Allocating Shared and Unknown Spend

Some cloud costs cannot be organized using metadata alone. A shared database serves multiple applications, but the billing data only shows a single line item. Other costs have no tags or account attributes at all, leaving them entirely unallocated. CloudZero handles both cases the same way: Allocation Dimensions split these costs across teams, products, customers, or any other grouping that fits your business.

This guide covers how to identify shared or unallocated costs, choose an allocation method, and combine the result into a single Dimension.

What you need

A Dimension where the costs you want to split are visible. This can be:

• A prebuilt Cloud Provider Dimension (such as Account, Service, or Region)
• A custom Grouping Dimension you built in Dimension Studio or CostFormation

Step 1: Identify what to split

Before creating an Allocation Dimension, identify the costs you want to split. You can allocate:

• Unallocated spend: Costs that do not match any rule in an existing Dimension, appearing under Not In Dimension. For example, if your Team Dimension leaves 30% of spend unallocated, you can split that 30% proportionally across the teams that do have spend.
• Identified elements: Costs already grouped into a named Dimension element. For example, a Shared Database element in a Grouping Dimension that isolates your shared database costs by service or account, or a Pre-Prod element that you want to redistribute across development, QA, and staging.

Step 2: Choose an allocation method

CloudZero supports two allocation methods. Both work in Dimension Studio and CostFormation YAML.

Option A: Allocate By Rules

Use Allocate By Rules when you want to use known spend that exists in CloudZero already to allocate the shared or unknown spend. This can be done evenly or proportionally:

• Even splits costs equally across all elements regardless of usage. For example, there's a $10,000 database that's shared across the Mobile, Web, and Platform teams. CloudZero splits the cost evenly at $3,333.33 to each team.
• Proportional splits costs based on each group's share of existing spend. For example, a serverless function that calls the shared $10,000 database is already broken down by metadata: Mobile at $1,200, Web at $600, and Platform at $200. CloudZero uses those spend amounts to calculate proportions automatically (60%, 30%, 10%) and splits the $10,000 database cost based on those calculated proportions: $6,000 to Mobile, $3,000 to Web, and $1,000 to Platform. Proportions recalculate for each time period as new data arrives.

For allocations that need different split logic per element, see ForEachElementOf.

Option B: Allocate By Streams

See Telemetry Streams to learn how to send non-cost data (such as observability metrics, security data, marketing data, or other usage data) into CloudZero.

Once your telemetry data is in CloudZero, the usage patterns in that data determine how costs are split proportionally. As new telemetry data arrives, the allocation updates automatically.

Example: You have $10,000 in shared database compute costs. Your observability platform shows 100,000 database queries total for January 25, where Mobile made 60,000, Web made 30,000, and Platform made 10,000. CloudZero calculates the proportions for that day (60%, 30%, 10%) and splits the $10,000 as $6,000 to Mobile, $3,000 to Web, and $1,000 to Platform.

How telemetry data maps to allocation

Every Allocation Stream has four components:

Example: A shared RDS database serves three teams. Your observability platform tracks compute and storage usage per team per day:

Team is the element name, Usage is the allocation value, and Service and Usage Type are filter columns that scope the allocation to RDS compute and RDS storage independently.

Send your telemetry data as an Allocation Stream via CSV upload or the API.

If your Allocation Dimension uses multiple streams, the first stream listed takes precedence.

Rate-based allocation

Allocate By Streams also supports a fixed rate per unit of usage, where each element's cost is their usage multiplied by a rate you set. This is useful for internal chargeback with agreed-upon transfer prices. The remaining spend goes to a default element so the total always reconciles with the actual bill. For syntax details, see CostFormation Reference.

Step 3: Set it up

There are multiple ways to design your allocation, but this is the approach we recommend:

1. Create an Allocation Dimension that splits the shared or unknown costs you identified in Step 1, using the method you chose in Step 2. Follow the guide for Dimension Studio or CostFormation YAML. Most organizations hide this Dimension (hidden: true in Dimension Studio or CostFormation) because it serves as an input to the combined view, not something users interact with directly.

2. Create a combined Grouping Dimension using a GroupBy rule that references both your Allocation Dimension and your Grouping Dimension. List the Allocation Dimension first, then the Grouping Dimension. The Allocation Dimension splits shared costs and redistributes them. The Grouping Dimension then organizes the remaining spend. Because the allocation already redistributed the shared costs, the shared element disappears from the final result.

   Example: Your Team Dimension has Mobile at $8,000, Web at $5,000, Platform at $2,000, and Shared at $10,000. Your Allocation Dimension splits the $10,000 Shared element as $6,000 to Mobile, $3,000 to Web, and $1,000 to Platform. With the Allocation Dimension listed first, the combined result is Mobile at $14,000, Web at $8,000, and Platform at $3,000. Shared is gone because the allocation already redistributed it.

A typical setup uses three Dimensions. Using the Team example:

Filtering noise and small elements

If an Allocation Dimension produces many elements where most account for very little spend, you can set a cutoff threshold to filter out the noise. Elements below the threshold are grouped into a single named element, or, when no name is set, their spend is spread across the elements that remain. This is available for proportional and standard telemetry allocations. For configuration details, see CostFormation Reference.

Keep allocations efficient

Most allocation setups process quickly as they are. As you add allocations, especially ones that build on each other, a few habits keep processing quick and your results easy to read.

Each Allocation Dimension turns a shared cost into one row per element. The way your allocations fit together affects how much work each publish takes, and how much CloudZero works through when you explore several Dimensions together.

Split each shared cost once

An Allocation Dimension can split the output of another Allocation Dimension. The second multiplies the rows the first produced, and processing time grows with the longest chain of Dimensions that build on one another. When you can, point each allocation at the original shared spend, such as a hidden Grouping Dimension that isolates it. For a ready-made example, see Split several shared costs from one source. When a later step needs an earlier allocation's result, keep the chain short.

A Grouping Dimension that combines an allocation with your other costs, as in Step 3, keeps one row per line and adds no extra rows.

To spot it: an allocation's SpendToAllocate source or conditions name another Allocation Dimension, directly or through a Grouping Dimension that reads one.

Give each shared cost one allocation

When two Allocation Dimensions split the same costs, CloudZero works through every combination of their elements. Isolate the shared spend in one hidden Grouping Dimension, with one element per kind of shared cost, and give each allocation its own element. If you keep two allocations of the same spend on purpose, such as one for showback and one for chargeback, that's a reasonable choice.

To spot it: two allocations whose SpendToAllocate conditions match the same service, account, or tag value.

Keep test copies out of processing

A copy of an allocation saved in another namespace is processed like the original, so the two split the same costs twice. When you finish testing, set Disable: true on the copy or delete it. Hide: true only removes a Dimension from view, and CloudZero still processes it.

Filter small elements

An allocation with a long tail of small elements adds rows that carry little spend. A cutoff keeps the elements that carry most of the spend and combines or spreads the rest, so choose a threshold that still shows the detail you report on. See Filtering noise and small elements.

Publish related changes together

Each publish reprocesses the Dimensions you changed, and any Dimensions that build on them, across your full cost history. Save related edits and publish them together.