Back to Articles|Published on November 8, 2026|28 min read

Docket Article

Government Contractor Agency Concentration in USAspending

How to measure government contractor agency concentration with USAspending in 2026: awarding vs funding agency, net obligations after de-obligations, the 90-day DoD lag, and a fictional worked example.

By Docket Research

Government Contractor Agency Concentration in USAspending

Summary

  1. 01USAspending obligations describe federal prime-award exposure within a frozen entity perimeter and action-date window. Company-wide federal revenue dependence requires a compatible recognized-revenue denominator.
  2. 02Awarding agencies administer awards; funding labels identify the largest funder rather than the complete funding mix. Allocate only reconciled account evidence and keep unsupported amounts in an unknown bucket.
  3. 03Keep transaction-level modifications and reversals. Report net shares only when total net obligations are positive and every group is non-negative; show positive-obligation shares as a labelled sensitivity.
  4. 04Use a common mature window and label recent quarters provisional. Publication delays and data vintage can distort comparisons, and a few weeks of a fiscal year should not be annualized.
  5. 05The hypothetical example has complete administering concentration but multiple confirmed funders and an unknown allocation. Agency diversity, funding exposure and award concentration answer different questions.
Inside this article
  1. 01Executive Summary
  2. 02Introduction and Background
  3. 03Defining the Claim and the Denominator
  4. 04Agency Roles, Funding Allocation and Vehicle Structure
  5. 05Building the Extraction: Windows, Award Types, Keys and Deduplication
  6. 06Formula Sheet: Gross, Reversals, Net and Concentration Measures
  7. 07Interpretation Traps, Timing and Data Limits
  8. 08Data Analysis and Evidence
  9. 09Case Study: Fictional Worked Example (Hypothetical Example)
  10. 10Implications and Future Directions
  11. 11Frequently Asked Questions (FAQs)
  12. 12Conclusion

Executive Summary

Government contractor agency concentration can be measured from public USAspending data, but the result answers a narrower question than most screens assume. Obligation shares describe reported federal prime-award exposure inside a frozen entity perimeter and a declared action-date window. They are not shares of company revenue and not a count of independent customers. The awarding agency creates and administers an award, while the funding agency provided the preponderance of the money, and the official guide says a multi-funded transaction lists only its largest funder [1].

The worksheet separates four claims that a single "customer concentration" label blurs: agency diversity, federal program exposure, prime contract concentration and share of company revenue. Each gets its own denominator. The arithmetic is plain: gross positive obligations P, absolute de-obligations D and net obligations N = P - D, built from transaction-level action obligations that can include de-obligations [2]. A net group share is reported only when total N is positive and every group's N is non-negative. Otherwise the worksheet reports amounts and an exception. Herfindahl-Hirschman Index (HHI) figures are not offered as a stable score, because the index is the sum of squared market shares [3].

Timing is the second hazard. Federal fiscal year 2026 (FY2026) ended on 30 September 2026, and as of 8 November 2026 only 39 days of FY2027 exist. Department of Defense (DoD) contract data are published on a 90-day delay [4] [5], and the data vintage behind the Government Accountability Office (GAO) FY2025 total of $793 billion was April 2026 [6]. A DoD awarding-agency pull for award types A to D on 8 November 2026 showed $449.8 billion for FY2026 against $541.8 billion for FY2025 [7] [8]. The DoD awarding-view totals are not comparable with GAO's presentation, so they are not a scale benchmark against the $793 billion figure, and the lower FY2026 total should not be read as a spending decline. The source does not explain the gap, but it shows why comparisons need a common mature window and why recent quarters are provisional.

The final step is an explicit unknown. Where Federal Account Funding evidence is missing, the allocation stays in an unknown bucket instead of an equal split. Where no independently sourced, same-period recognized-revenue denominator exists, company-wide federal dependence stays unresolved, and accounting rules treat entities under common control as a single customer [9]. The output is a public-data research status: accepted federal-exposure observations plus an open-questions register for management financials, contract documents and specialist review. It is not a valuation, a legal conclusion or a recommendation to acquire any company.

90 daysDelay before DoD contract action reports become publicly viewable
51 percentAgencies confirming completion of procurement data quality reports for FY2023

Introduction and Background

Private equity origination teams screening government-services candidates meet the phrase "diversified across the federal government" repeatedly. It hides four different claims: many agencies served, one program relied on, one large prime contract held, or a large federal share of revenue. Advisers treat the topic as a transaction issue. One firm lists customer and contract concentration, measured by the number of customers, contracts, contract vehicles and their relative significance, among the value drivers buyers weigh in a government-contractor exit [10], and flags "customer concentration without proper context" as a buyer red flag [10].

