XBRL Dimensions Explained: How Axes and Members Represent Segments, Products, Geographies, and More

What is an XBRL dimension?

A financial fact usually has several pieces of meaning attached to it.

At a simplified level:

Concept
Revenue

Value
$7.908bn

Entity
Broadcom

Period
FY2025

Unit
USD

Dimensions
Product or Service = Product
Geography = Americas

The first five items tell us what was reported, by whom, when, and in what unit.

The dimensions tell us which slice of the reported concept the fact represents.

A taxonomy-defined dimension is commonly referred to as an axis. A member identifies the category on that axis that applies to the fact.

The key idea is straightforward:

AXIS
What characteristic is changing?

MEMBER
Which value of that characteristic applies?

For example:

StatementGeographicalAxis
    ├── AmericasMember
    ├── AsiaPacificMember
    └── EMEAMember

or:

StatementBusinessSegmentsAxis
    ├── SemiconductorSolutionsMember
    └── InfrastructureSoftwareMember

The same financial concept can appear under any of those members.

That is how XBRL preserves disclosure structure without creating a separate accounting concept for every possible segment, geography, or product category.

Axis versus member

An axis defines the dimension along which a fact is being qualified.

A member identifies the particular position on that axis.

For example:

Axis:
StatementGeographicalAxis

Member:
AsiaPacificMember

The axis means:

The member means:

The fact might therefore be represented as:

Concept: Revenue
Value: $35.896bn
Axis: StatementGeographicalAxis
Member: AsiaPacificMember

The same Revenue concept could appear again with AmericasMember or EMEAMember.

Those are not duplicate revenue facts. They are separate members of the same dimensional view.

A simple one-axis example

Hilton's fiscal 2025 property, plant and equipment disclosure provides a clean example.

Hilton reported net PP&E of $684 million and split it into:

Domestic PP&E      $246m
Foreign PP&E       $438m
                   -----
Total PP&E         $684m

The dimensional structure is:

Concept
PropertyPlantAndEquipmentNet

Axis
GeographicDistributionAxis

Members
Domestic
Foreign

Conceptually:

The two member facts reconcile exactly to the dimensionless total.

This is the simplest useful mental model:

A naïve extractor that returns the consolidated $684 million plus the $246 million and $438 million member facts without preserving the dimensional relationship can make the data appear to contain $1.368 billion of PP&E.

The filing is not double counting. The extraction is.

Source: Hilton Worldwide Holdings, FY2025 Form 10-K, PP&E geographic distribution disclosure, filed February 11, 2026.
https://www.sec.gov/Archives/edgar/data/1585689/000158568926000007/hlt-20251231.htm ↗

Standard axes can contain company-specific members

A common pattern is:

standard axis
+
company-specific member

Broadcom provides a useful example.

For fiscal 2025 revenue, Broadcom uses the standard ProductOrServiceAxis. One member is the standard ProductMember. Another is a Broadcom-specific member representing Subscriptions and Services.

The resulting breakdown is:

Products                     $44.847bn
Subscriptions and services    19.040bn
                             ---------
Total revenue                $63.887bn

The axis is standard. One of the categories on that axis is company-specific.

That distinction matters because a custom member does not automatically mean the underlying fact is incomparable or unusable. Sometimes the custom member is simply the company's own economically meaningful category inside a standard dimensional framework.

Source: Broadcom, FY2025 Form 10-K, revenue disclosures, filed December 18, 2025.
https://www.sec.gov/Archives/edgar/data/1730168/000173016825000121/avgo-20251102.htm ↗

One total can have several independent dimensional views

Broadcom's fiscal 2025 revenue provides a stronger example of why dimensional data cannot be treated as one flat hierarchy.

The same $63.887 billion of revenue is presented through several different views.

