Reading the AWS Cost and Usage Report: the 10 columns that explain most of your bill

Your AWS bill is one table. Of its 125 possible CUR 2.0 columns (per the AWS CUR 2.0 table dictionary), ten tell you who spent the money, on what, and why it costs what it does. Read them in order: account, product, usage type, operation, charge type, amount, cost, resource, description, tags. With those ten you can trace any dollar on the invoice to an account, a product, a usage type and a resource.

Why is my AWS bill so hard to explain?

The Cost and Usage Report is not a summary. AWS delivers it to an S3 bucket you own as ZIP, GZIP or Parquet files, with one row per line item at HOURLY, DAILY or MONTHLY granularity: usage, commitment fees, credits, refunds and tax all sit in the same table.

Which CUR you have changes what you see. In legacy CUR the set of columns can vary from month to month, and each activated tag gets its own column. CUR 2.0, created in the Billing and Cost Management console under Data Exports, has a fixed schema and nests resource_tags, cost_category, product and discount as maps.

CUR 2.0 also adds bill_payer_account_name, line_item_usage_account_name, line_item_iam_principal and line_item_user_identifier. A CUR 2.0 export is defined as a SQL query on the COST_AND_USAGE_REPORT table. In both versions, line_item_usage_start_date (legacy lineItem/UsageStartDate) dates each line and bill_billing_period_start_date (legacy bill/BillingPeriodStartDate) gives you the month.

How do I check which CUR columns explain my AWS bill?

Start with these ten columns: CUR 2.0 name first, legacy name second.

  1. line_item_usage_account_id / lineItem/UsageAccountId: The AWS account that used the line item. In an organization, this tells you whether the charge came from the management account or a member account.
  2. line_item_product_code / lineItem/ProductCode: The code of the product measured, like AmazonEC2, AmazonS3 or AmazonRDS. This tells you the service.
  3. line_item_usage_type / lineItem/UsageType: The specific usage pattern. For example, USW2-BoxUsage:m2.2xlarge means an M2 High Memory Double Extra Large instance in US West (Oregon). DataTransfer-Out-Bytes means data transfer out to the internet.
  4. line_item_operation / lineItem/Operation: The specific AWS operation, like RunInstances for EC2 or GetObject for S3.
  5. line_item_line_item_type / lineItem/LineItemType: The charge type. Crucial for accurate summing. Valid values: Usage, DiscountedUsage, RIFee, Fee, SavingsPlanCoveredUsage, SavingsPlanNegation, SavingsPlanRecurringFee, SavingsPlanUpfrontFee, Credit, Refund, Tax, Discount, BundledDiscount.
  6. line_item_usage_amount / lineItem/UsageAmount: The usage in the period, in the unit of the usage type: hours for EC2 instance usage, GB-months for S3 storage, request counts for API calls. Check pricing_unit when you're unsure. Some subscription charges show 0 here.
  7. line_item_unblended_cost / lineItem/UnblendedCost: UnblendedRate multiplied by UsageAmount: the cost at the rate that line was charged. It is not a pre-discount on-demand figure. Reserved Instance covered usage shows 0 here, and the commitment cost sits on its own lines.
  8. line_item_resource_id / lineItem/ResourceId: The resource ID, like i-1234567890abcdef0 for an EC2 instance or my-bucket for S3. It is populated only if the report includes resource IDs (INCLUDE_RESOURCES in a CUR 2.0 export, "Include resource IDs" in legacy CUR). It is blank for data transfer, API requests, discounts, credits and taxes.
  9. line_item_line_item_description / lineItem/LineItemDescription: A human-readable description. AWS says its wording can change, so never parse it for automation; treat it as display text.
  10. resource_tags / resourceTags/user:<key>: Your cost allocation tags. In CUR 2.0, tags are nested maps with keys like user_department. In legacy CUR, each tag is a separate column named resourceTags/user:department. Only tags you’ve activated under Cost allocation tags in the Billing and Cost Management console appear.

Run this Athena query this week. It totals one billing period by charge type and product. Replace the table name with your own CUR table:

SELECT
  line_item_line_item_type,
  line_item_product_code,
  SUM(line_item_unblended_cost) AS unblended_cost
FROM your_cur_database.your_cur_table
WHERE bill_billing_period_start_date = TIMESTAMP '2026-10-01 00:00:00'
GROUP BY line_item_line_item_type, line_item_product_code
ORDER BY unblended_cost DESC;

Athena bills by data scanned. If your table is partitioned by billing period, add that partition filter too, so the query reads one month instead of your whole history. Then group the same month by usage type:

SELECT
  line_item_usage_type,
  SUM(line_item_unblended_cost) AS unblended_cost
FROM your_cur_database.your_cur_table
WHERE bill_billing_period_start_date = TIMESTAMP '2026-10-01 00:00:00'
GROUP BY line_item_usage_type
ORDER BY unblended_cost DESC;

Compare that second result with the checker. The checker reads legacy CUR or CUR 2.0 as .csv or .csv.gz (not .zip) and matches columns by header name. It needs only usage type and unblended cost, and it takes several part files at once, so you can drop in the whole month.