What these pieces do not supply is a reproducible public-data calculation. Another adviser says the contract waterfall and pipeline are key to understanding concentration [11]; a third notes that large customers split into subagencies with their own procurement offices and decision makers [12]. A vendor library publishes contractors with at least $1 million in cumulative net contract obligations and a federal customer count meant to show whether a vendor depends on a narrow buyer set [13]. That is a vendor claim, and this report neither reuses nor validates its totals. Its framing still shows the gap: a customer count does not reveal whose money funded the work, which program drove it, or how it relates to company revenue.

The practical question is narrower. Given a frozen legal-entity perimeter, what do reported prime contract obligations say about agency exposure, and when is the evidence too thin to continue screening? This report builds a worksheet for that question. It assumes the entity gate has been passed: Unique Entity Identifier (UEI) records and related recipient names matched, and included legal entities frozen. The worksheet does not rewrite that matching. It extends it with denominator discipline.

Scope is limited by design. The Congressional Research Service (CRS) tells researchers that search results may be incomplete or contain inaccuracies [14]. Findings are therefore research statuses: accepted, unresolved or unknown.

Defining the Claim and the Denominator

Four claims, four denominators

Name the claim before touching a download. Agency diversity asks how reported prime obligations spread across awarding agencies. Federal program exposure asks how much depends on one program, funding source or vehicle. Prime contract concentration asks how much sits in the largest award. Share of company revenue asks what fraction of recognized revenue is federal, which no obligation table answers alone.

The first three can be built from public obligation records inside the frozen scope. USAspending defines an obligation as a promise made by the government to spend funds and an outlay as money actually paid out [1] [15]. An obligation is a reported commitment, not recognized revenue, billings or profit.

Why revenue dependence stays unknown

Accounting practice reinforces the separation. The customer disclosure rule in Accounting Standards Codification (ASC) 280, as quoted in public Securities and Exchange Commission staff correspondence, treats a group of entities under common control as a single customer [9], and the same correspondence asks for federal government revenue in the aggregate across departments and agencies [9]. It also refers to a 10 percent revenue level [9]. The accounting view of "the federal government" can thus be one customer even when USAspending shows many awarding agencies. Agency diversity does not establish low federal dependence.

Backlog definitions differ as well. Remaining-performance-obligation disclosure covers the noncancellable contract term [16] and excludes optional orders above a contractual minimum even if such orders are expected [16], and one public contractor's annual report excludes potential orders under indefinite-delivery, indefinite-quantity (IDIQ) contracts and unexercised options from backlog [17]. Because definitions are company-specific, the worksheet never divides obligation totals by vendor-estimated company revenue. Company-wide federal dependence stays unknown unless an independently sourced, same-period recognized-revenue figure with a compatible federal measure exists.

The perimeter gate

The data dictionary carries separate fields for the recipient UEI and the ultimate parent UEI [18], and the official guide notes parent-recipient fields can group related recipients [1]. The General Services Administration (GSA) moved its systems from the DUNS number to the SAM-generated UEI in 2022 [19], so older records need a documented bridge. Federal Acquisition Regulation (FAR) 52.204-17 defines an immediate owner as the entity with direct control of the offeror [20], which helps describe a group but does not prove a subsidiary belongs in an accepted total.

Freeze three things before any arithmetic: the included legal entities, whether the screen covers one subsidiary or an independently corroborated group, and the list of uncertain entity edges. Uncertain edges stay outside accepted totals and go to the open-questions register.

Agency Roles, Funding Allocation and Vehicle Structure

Awarding versus funding agency

The official guide says the awarding agency creates and administers the award and that in most cases awarding and funding agency are the same [1]. The glossary adds that the funding agency provided the preponderance of the funds for an award or its individual transactions [2]. A program white paper says the awarding agency interacts with the recipient to make the award [21], and that in an assisted acquisition the funding agency has the funding and authority to carry out the activity [21]. The guide is dated 26 April 2023 [1]; the glossary and data dictionary were read on 8 November 2026 and carry no date of their own, so definitions were rechecked against the live dictionary.

The split exists because federal rules allow it. The FAR describes an assisted acquisition as a servicing agency awarding and administering a contract on a requesting agency's behalf [22], and Economy Act authority lets agencies obtain supplies or services from another agency [23]. GSA's FY2025 financial report states that the top five federal customers of its Acquisition Services Fund account for 86.2 percent of fund revenue [24]. One administering agency can serve many funders, and a small agency may require a larger federal entity to administer its award [25].

The largest-funder limitation

The guide states that a single prime award transaction may be funded by multiple agencies and that the agency funding the most is listed as the funding agency [1]. GAO's data-element table describes the funding agency the same way [26], and FAR 4.603 requires a Program/Funding Agency and Office code [27]. No cited source promises a complete funder list on the transaction row. A funding label need not disclose every funder.