Product or service

Products                     $44.847bn
Subscriptions and services    19.040bn
                             ---------
Total                        $63.887bn

Business segment

Semiconductor Solutions      $36.858bn
Infrastructure Software       27.029bn
                             ---------
Total                        $63.887bn

Geography

Americas                     $18.939bn
Asia Pacific                  35.896bn
EMEA                            9.052bn
                             ---------
Total                        $63.887bn

Each breakdown reconciles independently to consolidated revenue, but the members across those axes cannot be combined.

For example, Semiconductor Solutions revenue and Asia Pacific revenue overlap economically. A semiconductor sale in Asia Pacific can be represented in both views.

Conceptually:

Each branch answers a different question:

• What was sold?
• Which business segment earned the revenue?
• Where was the revenue attributed?

Broadcom also reports a separate country-level geographic grouping. That grouping reconciles to the same consolidated total, but it should not be mechanically combined with the regional presentation or treated as though one was necessarily a formal hierarchy of the other.

Source: Broadcom, FY2025 Form 10-K, product, segment, and geographic revenue disclosures.
https://www.sec.gov/Archives/edgar/data/1730168/000173016825000121/avgo-20251102.htm ↗

One fact can have multiple dimensions at the same time

An XBRL fact can carry more than one dimension simultaneously.

This is different from having several independent breakdowns.

Consider a fact that means:

To represent that precisely, the context needs to say both:

ProductOrServiceAxis = ProductMember

and

StatementGeographicalAxis = AsiaPacificMember

The fact is not merely product revenue. It is not merely Asia Pacific revenue. It sits at the intersection of both dimensions.

That is where dimensional data starts to resemble a multidimensional table rather than a simple parent-child list.

A two-dimensional matrix: what was sold and where

Broadcom's fiscal 2025 revenue disclosure gives a particularly clean example.

The filing reports revenue using product/service and geography simultaneously.

Each cell contains the same broad revenue concept, but with two dimensional qualifiers.

For example:

Concept
Revenue

ProductOrServiceAxis
ProductMember

StatementGeographicalAxis
AsiaPacificMember

Value
$33.596bn

The $33.596 billion fact means:

If a data pipeline drops either axis, the meaning changes.

Drop geography, and the system may confuse the cell with total product revenue.

Drop product/service, and it may confuse the cell with total Asia Pacific revenue.

The complete matrix reconciles to consolidated revenue, but the six cells must be interpreted as intersections—not as a second set of unrelated facts.

Source: Broadcom, FY2025 Form 10-K, Item 8 revenue tables.
https://www.sec.gov/Archives/edgar/data/1730168/000173016825000121/avgo-20251102.htm ↗

A three-axis fact: why every dimension can matter

Broadcom's customer-concentration disclosure goes one step further.

The company disclosed that one customer accounted for 32% of net revenue in fiscal 2025.

That percentage is qualified by three separate axes:

Each dimension contributes distinct meaning.

Customer axis

Which customer does the percentage refer to?

Benchmark axis

What is the concentration measured against?

In this case, net revenue.

Risk-type axis

What kind of concentration is being described?

Customer concentration.

Without those qualifiers:

Concentration = 32%

is ambiguous.

Thirty-two percent of what? Which concentration measure? Which customer?

The dimensional signature turns a generic percentage into a specific disclosure.

This also illustrates an important limitation of treating member labels as permanent entity identifiers. "Major Customer One" describes a role in the filing's dimensional structure; it should not automatically be assumed to identify the same real-world customer across reporting periods.

Source: Broadcom, FY2025 Form 10-K, "Concentrations of credit risk and significant customers."
https://www.sec.gov/Archives/edgar/data/1730168/000173016825000121/avgo-20251102.htm ↗

Dimensionless facts and dimensioned totals

In many disclosures, a fact without an explicitly reported taxonomy-defined dimension represents the consolidated or otherwise unqualified observation.

For example:

Broadcom revenue
No explicit product, segment, or geography member
$63.887bn

That is the natural reference point against which the dimensional breakdowns reconcile.