Check your own bill in 2 minutes: https://revelations.consulting/checker?utm_source=blog&utm_medium=organic&utm_content=blog-aws-cost-and-usage-report-columns

How do I total CUR costs without miscounting Reserved Instances and Savings Plans?

Don’t sum line_item_unblended_cost without filtering by line_item_line_item_type.

EC2 and RDS usage covered by a Reserved Instance shows up as DiscountedUsage with an unblended rate of 0, so its line_item_unblended_cost is 0. The cost of that reservation sits on separate RIFee lines. Group by product alone and you see cheap compute; add line_item_line_item_type and you see where the commitment cost landed.

Savings Plans work differently. SavingsPlanCoveredUsage lines carry the covered usage at its on-demand cost, and matching SavingsPlanNegation lines offset them, so the pair nets to zero. What you pay for the plan appears on SavingsPlanRecurringFee and SavingsPlanUpfrontFee lines.

line_item_net_unblended_cost (legacy lineItem/NetUnblendedCost) is the after-discount cost, and it is present only when the account has a discount that period. A query or dashboard that depends on it breaks or shows blanks in a month without a discount. Build your totals on line_item_unblended_cost and treat net cost as an extra view when the column exists.

What it usually looks like

Look for these patterns in your own CUR export:

  • EC2 or RDS usage with line_item_unblended_cost of 0 and line_item_line_item_type of DiscountedUsage: Reserved Instances cover it, and the cost sits on RIFee lines. If the RIFee lines are large but few lines are DiscountedUsage, check whether your reservations still match what you run.
  • line_item_usage_type containing DataTransfer-Out-Bytes with a blank line_item_resource_id: that is data transfer out to the internet, and AWS does not put a resource ID on data transfer lines. Group by line_item_usage_account_id and line_item_usage_type to see which account is sending it.
  • One line_item_product_code covering several kinds of spend: AmazonEC2 includes instance hours (BoxUsage), EBS volumes (EBS:VolumeUsage) and NAT gateways (NatGateway-Hours, NatGateway-Bytes). Group by line_item_usage_type inside each product code before you decide what to cut.
  • line_item_line_item_type of Tax: these lines carry no resource ID, so a tag-based or team-by-team breakdown will not include them. Total them separately so your breakdown still reconciles to the invoice.

If you'd rather walk through the output with someone, book a free 30-minute consult: we'll look at your bill with you, no obligation and no sales pitch.

Unzip any .zip parts first (the checker reads .csv or .csv.gz, not .zip), drop in every part file for the month, and see which usage types carry your unblended cost.

Run your own numbers: AWS bill checker

Frequently asked questions

What is the difference between line_item_unblended_cost and line_item_net_unblended_cost?

line_item_unblended_cost is UnblendedRate multiplied by UsageAmount: the cost at the rate each line item was charged, which is why Reserved Instance covered usage shows 0 there. line_item_net_unblended_cost is the after-discount cost, and that column is present only when the account has a discount in the billing period.

Why do some EC2 line items show zero unblended cost in the CUR?

EC2 usage covered by a Reserved Instance appears in the Cost and Usage Report as DiscountedUsage with an unblended rate of 0, so its line_item_unblended_cost is 0 and the reservation's cost sits on RIFee lines. Savings Plan covered EC2 usage is different: SavingsPlanCoveredUsage is offset by SavingsPlanNegation, and the plan's cost appears on SavingsPlanRecurringFee and SavingsPlanUpfrontFee lines. Group by line_item_line_item_type before you total anything.

Do I need to enable resource IDs in my Cost and Usage Report?

Turn on resource IDs in the Cost and Usage Report if you want to see which EC2 instance, S3 bucket or RDS database generated a charge. Without them (INCLUDE_RESOURCES in a CUR 2.0 export, "Include resource IDs" in legacy CUR), line_item_resource_id is blank. Expect a larger report with resource IDs on, which means more S3 storage and more data scanned per Athena query.

Why does my data transfer cost not have a resource ID?

AWS leaves line_item_resource_id blank in the Cost and Usage Report for data transfer, as it does for API requests, discounts, credits and taxes. To find data transfer out to the internet, group by line_item_usage_account_id and line_item_usage_type and look for usage types containing DataTransfer-Out-Bytes.

How do I know which tags are available in the CUR?

Only activated cost allocation tags appear in the Cost and Usage Report: in CUR 2.0 as user_<key> keys in resource_tags, in legacy CUR as one resourceTags/user:<key> column per tag. Check which keys are active in the Billing and Cost Management console under Cost allocation tags. A tag must be applied to the resource and activated there before it shows up.

Run the numbers on your own bill: Cloud Bill Gotchas Checker. Free, in your browser.

Want this kind of thing weekly? Get Cloud Bill Weekly. Want us to find it in your bill? We're taking three founding clients for a free Cost Assessment.