification table.

Go deeper: book chapters 20 and 21; for a team, the training offer covers the same ground in half a day to two days. Series: How to learn FinOps – Run the practice.

What the exercises look like at scale

The exercises are small on purpose; on large systems, a design change outweighs any discount.

• The case: at EDF Power Solutions, Mathieu Kerbel rewrote a simulation engine from C# to Rust. One run went from about 7 h 30 to under 30 seconds, and cloud costs fell by about half (the FinOps service page; chapter 6 of the book).
• In the framework's words: Architecting & Workload Placement, a change of design, not of rate. No commitment or tag produces that.
• Clients: from 2021 to 2023, more than 50 freelance clients, from embedded systems to high-performance computing (services). The FinOps offer applies these exercises to a company's own bill, starting with a free audit.

FAQ

How do I start learning FinOps?

With the free Introduction to FinOps course, about 30 minutes, then the framework pages, then the exercises in Applying each precept on your own cloud account.

Can I learn FinOps for free?

Yes, up to the certification: the introduction courses on FinOps and on FOCUS are free, and so are the framework pages, the providers' documentation and Cost Explorer's console. The certifications are paid: $325 for the Practitioner exam alone, $500 with its course, on November 3, 2026 (FinOps Foundation catalog).

Which FinOps certification should I take first?

The FinOps Certified Practitioner: it "Covers all fundamentals of FinOps" and has no formal prerequisites (course page). The others do not require it, except the Professional, which asks for the Practitioner or the Engineer, plus the FOCUS Analyst, AI Value and Technology Value certifications (Professional). Its exam and a study plan: FinOps Certified Practitioner.

What is the FinOps Certified Practitioner exam like?

Fifty multiple-choice questions in one hour, 75% to pass, not proctored, three attempts within 12 months of purchase; the certification is valid for 24 months (exam page).

Which book should I read to learn FinOps?

Cloud FinOps, second edition, by J.R. Storment and Mike Fuller of the FinOps Foundation, published by O'Reilly; its parts follow the Inform, Optimize and Operate phases (book page).

Do I need a cloud account to learn FinOps?

To pass the exam, no. To learn the practice, yes: every exercise in this guide reads or changes a real bill, and a personal account with a few resources is enough for most of them.