1. Introduction
Transfer pricing benchmarking under the TNMM (or CPM in the U.S.) depends on a credible comparables set. The standard workflow is familiar: screen a database by SIC code or industry keyword, apply financial filters, and document why each company is or is not comparable to the tested party. In practice, this often means analysts manually reviewing years of 10-K filings to determine whether a comparable that passed a database screen in 2018 still represents the same business in 2021. What that workflow tends to underweight is the functional dimension, not just what a company’s financials look like, but what the company actually does, how it earns revenue, and whether its business model in the audit year resembles the tested party’s.
That question is harder to answer than it sounds when the relevant period spans three to five years and the comparables set contains many companies. EdgarStat Annual Business Descriptions were built to make this challenge tractable. The dataset covers annual business descriptions for several thousands SEC-filing entities across multiple years, systematically generated from each company’s annual 10-K filing, enabling longitudinal functional analysis at a scale that manual filing review cannot match.
2. The Source Material: What Goes Into a Business Description
Each EdgarStat Annual Business Description is generated from the annual 10-K filing of a U.S.-listed company. The source sections are Items 1, 1A, 7, and 8, the Business section, Risk Factors, Management’s Discussion and Analysis, and the Financial Statements. Together these sections contain the most decision-relevant narrative in any annual report.
From that material, the description extracts and standardizes the following attributes:
| Attribute | Transfer Pricing Relevance |
| Business activities and operating model | Functional comparability with the tested party |
| Primary products and services | Product-line comparability |
| Revenue-generating business segments | Segment-level PLI analysis |
| Customer and partner ecosystem | Contractual and economic comparability |
| Key customer or supplier concentration risks | Identification of non-routine risk exposure |
| Geographic footprint and headquarters | Market comparability adjustments |
| Corporate structure (independent vs. subsidiary) | Elimination of captive or controlled entities |
| Year of incorporation and historical background | Assessment of company maturity |
| Significant M&A, divestitures, and name changes | Detection of structural breaks in financial data |
| Major organizational or branding changes | Year-specific functional profile verification |
The output is a concise, year-indexed summary that preserves the filing’s own terminology where that language most accurately describes the business.
Two important caveats apply. First, the descriptions capture only what the company discloses. If a filing does not name major customers, does not break out segment revenue, or consolidates information that practitioners might prefer to see separately, that gap will be reflected in the description. The quality of the output is bounded by the quality and completeness of the underlying filing. Second, some information may appear in sections outside Items 1, 1A, 7, and 8, in legal proceedings, exhibits, or supplemental schedules, and will not be captured unless it surfaces in the primary source sections.
3. The Comparables Problem in Multi-Year Benchmarking
TNMM analysis typically covers three to five years of financial results for each comparable, and audit disputes can reach further back. A company accepted into a comparables set in year one may have undergone a material transformation by year three: an acquisition that shifted its revenue mix, a divestiture that removed an entire business segment, or an internal reorganization that changed its functional profile entirely.
Standard database screens do not catch these events systematically. A company that was a pure-play distributor in 2015 and a vertically integrated manufacturer by 2018 will pass the same SIC-code filter in both years. The financial ratios may look continuous even when the underlying business has changed fundamentally.
EdgarStat Annual Business Descriptions address this directly. Because each year’s description is generated from that year’s filing and follows a consistent schema, it is straightforward to identify when a company’s functional profile changed and whether it remains a valid comparable for a specific audit year.
4. Longitudinal Consistency: The Core Analytical Attribute
The central design principle is year-over-year comparability. Raw 10-K filings vary substantially in length, structure, and level of detail, a company that filed a 40-page Business section in one year may file 90 pages the next following an acquisition, with no structural way to isolate what actually changed. The standardized descriptions compress each year into the same schema, making deltas visible without manual review of the underlying documents.
The table below shows selected fragments from EdgarStat Annual Business Descriptions for Amazon.com, Inc. across a ten-year period, chosen to highlight analytically significant changes. Each row is an excerpt, not the full description.
| Year | EdgarStat Annual Business Description (Excerpt) |
| 2013–2015 | Two segments: North America (retail + AWS bundled) and International. Revenue from product sales, third-party seller fees, subscriptions, advertising, and electronic devices (Kindle, Fire, Echo). AWS reported inside North America segment. |
| 2016 | Segment structure unchanged: North America and International. AWS net sales of $7,880 million (2015) reported within North America. No separate AWS segment disclosed. |
| 2017 | Segment restructuring: AWS broken out as a standalone third segment for the first time. Three segments reported going forward: North America, International, and AWS. AWS 2016 net sales: $12,219 million. Acquisition of Twitch (2014) noted. |
| 2018 | Three-segment structure confirmed. Whole Foods Market acquired August 2017, physical store revenue now part of North America segment. AWS 2017 net sales: $17,459 million. |
| 2019 | Three segments. Ring Inc. and PillPack acquired in 2018. Advertising services increasingly prominent as a revenue line alongside AWS and retail. AWS 2018 net sales: $25,655 million. |
| 2022 | Three segments. MGM Holdings acquired 2022 for $6.1 billion; 1Life Healthcare (One Medical) acquired 2023 for $3.5 billion. Advertisers added explicitly as a primary customer set. AWS 2021 net sales: $62,202 million. |
| 2025 | Three segments. AWS 2024 net sales: $107,556 million, now the largest profit contributor. Advertising services established as a distinct revenue driver across North America and International segments. |
For a transfer pricing analyst, the single most consequential event in this timeline is not an acquisition, it is the 2016–2017 segment restructuring. Before 2017, AWS financials were embedded in the North America segment. A comparables panel built on Amazon’s 2015 or 2016 financials is working with a blended retail-and-cloud number, not a clean retail or cloud benchmark. After 2017, AWS is reported separately, but the company now includes physical retail (Whole Foods) in its North America segment. Neither period is the same business as the one before it.
