Cloud cost and FinOps glossary

Definitions

Cost report

A saved or ad-hoc cost analysis: one time window, one grouping, one filter, and the settings the figures are computed with. Everything else in the product builds on it. See also: Cost reports and dashboards

Dashboard

A collection of saved cost reports for a recurring reader or workflow, such as an engineering team's weekly review or a monthly finance check. See also: Cost reports and dashboards

Report subscription

A recurring email delivery of one saved cost report. Monthly is a delivery frequency, not a separate kind of report: the email links to the same report the product shows, built from the same data and rules. See also: Cost reports and dashboards

Segment

The owner an allocation rule claims cost for: a team, a product, a cost centre, or the shared-cost bucket. See also: Cost allocation

Cost allocation

Assigning spend to teams, products or cost centres through segments and ordered rules, so each part of the bill has an owner. See also: Cost allocation

Allocation coverage

The share of spend that can be attributed to a team, product or cost centre. Higher coverage makes budgets, reports and ownership discussions more useful; unallocated spend should remain visible rather than being assigned by guesswork. See also: Cost allocation

Shared cost

Spend that serves several teams or products and cannot be attributed to one - a Kubernetes cluster, a shared database, an observability bill. Splitting it is a business decision rather than a technical one, so the method should be written down and agreed rather than inferred by a tool. See also: Cost allocation

Showback and chargeback

Showback reports allocated cost to the responsible team. Chargeback posts that cost to an internal budget or ledger. Costfluent provides showback: cost reports grouped by segment show each owner its allocated spend. See also: Cost allocation

Cost insight

A detected change in spend, such as an anomaly or a growing service. An insight shows where cost moved and which costs contributed. See also: Cost insights

Savings opportunity

A proposed change with an estimated value. Open, accepted, implemented and rejected are distinct states, and an estimate is never reported as a saving. Verification against actual cost is planned. See also: Savings

Realised savings

The reduction observed on the bill after a change was implemented, as opposed to the reduction estimated beforehand. Estimated, implemented, and verified savings are different states and should be reported separately. See also: Savings

Effective savings rate

The saving from commitment discounts, and depending on the basis negotiated discounts, as a share of what the same usage would have cost without them. FOCUS divides by contracted cost, which counts commitment discounts only; Microsoft's FinOps toolkit divides by list cost, which also counts negotiated discounts. The rate is meaningful only when its scope, period, and price basis are stated.

Effective cost

The FOCUS measure of what a charge actually costs once negotiated discounts apply and upfront commitment fees are spread over the period they cover. Billed cost is what the invoice shows when the charge is made. See also: Cost reports and dashboards

Amortized cost

A cost report setting that sums effective cost instead of billed cost, so a commitment paid upfront appears across the months it covers rather than as a spike in the month it was bought. See also: Cost reports and dashboards

FOCUS

An open specification of the Linux Foundation, supported by the FinOps Foundation, for normalizing billing data across providers. Costfluent normalizes AWS, Azure and Google Cloud billing data into FOCUS cost lines, so one report can compare them. See also: Integrations

File import

Billing data uploaded in a named provider format or as custom CSV. A file import is not a native connector: nothing is read from the provider's account. See also: Billing-file imports

Unit economics

Technology cost expressed per unit of business activity: per customer, per order, per document, per API request. Converts an infrastructure number into one that a pricing or product decision can be made against.

Technical gross margin

Revenue minus the technology cost of delivering it. Tells you whether growth is improving the business or just enlarging it, which a rising revenue line on its own does not.

FinOps

The practice of managing cloud cost as a shared responsibility of engineering, finance and management, with the FinOps Foundation's framework as its common vocabulary: inform, optimize and operate, repeated every month rather than finished once.

Tag

A key and value attached to a cloud resource, account or subscription that the billing data carries into every cost row, so cost can be grouped by owner, environment or product. AWS and Azure call them tags, Google Cloud labels; untagged cost is the share no key reaches. See also: Cost allocation

Budget

A planned amount for a scope and a period, compared with actual cost as the period runs. A budget alert fires when actual or projected cost crosses a threshold; an alerts-only budget changes nothing in the cloud, while AWS budget actions and Google Cloud spend cap budgets can restrict or pause usage. See also: Budgets

Forecast

An estimate of future cost from the usage to date, with a range that widens the further it reaches. A forecast is compared with the budget to see a gap early; it is not a commitment and is recomputed as data arrives.