But it is dangerous to turn this into a universal rule:

A fact can also appear as a dimensioned total, so dimensional presence alone does not determine whether an observation is consolidated.

Disney's fiscal 2025 revenue illustrates the latter point.

The filing contains dimensionless consolidated revenue of $94.425 billion. It also contains a dimensioned total using a consolidation axis and an operating-segments member with the same value.

The segment members are:

Entertainment                   $42.466bn
Sports                           17.672bn
Experiences                      36.156bn
Segment eliminations             -1.869bn
                                ---------
Total                           $94.425bn

So a system can encounter:

• the dimensionless total;
• a dimensioned operating-segment total;
• the underlying segment members;
• a negative elimination member.

All are valid.

A deduplication rule based only on concept, period, and value could incorrectly collapse meaningful structure—or count the total twice.

Source: The Walt Disney Company, FY2025 Form 10-K, segment disclosures, filed November 13, 2025.
https://www.sec.gov/Archives/edgar/data/1744489/000174448925000155/dis-20250927.htm ↗

The double-counting trap

Dimensional data becomes especially dangerous when a filing includes both subtotals and the components underneath them.

Hilton provides a clean example.

For fiscal 2025, Hilton reported total revenue of $12.039 billion.

The filing includes:

Total revenues excluding reimbursable revenues     $4.954bn
Cost reimbursement revenues                          7.085bn
                                                   --------
Total revenue                                      $12.039bn

The $4.954 billion subtotal is itself broken into components:

Base and other management fees                      $0.376bn
Franchise and licensing fees                         2.780bn
Incentive management fees                            0.313bn
Ownership                                            1.233bn
Other revenues                                       0.252bn
                                                   --------
Revenue excluding reimbursements                    $4.954bn

A naïve extraction might collect the total, subtotal, reimbursable revenue, and all five components, then treat every value as an independent category.

The result would be substantial double counting.

The mistake is assuming:

A robust dimensional model needs to preserve enough structure to distinguish totals, subtotals, components, eliminations, and overlapping views.

Source: Hilton Worldwide Holdings, FY2025 Form 10-K, consolidated revenue disclosure.
https://www.sec.gov/Archives/edgar/data/1585689/000158568926000007/hlt-20251231.htm ↗

Company-specific axes and members

Companies sometimes need a dimensional structure that the standard taxonomy does not provide.

In those cases, a filer can define a custom axis, custom members, or both.

Hilton's remaining-performance-obligation disclosure provides a useful example.

The filing uses a Hilton-defined:

RevenueTypeAxis

with company-specific members for categories including:

• Loyalty Program Revenues;
• Application, Initiation and Other Fees;
• Other Obligations.

The associated remaining-performance-obligation values include:

Loyalty Program Revenues                 $1.514bn
Application, initiation and other fees    0.825bn
Other Obligations                         0.015bn

Here, both the organizing axis and the members are company-specific.

The challenge for financial-data systems is to preserve that structure without pretending the custom categories are automatically equivalent to standardized dimensions used by other issuers.

There is also an important caution: the loaded facts do not provide an unqualified total that should be inferred from these rows.

Even when several members sit on one axis, a system should not assume that the members form a complete partition of some total unless the filing and taxonomy structure support that conclusion.

Source: Hilton Worldwide Holdings, FY2025 Form 10-K, "Revenues from Contracts with Customers—Additional Information."
https://www.sec.gov/Archives/edgar/data/1585689/000158568926000007/hlt-20251231.htm ↗

Dimensions are not limited to financial statement dollars

Dimensions can also qualify operational facts.

CoreWeave's Q2 2026 filing includes a company-specific fact describing 393 MW of electrical power access not yet commenced at a single-site data center.

