Enable AWS Data Exports (CUR 2.0): A Practical Setup Guide

I've seen teams evaluate FinOps tools for months before enabling the billing export those tools eventually depend on.

This is the first guide in the AWS Cost Foundations series.

Why now? Turn this on today. Future-you will be glad the data exists.

Prerequisites

Billing permissions on the management account (or delegated billing admin access)
An S3 bucket to receive the export (you can create one during setup)

If you use AWS Organizations, start from the management/payer account unless delegated billing administration is already configured.

Nothing else. No other services, tools, or infrastructure required.

Step 1: Open Data Exports

AWS Data Exports landing page showing an empty exports table with a blue Create button — Figure 1. AWS Data Exports landing page

Navigate to Billing & Cost Management → Data Exports
Click Create
If you see "Cost and Usage Reports" instead, that is the legacy page being deprecated. Look for "Data Exports" under Cost and Usage Analysis.

Step 2: Select Standard Data Export

Create export page with Standard data export selected and CUR 2.0 chosen — Figure 2. Standard Data Export with CUR 2.0 selected

Select Standard data export (not QuickSight, not Legacy CUR)
Under "Data table content settings", select CUR 2.0
Give your export a name: MyCostExport or similar. No spaces.

Why CUR 2.0? Native AWS format. What most tools expect. What most documentation references. FOCUS is a valid multi-cloud alternative covered in Part 2 of this series.

Step 3: Configure additional content

Additional export content options with Include resource IDs and Split cost allocation data checked — Figure 3. Additional export content settings

Check Include resource IDs - so you can trace costs to individual resources later
Check Split cost allocation data - breaks out shared ECS/EKS costs by container
Check Refresh automatically - keeps data accurate when AWS applies retroactive credits or refunds

Important: If you enable "Split cost allocation data", you must also opt in under Cost Management Preferences. AWS shows a blue info banner reminding you. Skip this and the column will be empty.

Step 4: Configure time granularity

Time granularity section with Daily selected and all 131 columns included — Figure 4. Daily granularity with all columns included

Select Daily granularity (one row per resource per day)
Leave all 131 columns selected (the default). You cannot retroactively add columns to historical exports.
Hourly produces far more data and is rarely needed at this stage. Monthly is too coarse.

Step 5: Configure delivery options

Delivery options with Overwrite existing file selected and Parquet compression format — Figure 5. Delivery options using Parquet format

File versioning: Overwrite existing data export file (saves S3 storage costs)
Report data integration: Leave all unchecked (no Athena, no Redshift)
Compression and format: Parquet (columnar, compressed, what every modern tool expects)

We are collecting data, not querying it. Integrations can be enabled later without recreating the export.

Step 6: Configure S3 destination

S3 storage settings showing bucket name field, path prefix field, and Configure button — Figure 6. S3 destination settings for the export

Storage account: This account
S3 bucket: your bucket name (e.g. mycompany-billing)
S3 path prefix: a prefix like cur/

If the bucket does not have the correct policy, click Configure. AWS generates and applies the required bucket policy automatically. It grants only the billing service write access.

Click Create.

Verification

Expected timing: AWS delivers the first export within 24 hours. Do not expect immediate data.

What success looks like: Parquet files in your S3 bucket, organized by billing period:

s3://mycompany-billing/cur/MyCostExport/...

After 24 hours, check your bucket. Files present means the export is working. Your billing history is now being preserved.

Common mistakes

No billing permissions: The IAM user or role must have Billing console access. In an AWS Organization, this often requires a policy on the management account, not member accounts.
Wrong bucket permissions: The S3 bucket needs a policy allowing billingreports.amazonaws.com to write objects. Use the "Configure" button during setup to apply this automatically.
Waiting less than 24 hours: The first delivery takes up to 24 hours. An empty bucket immediately after creation is normal, not a failure.
Confusing CUR 2.0 with FOCUS: Different formats, different column schemas. CUR 2.0 is native AWS. FOCUS is a multi-cloud FinOps standard. Both are valid. This guide uses CUR 2.0. Part 2 covers FOCUS.

What's next

Most teams delay this because it sounds bigger than it is.

Next in this series: CUR 2.0 vs FOCUS - which format to choose, when FOCUS matters, and whether you should run both.

Back to Guides