Anomaly

A cost movement outside the range a service's own history makes likely: a jump, a drop or a new charge with no counterpart in the previous period. Detection names the service, the account and the day; whether it was intended is the owner's call. See also: Cost insights

Reserved Instances

A one- or three-year commitment to a specific instance type in a region, at AWS and Azure also to a scope, in exchange for an hourly rate below on-demand. The reservation bills whether the instance runs or not, which is why coverage and utilisation are tracked separately. See also: Savings

Savings Plans

An AWS or Azure commitment to a fixed hourly spend on compute for one or three years, applied automatically to eligible usage at a discounted rate. Usage above the commitment bills on demand, and the commitment bills even when usage falls below it. See also: Savings

Rightsizing

Matching a resource's size to its measured use: a smaller instance, fewer nodes or a lower storage tier, once the utilisation history shows headroom that was never used. It changes the usage, not the rate, and is verified against the next bill. See also: Savings

Data transfer

The charge for bytes that cross a boundary the provider meters: out to the internet (egress), between regions, between availability zones or through a gateway. Inbound traffic is mostly free; the direction and the path decide the price. See also: Cost reports and dashboards

Token

The unit in which language models meter text: a fragment of a word, a few characters long, counted separately for input, output and cached context. A price list quotes a rate per million tokens, so cost follows the length of prompts and answers, not the number of requests.

European sovereign cloud

A provider's offering that keeps a workload's data, operations and support inside Europe and under a European legal entity, such as the AWS European Sovereign Cloud or Microsoft Cloud for Sovereignty. It is a product with its own regions, price list and limits, not a legal status of the customer. See also: Digital sovereignty

FinOps principles

The six statements the FinOps Foundation's framework rests on: teams collaborate, business value drives technology decisions, everyone takes ownership for their usage, FinOps data is accessible, timely and accurate, the practice is enabled centrally, and the cloud's variable cost model is used deliberately.

FinOps phases

The three recurring stages of the FinOps Foundation's framework: Inform makes cost visible, allocated, budgeted and forecast; Optimize identifies savings opportunities and commitments; Operate implements the chosen options and matures the practice. A scope passes through them continuously rather than once. See also: Cost reports and dashboards

FinOps persona

A stakeholder perspective the FinOps Foundation's framework names, such as engineering, finance, product, procurement, leadership or the FinOps practitioner. A persona describes what someone needs from cost data and which decision they own, not a position; one person can hold several.

FinOps capability

A functional area of activity the FinOps Foundation's framework defines, grouped into four domains; examples are allocation, reporting and analytics, budgeting and anomaly management. Each capability is practised at its own maturity and is carried by a concrete artefact such as a report or a rule.

FinOps maturity

The level at which a team practises one FinOps capability, which the FinOps Foundation's framework names crawl, walk and run. Maturity is judged per capability, not per company, and the right level is the one whose effort the business value justifies.

FinOps scope

A segment of technology spend to which the FinOps practice is applied, as the FinOps Foundation's framework defines it. The framework groups scopes by technology category, such as public cloud, SaaS, data cloud platforms, data centre and AI; each category brings its own billing data and allocation keys. See also: Integrations

Cloud cost management

The activity of making cloud spend visible, allocated and controlled, and the name the providers give their billing consoles and tools for it. FinOps builds on cloud cost management but adds the collaboration, ownership and decision practice the FinOps Foundation's framework describes. See also: Cost reports and dashboards

Line item

One row of an invoice or a billing export: a service, a usage type or meter, a quantity in a pricing unit, a rate and an amount for one period and one account. The invoice aggregates many export rows into one line; the export keeps them apart. See also: Cost reports and dashboards

Credit

A negative amount the provider applies against charges: promotional, programme or contractual credits that reduce what is payable before the invoice, and refunds that return money for a charge already invoiced. Credits are not discounts; they expire, have a scope and belong to a period. See also: Cost reports and dashboards

Billing period

The span of usage one invoice covers, on AWS, Azure and Google Cloud normally a calendar month, closed a few days after month end when the invoice is issued. Usage dated inside the period can still be invoiced later; the invoice month and the usage month are two calendars. See also: Cost reports and dashboards

Invoice currency

The currency in which the provider issues the invoice and expects payment, set per billing account or billing profile and distinct from the pricing currency, usually US dollars, in which rates are published. The conversion rate and its date are the provider's and appear in the billing data. See also: Cost reports and dashboards