5. Structural Break Detection and Comparables Rejection
One of the most common documentation deficiencies in TNMM benchmarking is the failure to identify and exclude years in which a comparable underwent a structural break. Mergers, acquisitions, and divestitures distort the profit level indicators used as benchmarks: an acquisition inflates assets and can suppress return on assets in the transition year; a divestiture removes an entire revenue stream and can produce artificially high or low operating margins. Segment reorganizations are subtler still: revenue may remain unchanged at the consolidated level while the underlying segment definitions, and therefore the comparability of any segment-level PLI, change entirely.
The Amazon example illustrates both types. The Whole Foods acquisition in August 2017 added a material physical retail operation to the North America segment in the same year AWS was first reported separately. An analyst using Amazon’s North America segment operating margin as a benchmark for a tested party engaged in online retail distribution would need to identify both events, assess their directional impact on the PLI, and either exclude the affected years or document why the distortion is immaterial.
EdgarStat Annual Business Descriptions flag these events explicitly and consistently. The 2018 description notes the Whole Foods acquisition and its integration into the North America segment. The 2017 description marks the first year of the three-segment structure. An analyst reviewing a ten-company comparables set over five years, fifty company-year observations, can identify contaminated observations systematically rather than relying on footnote review of individual filings.
Two further examples from the EdgarStat dataset illustrate different types of structural breaks that the descriptions capture.
Abbott Laboratories: spinoff removes the dominant business. Through 2012, Abbott operated five segments including a Proprietary Pharmaceutical Products segment that generated $18 billion in revenue, more than half the company’s total. In January 2013, Abbott completed the separation of that business as AbbVie Inc., an independent public company. By 2014, Abbott’s descriptions show four segments with no proprietary pharmaceuticals presence, and the Vascular Products segment as the largest device business. An analyst using Abbott as a comparable for a pharmaceutical entity in 2012 and again in 2014 is benchmarking against two structurally different companies. The consolidated financials show continuity in the Abbott ticker; the descriptions show the break immediately.
Hewlett Packard Enterprise: serial divestitures over multiple years. HPE was incorporated on November 1, 2015 following the split of the original Hewlett-Packard Company. Its first full-year filing shows five segments including Enterprise Services ($19.8 billion) and Software ($3.6 billion). By April 2017, Enterprise Services had been spun off into DXC Technology; by September 2017, Software had been merged into Micro Focus International. Together these two divestitures removed approximately $22 billion in annual revenue from the company’s scope in a single fiscal year. The 2018 descriptions show four segments organized around Hybrid IT, a fundamentally different functional profile from 2016. By 2020, segment reporting had been reorganized again into seven units following the Cray acquisition, before consolidating once more in subsequent years. Each of these transitions is visible in the year-specific descriptions; none of them would be detected by a static SIC-code screen.
This is particularly relevant under the OECD’s guidance on comparability adjustments and the requirement to document that selected comparables performed comparable functions, engage comparable risks, and used comparable assets in the years used for benchmarking.
6. Functional Analysis Documentation
Beyond comparables selection, EdgarStat Annual Business Descriptions support the functional analysis documentation requirement directly. Transfer pricing documentation must characterize the functions performed, assets employed, and risks assumed by both the tested party and the comparables. For the comparables to the tested party, that characterization is typically supported by brief descriptions drawn from publicly available information; in practice, from the Business section of the 10-K.
The standardized descriptions provide that information in a format that is already organized around the attributes most relevant to functional analysis: business segments, revenue drivers, customer and supplier relationships, geographic presence, and corporate structure. For a practitioner preparing documentation across multiple jurisdictions and multiple audit years, having this material pre-extracted and indexed by year reduces a significant portion of the manual research burden.
7. Practical Applications in Transfer Pricing
The table below summarizes where EdgarStat Annual Business Descriptions fit into standard transfer pricing workflows:
| Workflow Stage | How Annual Business Descriptions Help |
| Initial comparables screening | Functional profile verification before financial filters are applied |
| Multi-year panel construction | Year-specific acceptance or rejection based on structural break detection |
| Functional analysis write-up | Pre-extracted, standardized attributes for documentation |
| Audit defense | Year-by-year evidence that comparables were functionally similar in the benchmark period |
| Comparability adjustments | Identification of differences (e.g., customer concentration, geographic mix) requiring adjustment |
8. Conclusion
Profit indicators are the arithmetic of transfer pricing benchmarking. Functional analysis is the logic that makes those profit indicators meaningful, the justification for why a particular company’s operating profit margin in a particular year is a relevant reference point for the tested party’s compensation.
EdgarStat Annual Business Descriptions are designed to reduce the time it takes to get that logical layer in place. By standardizing the functional profile of each SEC-filing company year by year, they make it easier to identify when a comparable’s business changed, flag acquisitions and restructurings that may contaminate a benchmark year, and support the narrative documentation that transfer pricing regulations require.
They do not replace the analyst’s judgment. A description that notes an acquisition still requires the analyst to assess whether the acquisition is material enough to affect the selected profit indicator, whether a comparability adjustment is warranted, and whether the company should be retained or rejected for that year. Nor do they substitute for reading the underlying filing when a specific fact needs to be verified or a nuanced comparability question needs to be resolved.
What they offer is a more efficient and defensible starting point, one that reduces the manual effort of tracking how dozens of companies evolved across a multi-year benchmark period, so that the analyst’s time can be spent on the judgments that actually require it.