Common Bill Format (CBF)

CloudZero has developed a standard data model to ingest cost data from any source, called the Common Bill Format. The CBF is modeled after the AWS CUR format, but simplier and more generic. This allows it to be used by many different cloud providers.

Data File General Guidelines

  • Provided as drop of gzipped CSVs and JSON manifest file (Common Bill Format (CBF) File Drop Specs)
  • Max 1,000,000 rows per file
  • datetime values should be ISO formatted strings in the UTC timezone.
    • Note: Excel will often auto-convert ISO formatted dates to a human-readable format. Ensure dates have saved correctly in your CSV file.

Data File Columns

CategoryColumnTypeCZ Dimension or Cost TypeDescriptionRequired
Line Itemlineitem/idstringUniquely identifies this specific line item in this specific bill.
Line Itemlineitem/typestringLineItemTypeBroad category of this line item: Usage, Tax, Discount, etc. Values have special meaning (see below). If not provided, assumed to be Usage.
Line Itemlineitem/descriptionstringDescriptionAn optional description specifying additional information about this line item.
Timetime/usage_startdatetimeThe hour during which the charged usage applies. This value should be aligned to the start of the hour and will be treated as such. Note, the current version of the Common Billing Format does not currently use usage_end date.
Timetime/usage_enddatetimeThe end of a timespan to which the charged usage applies. Note that although this may be specified, currently all changes are treated as occurring in a single hour at usage_start. Future updates to the ingest process may take advantage of the usage_end value.
Resourceresource/idstringResourceUniquely identifies the object for which this charge applies. This column is required if any resource/tag:<key> columns are included. Read more about Resource IDs below.✅*
Resourceresource/servicestringServiceThe category to which this resource belongs. Generally represents different kinds of services provided by the Cloud Provider for which the customer is charged.
Resourceresource/accountstringAccount (if action/account is not specified)The account or project to which this resource belongs (if applicable).
Resourceresource/regionstringRegion (if action/region is not specified)The region to which this resource belongs (if applicable).
Resourceresource/usage_familystringUsageFamilyCommonly a subdivision of resource/service
Resourceresource/tag:<key>stringTag:<key>Custom attributes associated with the resource. The <key> is case sensitive, and can consist of letters, numbers, and the following special characters _ . : / = + - @.. Using this column also requires resource/id to be populated. Read more about tags below.
Actionaction/operationstringOperationThe thing that was done to the resource for which you're being charged.
Actionaction/usage_typestringUsageTypeCommonly a subdivision of action/operation
Actionaction/regionstringRegionIn what region the operation was performed. (May not match the resource/region for cross-region operations.)
Actionaction/accountstringAccountIn what account the operation was performed. (May not match the resource/account for cross-region operations.)
Usageusage/amountnumberA numeric value describing an amount consumed/used. E.g. GB stored/transferred, seconds executed, credits consumed
Usageusage/unitsstringA description of the units used for usage/amount
Costcost/costnumberBilled CostThe cost associated with this line item. (May be negative for line items which represent discounts / credits.) Sometimes referred to as unblended cost, billed cost, or invoiced cost.
Costcost/discounted_costnumberDiscounted CostThe net cost associated with this line item after any discounts, credits, or private pricing are applied. Sometimes referred to as net cost or net unblended cost.
Costcost/amortized_costnumberAmortized CostThe effective cost associated with this line item after any "committed use" purchases (e.g. AWS RIs or Savings Plans, GCP CUDs, etc.) are amortized over all the "Usage" line items to which they apply.
Costcost/discounted_amortized_costnumberDiscounted Amortized CostThe net effective cost associated with this line item after any discounts, credits, or private pricing are applied. It also includes any "committed use" purchases (e.g. AWS RIs or Savings Plans, GCP CUDs, etc.) amortized over all the "Usage" line items to which they apply. Sometimes referred to as net amortized cost, net effective cost, or fully-loaded cost.
Costcost/on_demand_costnumberOn-demand CostWhat the cost of this line item would have been if no mechanism for reducing costs applied. In other words, it is the cost as if there were no discounts, applicable "committed use" purchases, private pricing, etc. This is often useful for determining one's effective savings rate.
Billbill/invoice_idstringInvoiceIdUniquely identifies a particular bill. A single billing data ID for a single month may include multiple invoices. Ideally this field will not be populated until an Invoice is closed.
Kubernetesk8s/clusterstringKubernetes ClusterThe Kubernetes cluster associated with the resource.
Kubernetesk8s/namespacestringKubernetes NamespaceThe Kubernetes namespace associated with the resource.
Kubernetesk8s/deploymentstringKubernetes DeploymentThe Kubernetes deployment associated with the resource.
Kubernetesk8s/labelsstringKubernetes LabelsThe Kubernetes labels associated with the resource.