Billing account

The provider-side entity that receives invoices and pays for a set of accounts, subscriptions or projects: the management account of an AWS organization, an Azure billing account with its billing profiles, a Google Cloud Billing account. It decides who sees cost and which discounts are shared. See also: Cost allocation

Support plan

A paid tier of provider support, priced as a fixed monthly fee or as the greater of a minimum and a percentage of the monthly usage charges. It appears as its own line on the invoice at the paying account and is a shared cost that no single workload causes. See also: Cost allocation

Private pricing

Rates below the public list price that a customer negotiates with the provider, usually against a spend commitment over a term: AWS private pricing and its enterprise discount programme, Microsoft's enterprise agreement and consumption commitment, Google's negotiated contracts. The billing data shows the contracted rate beside the list rate. See also: Cost reports and dashboards

Usage export

A provider's machine-readable record of consumption, as opposed to its invoice: an API or file that lists units used (tokens, requests, hours, bytes) per time bucket and per key, project or model. AI providers deliver usage this way; the cost follows from it, not the other way round. See also: Billing-file imports

Restatement

A change to billing data for a period that was already delivered or invoiced: a credit, refund, tax or support charge added later, a corrected rate, or a line the provider reissues. A restated month no longer matches the number a report showed before, so reports record which load they used. See also: Cost reports and dashboards

Granularity

The resolution of billing data in two dimensions: time (hour, day, month) and object (account, service, SKU, resource, container). Finer granularity answers more questions and multiplies rows and storage; a level not chosen when the export was set up cannot be reconstructed afterwards. See also: Cost reports and dashboards

Reconciliation

The check that a billing export and the invoice describe the same money: export rows are tied to the invoice by its id and period, summed, and compared with the invoice total, with tax, credits, refunds, marketplace and currency explained as separate classes. What remains is a question for the provider. See also: Cost reports and dashboards

On-demand pricing

The price a provider charges for usage with no commitment: a published rate per unit such as an instance-second, a gigabyte-month or a request, billed in arrears. It is the baseline every commitment discount and spot price is measured against. See also: Cost reports and dashboards

Spot capacity

Spare compute capacity a provider sells below the on-demand price on condition that it can reclaim it at short notice. AWS, Azure and Google Cloud each set the price and the warning period differently, and the workload has to tolerate the interruption. See also: Savings

Tiered pricing

A price per unit that falls as usage within a billing period crosses published thresholds, as for object storage or data transfer. The tiers apply automatically and are counted per account or across an organization, depending on the provider. See also: Cost reports and dashboards

Committed use discount

Google Cloud's discount in return for a one- or three-year commitment, either to a quantity of resources in a region or to an hourly spend on eligible services. It applies automatically and is billed for the whole term whether the usage occurs or not. See also: Savings

Commitment coverage

The share of eligible usage, measured in on-demand cost or hours, that ran under a commitment discount in a period. Its counterpart, utilisation, is the share of the purchased commitment that usage actually consumed; together they show whether a commitment is too small or too large. See also: Savings

Commitment

A promise to a provider to use a quantity of resources or to spend an amount per hour for a fixed term, usually one or three years, in return for a lower rate. Reserved Instances, Savings Plans, Azure reservations and committed use discounts are its forms. See also: Savings

Serverless

A way of running code in which the provider supplies capacity per request or event and bills invocations and the memory and time they consume, not servers. Nothing runs between requests unless capacity is kept ready, which is billed separately. See also: Cost reports and dashboards

Object storage

Storage that keeps data as objects in buckets or containers, addressed by key over an API rather than mounted as a disk. Providers bill stored gigabytes per month, requests by type, retrieval from colder classes and data transferred out. See also: Cost reports and dashboards

Storage class

The tier an object is stored in, which sets its price per gigabyte-month against the cost of reading it: colder classes store for less but charge retrieval fees and a minimum storage duration. Azure calls them access tiers. See also: Savings

Snapshot

A point-in-time copy of a disk or volume that the provider stores and bills per gigabyte-month. Most snapshots are incremental, so only changed blocks add storage, and they keep billing after the source disk is deleted until they are deleted themselves. See also: Savings

NAT gateway

A managed network service through which resources without public IP addresses reach the internet or other networks. AWS, Azure and Google Cloud bill it by the hour and per gigabyte processed, on top of any data transfer the traffic causes. See also: Cost insights

