The short answer
Financial statements follow accounting rules. That does not mean the data extracted from every filing arrives in one uniform, directly comparable shape.
Companies can use different labels for economically similar items, choose different XBRL concepts, disclose different levels of detail, and reorganize segments over time. Within a single filing, a consolidated total, its components, and a product or segment breakdown may all appear as valid facts.
As-reported financial data preserves what the company disclosed: its label, value, XBRL concept, period, unit, dimensions, filing, and context.
Standardized financial data maps selected disclosures into a shared schema so analysts can compare companies and periods without manually reconciling every label and tag.
| As reported | Standardized | |
|---|---|---|
| Main question | What did the company disclose? | What common concept can this support? |
| Labels and tags | Preserved from the filing | Mapped to a shared vocabulary |
| Context and dimensions | Retained at source detail | Used to organize comparable observations |
| Main strength | Fidelity to the source | Comparability and consistency |
| Main risk | Semantics take more work to reconcile | Mapping can imply too much equivalence |
| Typical uses | Reconciliation, accounting research, audit trail | Screening, models, peer analysis |
The practical goal is to make a standardized value comparable while preserving the route back to the original disclosure.
Why financial data needs standardization
Consider three fiscal 2025 top-line observations: Carvana reported $20.322bn as “Net sales and operating revenues”; Community Health Systems reported $12.485bn as “Net operating revenues”; Kohl’s reported $15.527bn as “Total revenue,” made up of $14.775bn of net sales and $0.752bn of other revenue. Carvana 10-K ↗ Community Health Systems 10-K ↗ Kohl’s 10-K ↗
The labels differ. In SourceState’s current mapping, the Carvana and Kohl’s observations use us-gaap:RevenueFromContractWithCustomerExcludingAssessedTax; Community Health Systems uses us-gaap:Revenues. SourceState maps these selected observations to its common Revenue concept while retaining their original labels and tags.
us-gaap:us-gaap:us-gaap:A common revenue field makes growth, margins, screens, and cross-sectional models easier to calculate. It does not make a vehicle retailer, a hospital operator, and a department store economically identical. The analyst still needs to understand each company’s business and presentation.
A tag is evidence, not the whole interpretation
An XBRL fact is more than a number beside a concept name. Its meaning also depends on the entity, period, unit, dimensions, and its location and relationships within the filing.
A filing may contain a consolidated total and facts disaggregated by product, geography, or segment. Each can be valid, yet represent a different level of the same reporting structure. Summing every fact that sounds like “revenue” can count the same economics twice.
Kohl’s: dimensions and reporting hierarchy
Kohl’s fiscal 2025 filing reports $15.527bn of total revenue: $14.775bn of net sales plus $0.752bn of other revenue. It separately reports net sales by line of business. Kohl’s FY2025 Form 10-K ↗
The six line-of-business values sum to net sales. They are its breakdown, not additional amounts to add on top of total revenue. A flat extraction that adds every displayed row would double-count; the problem is the lost relationship between facts, not an inconsistency in the filing.
Standardized does not mean interchangeable
UnitedHealth Group reported $352.229bn of premiums and $447.567bn of total revenues for 2025. Premiums are a related revenue measure, but they are not the consolidated total. UnitedHealth Group 2025 Form 10-K ↗
In SourceState’s model, premiums are related to the broader revenue concept as a selected subtotal. An ontology can express that relationship without making the subtotal a synonym for total revenue. The same care applies to segment revenue, product revenue, and company-defined operating metrics.
Similar labels, different facts
The reverse problem occurs too: the same displayed label can describe different accounting roles. DXC’s fiscal 2026 filing reports “Depreciation and amortization” of $1.160bn on the income statement and $1.182bn as a cash-flow adjustment. They are not duplicate values. DXC FY2026 Form 10-K ↗
SourceState distinguishes the observations using statement role, concept, and context. A label helps a reader, but does not settle whether two facts are equivalent.
Company-specific extensions call for restraint
A company-specific XBRL extension can represent a disclosure for which a filer does not consider a standard concept suitable. Mapping it to a broad standard item may hide an important distinction.
CoreWeave’s Q2 2026 filing separately reports current and non-current, recourse and non-recourse debt. SourceState’s example treatment retains those distinctions and source identities instead of collapsing them into one generic debt value. This is SourceState’s interpretation of the reported structure, not a company-provided normalized concept. CoreWeave Q2 2026 Form 10-Q ↗
An unmapped fact is not necessarily a data-quality failure. Where equivalence is unsupported, preserving the extension and its lineage can be more informative than forcing a broad mapping.