Its structure includes:

Concept
ElectricalPowerAccessNotYetCommenced
(company-specific)

Value
393 MW

Period
June 30, 2026
instant

Axis
PropertyPlantAndEquipmentByTypeAxis
(standard)

Member
SingleSiteDataCenterMember
(company-specific)

The dimensional member tells us that the 393 MW applies to a specific type of data-center asset.

The filing explains that the power remained undelivered and was expected in phases.

Without that context, it would be easy to misread the figure as total CoreWeave power capacity, operating capacity already in service, or a standardized financial statement measure.

It is none of those.

This matters because many economically important disclosures are operational rather than purely financial: power, capacity, customers, units, subscribers, bookings, backlog, utilization, concentration, and contracted commitments.

Source: CoreWeave, Q2 2026 Form 10-Q, "Leases Not Yet Commenced," filed August 12, 2026.
https://www.sec.gov/Archives/edgar/data/1769628/000176962826000366/crwv-20260630.htm ↗

How to work with dimensions safely

When you encounter dimensioned XBRL facts, ask at least the following questions.

1. What is the concept?

What economic fact is being reported?

2. Which axes qualify it?

Do not reduce the context to a single optional "segment" field if several axes are present.

3. Which member belongs to each axis?

Keep the axis-member pairing intact.

GeographyAxis → AsiaPacificMember

is meaningful. AsiaPacificMember without its axis is less informative.

4. Is the member standard or company-specific?

A custom member may still represent a clear economic category. It simply requires preservation of the company's semantics.

5. Is the fact a total, subtotal, leaf, or elimination?

Do not assume every member can be summed.

6. Are there several independent views?

Segment, product, and geography often represent overlapping ways of slicing the same company.

7. Does one fact contain several axes simultaneously?

A product-by-geography cell is different from a standalone product total or geography total.

8. Do the members actually reconcile?

Check the filing rather than assuming that every axis forms a complete partition.

9. Is there a dimensionless or dimensioned reference total?

Both may exist.

10. What does the filing say the dimension means?

For geography in particular, the attribution rule can vary.

Why dimensions matter for investors and financial-data systems

Dimensions are sometimes treated as a technical XBRL feature.

For investment research, they are often where the economically useful detail lives.

A consolidated income statement might tell you:

Revenue = $63.887bn

Dimensions can reveal:

Semiconductor Solutions = $36.858bn
Asia Pacific = $35.896bn
Subscriptions & Services = $19.040bn
Products in Asia Pacific = $33.596bn

Those observations answer different research questions.

They can support segment growth analysis, geographic exposure, product mix, customer concentration, business-mix analysis, operational KPI extraction, custom screens, and peer comparisons.

But only if the system preserves the structure that gives each observation meaning.

Flatten the dimensions, and much of that information becomes ambiguous.

Sum them naïvely, and it becomes wrong.

How SourceState treats dimensional data

SourceState preserves dimensions as part of the structure of an as-reported fact rather than treating them as decorative XBRL metadata.

A dimensioned observation can remain connected to:

• its financial or operational concept;
• reporting period;
• unit;
• every axis in the context;
• the member associated with each axis;
• whether an axis or member is standard or company-specific;
• source filing;
• source location;
• standardized concept where appropriate.

This makes it possible to distinguish:

Broadcom Revenue
FY2025
$63.887bn
no explicit product/geography dimension

from:

Broadcom Revenue
FY2025
$44.847bn
ProductOrServiceAxis = Product

from:

Broadcom Revenue
FY2025
$33.596bn
ProductOrServiceAxis = Product
StatementGeographicalAxis = Asia Pacific

All three belong to the broader revenue concept.

They are not interchangeable facts.

The purpose is to make detailed disclosure usable for research without destroying the distinctions the company actually reported.

The simplest way to remember axes and members

An axis answers:

A member answers:

And a single fact can sit at the intersection of several axes:

Revenue = $33.596bn

Product axis
→ Product

Geography axis
→ Asia Pacific

That is not metadata surrounding the number.

It is part of what the number means.