Content delivery network (CDN)

A network of edge locations that caches content and serves it close to users. Providers bill data delivered from the edge and the requests served, at rates that differ from regional egress; a cache miss adds the fetch from the origin, which some providers charge for. See also: Cost insights

Bring your own licence (BYOL)

Using software licences a company already holds on cloud resources instead of paying the provider's licence-included rate. The software vendor's licence terms decide whether this is allowed, which licences qualify and whether dedicated hardware is required. See also: Savings

Account structure

The arrangement of AWS accounts, Azure subscriptions and resource groups, and Google Cloud projects, with the organizational units, management groups or folders above them. Every billing line carries its account, so the structure is the first allocation dimension and needs no tagging. See also: Cost allocation

Tagging policy

The written rule set for tags: which keys are mandatory, which values are allowed, how keys are spelled and cased, on which resources they must sit and who maintains them. It is what makes tags comparable across teams and providers. See also: Cost allocation

Allocation rule

A rule that assigns cost to an owner by conditions on the billing data, such as account, service, resource name or an existing tag, instead of a tag on the resource. Rules run in a fixed order and cover what tagging cannot reach. See also: Cost allocation

Unallocated cost

The part of the bill that no account, tag or rule assigns to an owner, kept as its own line rather than spread over everyone. Its size is the complement of allocation coverage, and someone has to own reducing it. See also: Cost allocation

Idle cost

The cost of capacity that is paid for but used or requested by no workload, such as the part of a Kubernetes node that no pod requested. It stays with the owner of the shared resource unless a rule spreads it across the workloads. See also: Cost allocation

Cost dimension

An attribute of a billing line by which cost can be filtered and grouped, such as provider, account, service, region, resource, tag or segment. Grouping by a dimension splits the same total into parts; filtering by it narrows the total. See also: Cost reports and dashboards

Reporting currency

The single currency in which a report shows cost from providers that invoice in different currencies. Each converted figure depends on a stated rate source and date, so a change in the total can come from usage or from the exchange rate. See also: Cost reports and dashboards

Data cut-off

The point in time up to which a report's billing data has arrived, stated with the report so its figures can be reproduced. Figures before the cut-off can still change through late charges and restatements until the provider finalises the period. See also: Cost reports and dashboards

Cost driver

A quantity whose change moves cost, such as instance hours, requests, stored gigabytes or a business volume like customers or orders. A change in cost splits into changes in drivers, in rates and in mix. See also: Cost insights

Budget alert

A notification sent when a budget's actual or forecasted cost crosses a threshold set as a percentage or an amount of the budget. It reports where cost stands against the amount; it changes usage only when an action is attached to it. See also: Budgets

Budget variance

The difference between actual or forecast cost and the budget for the same scope, period and cost basis, as an amount and a percentage. Variance analysis splits it into price, volume, mix, scope and exchange-rate effects. See also: Budgets

Budget exchange rate

The exchange rate fixed when a budget is set, at which cost billed in another currency is converted for the comparison with the budget. Comparing at the budget rate shows the usage and price variance; the difference to the actual rate is the currency effect.

Capex and opex

Capital expenditure (capex) acquires or creates an asset that is used over several periods and, where its useful life is limited, written off over that life; operating expenditure (opex) is expensed in the period it is incurred. Which applies to a cloud cost follows the accounting framework, and the auditor judges it.

Cost of goods sold (COGS)

The cost of producing what a company sold in a period, set against revenue to give gross profit. Cloud cost of the environments that serve customers can belong here, development and internal use elsewhere; the split follows the company's accounting policy. See also: Cost allocation

Accrual

An entry that records in a period the cost incurred in it whose invoice has not yet arrived, so the period carries its own cost. For cloud cost the estimate can come from the billing data before the invoice is final; how it is booked follows the accounting policy.

Auto-scaling

Adding and removing compute capacity automatically as load changes, driven by a metric, a schedule or a prediction. Billed capacity then follows demand instead of the peak, while the minimum size, the scaling metric and the cool-down decide how closely the bill follows the load.