DXC: one historical total, two segment presentations
DXC originally reported fiscal 2024 revenue across two segments: Global Business Services (GBS) at $6.820bn and Global Infrastructure Services (GIS) at $6.847bn, totaling $13.667bn. DXC FY2024 Form 10-K ↗
DXC later adopted a three-segment structure. Its fiscal 2026 filing recasts fiscal 2024 as Consulting & Engineering Services (CES) at $5.274bn, GIS at $7.230bn, and Insurance Software & Services at $1.163bn—again $13.667bn in total. DXC FY2026 Form 10-K ↗
FY2024 Form 10-K
Originally reported
FY2026 Form 10-K
Prior period recast
The recast view may be more useful for analysis under the current segment structure. The original view answers what investors could see in the fiscal 2024 filing. Both describe that period from different filing vintages.
Financial history has more than one timeline
A later filing can change how an earlier period is presented. A formal correction, restatement, segment recast, new requirement, and presentation change are different events; a dataset should identify which applies.
Replacing an earlier observation with a later one may preserve the latest view but make the original disclosure impossible to reconstruct. Filing date, source document, and value vintage let an analyst ask either question.
Period end and public availability are also different dates. A quarter can end on March 31 and be reported weeks later. Historical analysis using only period end can introduce later information into an earlier decision date.
When each representation is useful
As-reported data
Use it when exact disclosure or presentation matters:
- reconcile a model to filed statements;
- investigate accounting and disclosure changes;
- analyze segments, products, geographies, and company metrics;
- check a standardized value against its source;
- rebuild what was public at a historical date.
Its strength is fidelity. Issuer-specific labels and structures take more work to compare at scale.
Standardized data
Use it when analysis spans companies or periods:
- screen issuers using common fields;
- compare growth, margins, leverage, and returns;
- supply models and portfolio analytics with consistent definitions;
- build series across label and taxonomy changes.
Its strength is comparability. Quality depends on mapping rules, aggregation, and reviewable lineage.
Why serious research needs both
A useful data model connects the layers. The standardized concept supports comparison; the reported label preserves the company’s presentation; the XBRL fact and context retain machine-readable meaning; the filing supplies the evidence.
What a reliable data model should preserve
- Source filingIssuer, form, accession, filing date, and document locator.
- As-reported factOriginal label, concept or extension, value, unit, period, and dimensions.
- Standardized interpretationCommon concept, aggregation level, mapping status, and subtotal relationships.
- LineageA path from the standardized observation to the original fact and source passage.
- Filing vintageWhether the view comes from an original, amended, restated, or recast presentation.
These fields make it possible to compare, inspect, explain, and revise a mapping without losing the evidence behind it.
Frequently asked questions
Is standardized data more accurate than as-reported data?
Not by definition. Standardization is an interpretation for comparison. Accuracy depends on the question and on whether the reported value and context remain inspectable.
Does the same XBRL tag always mean the same thing?
No. Entity, period, unit, dimensions, statement location, and reporting context also matter.
Can different XBRL tags map to one standardized concept?
Yes, when the observations support the stated comparison. Original tags and the mapping scope should remain available.
Should every company-specific extension be mapped?
No. Preserve the extension and lineage when equivalence to a standard concept is not supported.
Why keep old filing versions after a recast?
The later recast supports current comparisons; the earlier filing shows what was reported at the time. Both are needed for different questions.
Primary references
External links open the SEC filing in a new tab. Company-reported figures come from these sources; SourceState mappings and figure organization are SourceState interpretations.
- Carvana Co., 2025 Form 10-K ↗ — net sales and operating revenues.
- Community Health Systems, Inc., 2025 Form 10-K ↗ — net operating revenues.
- Kohl’s Corporation, 2025 Form 10-K ↗ — total revenue, net sales, and lines of business.
- DXC Technology Company, FY2024 Form 10-K ↗ — original segment presentation.
- DXC Technology Company, FY2026 Form 10-K ↗ — recast prior period and depreciation examples.
- UnitedHealth Group, 2025 Form 10-K ↗ — premiums and total revenues.
- CoreWeave, Q2 2026 Form 10-Q ↗ — recourse and non-recourse debt presentation.
For background on machine-readable filing data, see the SEC’s EDGAR API overview ↗ and XBRL International’s Inline XBRL guidance ↗.