On this page

Google Cloud gives every billing account a FinOps score from 1 to 5. It tells you how well you use Google’s own tools to see, control and reduce your cloud bill. This guide explains what the score looks at and the habits that raise it, with a 30-day plan you can follow.

In short

  • The score rewards habits, not low spend: watching costs, labelling, budgets, the BigQuery export, removing idle resources and committing where usage is steady.
  • Start with visibility. Labels and the billing export make every later step easier.
  • Google does not publish the formula. Treat the score as a checklist, not a target to game.

What the FinOps score is

You find the score in the Google Cloud console under Billing → FinOps hub. According to Google’s documentation, it reflects how often your team does these things:

  • uses the Cloud Billing tools to monitor spend,
  • uses labels and tags to allocate costs,
  • turns off idle resources and right-sizes machines,
  • buys committed use discounts, including the ones Google recommends,
  • creates and checks budgets,
  • automates with the billing export to BigQuery and the Budget API.

The score updates daily, using data from two days earlier. Next to it you see a peer benchmark: the average for companies in your industry. The Improve your score link opens recommended actions in three stages: inform, operate and optimize. These match the FinOps Foundation phases.

Two practical notes. You need full billing-account access (Billing Account Viewer or Administrator) to see the score. And Google does not publish how each habit is weighted, so nobody can promise exactly how much one change will add.

Our score: 4.3 out of 5

FinOps score of 4.3 out of 5 next to a peer benchmark of 2.7

On 1 November 2026, our own Google Cloud billing account showed a FinOps score of 4.3 out of 5 (maturity: High). The peer benchmark for our industry was 2.7 (maturity: Medium).

The score depends on habits, not on how much you spend, so a small account can score as high as a large one. The rest of this guide covers those habits, step by step.

See your spend

Labelled cloud resources grouped into cost by team

The first stage is simply looking. Open Billing → Reports every week and group costs by project and by service. Regular use of the billing tools is part of what the score measures.

Then add labels. A label is a key and a value on a resource, such as team:web or env:prod. Labels appear in billing reports and in the BigQuery export, so you can see who or what each euro is for.

Label key
Example values
Answers
env
prod, staging, dev
What does testing cost?
team
web, data, marketing
Who owns this spend?
product
app, website, api
What does each product cost?

Keep the keys short, lowercase and the same everywhere. A resource can have up to 64 labels, but three or four used consistently are worth more than twenty used now and then.

Export billing data to BigQuery

Cloud Billing data exported to BigQuery and shown in a cost dashboard

The billing export copies your cost data into BigQuery every day. It is one of the habits the score names, and it lets you build your own reports and alerts.

  1. Create a separate project for FinOps, linked to the same billing account.
  2. Create a BigQuery dataset in a multi-region (US or EU) location. Only multi-region datasets get earlier data filled in, from the start of the previous month. The location cannot be changed later.
  3. In Billing → Billing export, turn on detailed usage cost and pricing, and choose that dataset.

Turn it on early: data from before you enable it is mostly not filled in later. Do not set table expiry on the dataset. If you use Looker Studio, connect it to this dataset for a cost dashboard your team will actually open.

Budgets, alerts and anomalies

Monthly budget chart with a forecast alert before spend passes 100 percent

Create a budget for the whole billing account and one for each important project. Add alerts on forecasted spend as well as actual spend, so you hear about a problem before the month ends.

A normal budget only sends alerts. It does not stop spending. For some services you can now create a spend cap budget, which pauses new usage of one service in one project for the rest of the month. Today this covers the Gemini API, Vertex AI, Cloud Run and Cloud Run functions. Set the cap below your real limit, because cost data can lag by more than a day.

Also open Billing → Anomalies. Google flags unusual spikes and shows which service, region and SKU caused them. Set the thresholds so alerts mean something, and send them to the people who can act.

To automate, connect budgets to Pub/Sub. You can then post alerts to Slack or trigger an action. Using the Budget API this way is another habit the score names.

Cut idle and oversized resources

The FinOps hub lists Google’s recommendations with their monthly savings. Start with the easy ones:

  • Idle VMs. Google flags machines with very low use, based on the last 1 to 14 days (idle VM recommendations). Stop them, or delete them after checking with the owner.
  • Unattached disks and unused IP addresses. They cost money even when nothing uses them.
  • Oversized machines. Apply right-sizing recommendations to change the machine type.
  • Dev and test at night. Schedule non-production machines to stop outside working hours.

Apply recommendations in the hub, or dismiss them with a reason. Leaving a long list untouched suggests nobody is looking.

Commit where usage is steady

Committed use discounts (CUDs) lower the price when you promise steady use for one or three years. There are two kinds:

Type
You commit to
Good for
Resource-based
vCPUs, memory and similar, in one region
Compute Engine machines that run all the time
Spend-based
A minimum spend per hour or per month
Changing workloads across Compute Engine, GKE and Cloud Run, plus services such as BigQuery and Cloud SQL

Only commit to the base you use every hour of every day. Look at at least 30 days of usage first. The hub shows your CUD optimization rate: the share of eligible spend that is already covered. Google’s own CUD recommendations are a safe starting point. Buying them is also one of the habits the score looks at.

A 30-day plan

When
Do this
Day 1
Open the FinOps hub and note your score and benchmark. Turn on the BigQuery billing export.
Week 1
Agree on three or four label keys and label every project's main resources.
Week 2
Create budgets with forecast alerts. Set anomaly thresholds. Connect alerts to Slack or email.
Week 3
Work through the recommendations list: idle machines, disks, IPs, right-sizing.
Week 4
Review 30 days of usage and buy commitments for the steady base. Book a monthly 30-minute cost review.

GOOGLE CLOUD FINOPS

Want a higher score and a smaller bill?

We can set up labels, the BigQuery billing export, budgets and a monthly cost review for your Google Cloud account.

Review my cloud costs

Fresh from the blog

Short, practical reads on design, AI, and growth from Growth Geeks, an AI growth agency in Berlin.

Leave a Reply