Resource requests

The CPU and memory a Kubernetes container declares it needs. The scheduler places pods by their requests, so requested capacity occupies a node whether it is used or not, and the gap between requests and actual usage is what efficiency work on a cluster reduces. See also: Cost allocation

Commitment portfolio

All reservations, savings plans and committed use discounts a company holds, seen together with their types, scopes, terms and end dates. Viewed as one portfolio, they show how coverage, flexibility and expiry combine, which no single purchase decision shows. See also: Savings

Cost policy

A written rule set stating what teams may use and spend without asking: allowed regions, services and sizes, mandatory tags, approval thresholds, commitment purchases and access to cost data. Guardrails, budgets and reviews enforce it, and every exception carries a name and an end date. See also: Budgets

Policy as code

Governance rules such as required tags, allowed regions or instance sizes written as machine-readable definitions, kept in a repository, reviewed like code and deployed through a pipeline. The same definition can then block a non-compliant resource at creation and report existing resources that break the rule.

Cost owner

The named person who answers for the cost of a scope such as a team, product or environment: they see its cost report, explain its variance and decide within its budget. A scope without a cost owner has nobody to accept a saving or explain a change. See also: Cost allocation

Rate card

A provider's published table of rates per unit for each model or service, split by unit type such as input, output and cached tokens, and by tier, region or processing mode. The bill is usage times the rate in force on the day of use, so a new model or a price change moves cost even when usage stays the same.

Inference

Running a trained model on new input to produce an output, such as a language model answering a prompt, as opposed to training and fine-tuning, which create or adapt the model. Through a model API it is billed per token or request; on a dedicated endpoint or provisioned throughput it is billed per hour, whether requests arrive or not.

Capacity reservation

Compute capacity of a given instance type that a provider holds for a customer in a region or zone, as AWS, Azure and Google Cloud offer for virtual machines and GPU instances. It usually costs the on-demand rate for as long as it exists, used or not; unlike a commitment it secures capacity rather than a lower rate, though commitment discounts can apply to it.

Spend limit

A ceiling on spend that a provider enforces for an account, workspace or project: once it is reached, requests are refused or the service pauses until the period ends or the limit is raised, as with Anthropic's workspace spend limits or the spend caps on Google's model API. Unlike a budget alert it stops usage rather than warning about it, protecting the bill at the cost of availability.

Embedding

A list of numbers that an embedding model produces from a text or image so that similar content lies close together, used for semantic search and retrieval. Creating embeddings is billed per input token, storing and querying them in a vector database adds cost that grows with the number and dimensions of the vectors, and a new embedding model means embedding everything again.

AI agent

A program in which a language model decides step by step which tools to call and what to do with their results until a task is done, instead of answering a single prompt. Each step is a model request that resends the growing context, and tool calls can carry charges of their own, so the cost per task depends on how many steps it takes.

Prompt caching

A model provider feature that stores the processed beginning of a prompt, such as a system prompt or a document, so later requests starting with exactly the same text pay the reduced cached-input rate for that part. Cache writes can cost more than normal input and some providers bill storage per hour, so it saves money only when the same beginning recurs often within the cache lifetime.

Model routing

Sending each request to the model that a rule or classifier picks for it, such as a smaller, cheaper model for simple requests and a larger one for hard ones, either in the application or through a provider's router. The bill then follows the mix of models used, and a misrouted request costs either quality or money.

Contracting entity

The provider company that is party to a customer's agreement and issues its invoices, set by the customer's country, billing address and type of agreement; for European customers usually a European subsidiary such as Amazon Web Services EMEA SARL. It decides who bills and how VAT or the reverse charge appears, so the same service can come from different companies on different accounts.

Cloud region

A geographic area in which a provider runs data centres, such as Frankfurt or Ireland, usually divided into several isolated availability zones. Prices, available services and data transfer charges are set per region, so the same resource can cost a different amount depending on where it runs, and moving data between regions is charged on top.

Data residency

The commitment, by choice or because a rule requires it, that data is stored and processed only in a chosen country or area such as the EU, which providers support through region selection and through commitments such as Microsoft's to store and process customer data within the EU and EFTA. It limits which regions and services can be used and can rule out cheaper locations and global services. See also: Digital sovereignty

Switching charges

Charges a provider of cloud or other data processing services imposes for moving a customer to another provider or to its own infrastructure, data egress charges included, as the EU Data Act defines them; standard service fees and early termination penalties do not count. Until 12 January 2027 they may not exceed the provider's costs directly linked to the switch, and after that date they are not allowed. See also: Digital sovereignty

Cloud carbon footprint

The greenhouse gas emissions a provider attributes to a customer's cloud use, in CO2 equivalent, as AWS, Microsoft and Google report them in their carbon tools across scopes 1, 2 and 3 of their operations; for the customer they are indirect emissions in its supply chain. Each provider shares its emissions out among customers by its own method, and the tools differ in units and in which scope 2 method they show, so two providers' figures do not compare directly.