*resource/id is only required when a resource/tag:<key> value is included for the charge.

Resource IDs

  • The resource/id is the unique identification of the resource for which the charge applies.
  • The following special characters : / have special meaning and should be used carefully
    • A resource/id containing one of the above characters will be split and the second element will be taken as the resource/id (e.g. a resource/id of a:b:c will become b)
  • When resource/id is specified, you may also provide resource/tag:<key> columns to categorize the resource by <key>, which may be a purpose, owner, environment, or other criteria.
    • When including resource/tag:<key> data with a charge, the resource/id column is required.
    • Only one resource/tag:<key> column <value> will be associated with each unique resource/id value. For more information, see the Resource Tags section below.

Resource Tags

  • resource/tag:<key> columns allow you to categorize resources by purpose, owner, environment, or any other criteria.
  • The <key> is case sensitive, and can consist of letters, numbers, and the following special characters _ . : / = + - @..
    • No spaces, tabs or returns are allowed.
    • For example, resource/tag:team or resource/tag:Product-Type are valid tag keys, while resource/tag:Team Name and resource/tag:Product&Dev are not.
  • resource/tag:<key> will appear as Tag dimensions in the CloudZero platform, and can be used to filter or aggregate the cost of the tagged resources.
  • You may include as many unique resource/tag:<key> columns as you need.
  • Only one <value> will be associated with a resource/tag:<key> column for each unique resource/id.
    • If multiple values are supplied for the same ID, the value with the latest time/usage_start will be used. If all instances for a given key have identical time/usage_start dates, then one is arbitrarily selected.
  • The latest <value> of a given resource/tag:<key> for each resource/id will be retroactively applied to all data in the CloudZero system with the same resource/id.

Resource ID & Tag Usage Example

B Large, Inc. is using the CloudZero platform to monitor their cloud costs, and needs to easily identify their resources by the teams that own them. To assist with this, B Large creates a Resource Tag called resource/tag:team, and populates it with the names of the various teams that own the AWS resources in question.

Team Alpha owns an AWS Lambda function with the Resource ID of making-great-calculations. Since these resources have been tagged as owned by Team Alpha using the resource/tag:team key with the value of Alpha, whenever anyone logs into CloudZero, they can easily use the team Tag dimension to view what Alpha's Lambda is costing B Large, and monitor trends over time.

Recently, B Large promoted more personalized team names, and Team Alpha voted to name themselves Team Batman. With this change, the resource/tag:team value in AWS is updated to reflect their new name, and CloudZero receives a new data drop that associates a new resource/tag:team value of Batman for the AWS Lambda making-great-calculations.

The CloudZero platform receives this updated tag information, and because only one value per resource/tag:team can be associated with a single resource/id of making-great-calculations, the platform retroactively applies this new tag to all historical data in the system for the resource/id making-great-calculations.

Now, when anyone at B Large logs into CloudZero, they can still see what Team Batman owns, and because the tag change was retroactively applied to all historical data, the monitoring of trends over time for the making-great-calculations Lambda sees no interruption, and reports continue to accurately reflect totals and trends, including those that were ingested prior to Team Batman's name change.

Line Item Types

The lineitem/type column defines how each line items should be interpreted.

The following values are supported:

Line Item TypeDescription
UsageThe most common type and the default. The line item represents a charge for the use of some cloud resource. “Real Cost” includes only Usage line items and uses the first available cost type provided from: amortized_cost, discounted_cost, or cost
TaxThis line item represents any tax charges. Columns in the time, resource, and action categories should only be populated if the tax is associated with applicable Usage changes. Otherwise these columns should be left blank.
SupportCharges for support or other human services. Columns in the time category should specify the start and end of the support contract.
PurchaseCharge for a one-time purchase or non-usage based subscription; for example, software bought from the AWS Marketplace. The time category columns should represent the span of time over which the purchase applies (for subscriptions with renewal) or the time of the purchase (for one-time charges).
CommittedUsePurchaseCharges for a "committed use" (RI, savings plan, etc.) purchase. "Committed use" includes any instrument for which payment is made to receive a reduced rate on future usage. This may include one-time upfront purchases or recurring monthly charges. The time category columns should represent the span of time over which the charge applies (for example, one month for monthly recurring charges.) The amortized_cost cost column of this line item should always be zero (if used).
DiscountA negative value cost associated with some Usage line item. Columns in the time, resource, and action categories of this line item should match those of the applicable Usage line item and the cost should be negative. The discounted_cost for this line item should be zero if this discount is fully accounted for in the discounted_cost of other Usage line items. (Please note: you must still include discounted_cost if it should be zero)
CreditA negative value cost not associated with any specific Usage line items. May represent an adjustment due to an error, rounding, refund, etc.
FeeA positive value charge for which no other line item type applies.
AdjustmentAn alteration made to the bill to correct for some error or rounding issue.