Cost as code

Cost controls such as budgets, alerts, allocation rules and cost estimates kept as version-controlled definitions, applied through infrastructure as code and checked in the delivery pipeline. A change to a budget or rule is then reviewed and traceable like any code change, and the cost of a resource can be estimated before it is deployed. See also: API and Terraform

Model Context Protocol (MCP)

An open protocol through which AI applications connect to tools and data sources: a server exposes functions and data that the model calls during a conversation. Anthropic introduced it in 2024, and since December 2025 it has been governed under the Linux Foundation; every call adds the data it returns to the context, where it is billed as tokens.

Usage type

The AWS billing field that names what a line item metered, such as BoxUsage:m6i.large for an instance hour or EBS:VolumeUsage.gp3 for a gigabyte-month of storage. Outside the Northern Virginia region it carries a region prefix (EUC1 for Frankfurt), and it is the key by which one service's charges are told apart in the Cost and Usage Report and in Cost Explorer. See also: Integrations

Extended support

A surcharge a provider bills once a managed service runs a version past its standard support window, instead of ending support: Amazon EKS charges a higher cluster-hour rate for Kubernetes versions in extended support, Amazon RDS a fee per vCPU-hour for database engine versions past their standard support, and GKE an hourly surcharge on its extended release channel. It appears as its own line and ends with the upgrade. See also: Savings

Node pool

A group of worker nodes in a Kubernetes cluster that share one machine type, disk, operating system and scaling range, called a node group on Amazon EKS and a node pool on AKS and GKE. It is the unit the cloud bills: every node in the pool is a virtual machine charged by the hour whether pods run on it or not, so the pool's machine type, minimum size and scale-down settings set the cluster's compute cost.

Meter

The unit of measure Microsoft bills an Azure resource by. Each meter has a category, a subcategory, a name, a unit and a rate, and one resource produces several of them: a virtual machine bills compute hours, disks, a public IP address and bandwidth as separate meters. Cost exports carry the meter identifiers on every line, which is how a line is traced back to the service and the price that produced it.

SKU

A stock keeping unit: the smallest priced item in a provider's catalogue, with its own identifier, unit and price. Google Cloud bills every service as SKUs and names the SKU on each line of the billing export; Azure uses the word for a resource's size or tier, with the meters underneath it. Reading a bill at SKU level shows which priced item a service turned into.

Provisioned throughput unit (PTU)

A unit of reserved model capacity on Azure OpenAI and Microsoft Foundry, billed per hour or as a monthly or yearly reservation whether or not tokens flow through it, instead of per token. A deployment is sized in units by its expected tokens per minute, and capacity left unused still costs the hourly rate, so the unit pays off only above a steady utilisation.

Lifecycle rule

A rule on an object storage bucket or container that moves objects to a colder storage class or deletes them after a set age, by prefix, tag or version state. S3 lifecycle configuration, Azure Blob Storage lifecycle management and Cloud Storage lifecycle rules are the providers' forms; a bucket without one keeps every version, every incomplete multipart upload and every log file at the price of the class it was written to. See also: Savings

Provisioned capacity

Capacity a service bills for because it was reserved for a resource, whether or not it is used: DynamoDB read and write capacity units, Azure Cosmos DB request units per second, Kinesis shards, BigQuery slots, provisioned IOPS on a volume. The bill follows the provisioned figure, so the saving lies in the gap between what is provisioned and what is consumed, and the provider's metrics show both. See also: Savings

RACI

A matrix that assigns one of four roles to each person or team for each activity: responsible, who does the work; accountable, who owns the outcome and is exactly one; consulted, whose input is sought; informed, who is told the result. For cloud cost it fixes who tags, who reviews, who buys commitments and who answers for the bill, so an unowned saving has somewhere to go.

Included traffic

The outbound data volume a provider includes in a server's or plan's monthly price, per server or pooled per account, with traffic beyond it billed per unit and inbound traffic usually free. It keeps the bill flat until the allowance is crossed, and it is the figure to check before a comparison with a provider that meters every gigabyte.