Cost Category Columns

There are five cost category columns cost, discounted_cost, amortized_cost, discounted_amortized_cost, and on_demand_cost. Only cost is required, but the others may be used to help “smooth out” charges not directly associated with the use of cloud services.

To ensure all cost columns are fully populated and may be used in the platform, blank values will "fall back" and use a different cost column. For example, if discounted_amortized_cost is left blank, discounted_cost will be used instead. The assumption is that no line items required amortization (i.e. no committed use purchases) and so the discounted cost could be used instead. Only blank or empty values trigger this behavior, if the value is set to 0 it will be used as is. The complete set of fallback logic is:

  • discounted_amortized_costamortized_costdiscounted_costcost
  • amortized_costcost
  • discounted_costcost
  • on_demand_costcost

The total value across all line items for each of cost, discounted_cost, amortized_cost, and discounted_amortized_cost should be the same. Each represents a redistribution of charges (discounts and/or purchases) not the inclusion or exclusion of charges. This equivalence is not currently enforced by the billing ingest system, but may be in the future.

Cost Types

CloudZero cost types are defined in terms of the lineitem/type and different cost category columns.

  • Real Cost - Only Usage line item types are included. The discounted_amortized_cost column is used, if present, with “fallback” logic to the other cost categories if it is not.
  • Discounted Amortized Cost - All line item types are included. The discounted_amortized_cost column is used, if present, with “fallback” logic to the other cost categories if it is not.
  • Amortized Cost [Coming Soon] - All line item types are included. The amortized_cost column is used, if present, with “fallback” to cost if not.
  • Discounted Cost - All line item types are included. The discounted_cost column is used, if present, with “fallback” to cost if not.
  • On-Demand Cost - All line item types are included. The on_demand_cost column is used, if present, with “fallback” to cost if not.
  • Billed Cost - All line item types are included. The cost column is used.



In this example we purchased some Compute from “Simple Cloud”. We also made an up-front “committed use” which lowered the normal hourly rate by 50%. We received the “SpecialCompute” discount for using certain Compute instance type (Simple Cloud is running a promotion.) We also received the “MVP Discount” because we are so cool.

This first version of the bill just includes the actual charges. The cost for our Compute instances is at the reduced rate because of the “committed use” purchase. (For example, instance-0000 would have otherwise cost $24 during that hour.) The sum of the values in the cost/cost column should equal exactly what I see on my bill for March: $105.30


Discounted Cost

It would be helpful if the discount we received for using the promoted Compute instances was actually represented in the cost of those instances. In this example we’ll include the cost/discounted_cost


The SpecialCompute discount was distributed evenly over the individual instances. We also zeroed out the cost for the SpecialCompute line item because that discount has already been accounted for. We chose not to distribute the MVPDiscount since it was unrelated to the Compute instances. We copied over the values for the CommitedUse purchase and MVPDiscount, but we could have also left them blank and the cost/cost value would have been used by default.

Amortized Cost

To help our engineers better understand the real cost of those Compute instances we also want to include what we payed for the CommitedUse purchase.


The cost of the CommitedUse purchase was evenly distributed over all the Compute instances to which it applied (which in this case was all of them.) We left the cost/amortized_cost for the Discounts blank. We could have also copied over the values from the discounted_cost (0 and -20). The important this is the the sum of the values in the cost/amortized_cost column (including the default values for the blank cells) is still: $105.30

Kubernetes-Related Columns

In this example, we're adding Kubernetes-related columns to the existing CBF data to provide more context on resource usage within a Kubernetes environment.

UsageComputeinstance-00002022-03-16T13:00:00Zmy-clusterdefaultweb-app{"app": "frontend", "env": "production"}12
UsageComputeinstance-00012022-03-16T13:00:00Zmy-clusterdefaultweb-app{"app": "frontend", "env": "production"}20
UsageComputeinstance-00022022-03-16T13:00:00Zmy-clusterdefaultworker-app{"app": "backend", "env": "production"}15.3

In this example, we're including Kubernetes-related information such as cluster, namespace, deployment, and labels. This allows us to track usage and costs within specific Kubernetes contexts.