To see all funders, the guide points to the Funding Agency column in the Federal Account Funding table on the award profile [1]. The API contract for that table lists federal account financial data for the award [28], and a companion endpoint returns counts of awarding agencies, funding agencies and federal accounts [29], which makes a cheap multi-funder flag.

Allocation evidence has limits

Federal Account Funding is File C data. The download readme says it is only available from January 2017 onward and that a subset is not definitively linkable to a single federal award [30]. A USAspending methodology page defines unlinked awards as those lacking a shared award ID that allows two systems to match records [31], and GAO's DATA Act review found that some agencies omitted the file linking budget and award information [32].

The rule follows. Allocate a transaction to funders only where Federal Account Funding evidence covers the same period and reconciles to the transaction. Use a completeness column with three states:

  • Supported: account lines cover the same period and reconcile to the transaction.
  • Partial: some account lines exist; the reconciled part is allocated and the remainder goes to unknown.
  • Unsupported: no usable lines; the whole amount goes to the unknown bucket.

Never spread a transaction equally across visible funding agencies, because the label names the largest funder and says nothing about proportions.

Hierarchy levels

Keep agency code, subagency and office as separate levels. The dictionary defines the awarding sub-agency code as a level-2 identifier [18], and two federal hierarchies relate agencies to subcomponents, one for account data and one for award data [4]. A count of procurement offices is not a count of independent customers.

Table 1 consolidates the field-level choices, with permissible and prohibited claims.

Field choicePermissible claimProhibited claim
Transaction vs award summaryWindowed sums of transaction-level action obligations; a prime award summary is a roll-up of related transactions [1].Summing award-level cumulative fields across rows; the dictionary describes total dollars obligated as a system-generated sum over the award [18].
Obligation vs outlayObligations as reported commitments in the window.Calling obligations cash or revenue; outlay data before the mandatory period may be incomplete [1].
Awarding vs funding agencyAwarding view as administering exposure; funding label as largest funder only.Reading the label as the full funding mix.
Top tier vs subagency vs officeSeparate views at each level.Treating office count as independent customer count.
Prime vs subawardPrime panel only; subawards are modeled separately [33].Adding subaward amounts to prime obligations; prime recipients submit subaward data [15].
Current vs lifetime vs potentialCurrent obligated amount [2]; potential amount only as a ceiling.Using lifetime or ceiling values in a fiscal-year denominator or as funded backlog.
Funding allocationAllocation only where Federal Account Funding reconciles.Equal splits; filling gaps with the largest-funder label.

Every field has a defensible use and a tempting misuse. The misuse is usually quicker to execute, which is why the prohibited column matters more during review.

Figure 01
Awarding and funding agencies answer different questions
Awarding agencyAdministering exposure
  • Creates and administers the award.
  • One administering agency can serve many funders.
Funding agencyLargest funder on the row
  • A funding label need not disclose every funder.
  • Allocate only where Federal Account Funding evidence covers the same period and reconciles to the transaction.

Keep agency code, subagency and office as separate levels. A count of procurement offices is not a count of independent customers.

A defensible agency-concentration screen begins with a claim and a denominator, not a download.

Building the Extraction: Windows, Award Types, Keys and Deduplication

Declare the window, types and snapshot

Start with an action-date window. The transaction search contract applies its time period to the transaction's action date [34], defined in the dictionary as the date the action was issued or signed, or a binding agreement reached [18]. The award-level filter differs: for award searches the action date is that of the latest transaction on the award [35], and CRS notes that award search results include awards active during the period regardless of start date [15]. An award-level filter can therefore pull lifetime amounts into a fiscal-year denominator. The worksheet filters transaction rows by action date.

Declare the scope as a short checklist:

  • Window: transaction action dates only, with start and end stated.
  • Award types: prime contract types A to D.
  • Snapshot: download date, data-as-of marker and dictionary version.
  • Completeness flags: one per agency, marking provisional periods.
  • Panels: subcontract work and grants kept separate.

Declare prime contract award types A to D. The API contract searches contracts with type codes A, B, C or D [33]. Parent indefinite delivery vehicle (IDV) records sit in a separate code family, so A to D avoids mixing a vehicle with the orders placed under it. Record the snapshot date, a data-as-of marker (the awards last-updated endpoint returned 11/06/2026 on 8 November 2026 [36]) and completeness flags per agency.

Keep subcontract work and grants in separate panels. Prime awards are distinct from sub-awards [33], and USAspending states that duplication of subaward records is common [37].

Pick the right download

CRS notes that results can be downloaded at the award or transaction level [15]. The per-award readme describes a file of transaction-level data for all modifications, including the base award [30]. Monthly archive files are uploaded by the 15th of each month [38], and the archive page directs users to its Data As Of column [38]. Delta files carry a correction indicator marking records modified, deleted or added [38], so a delta is applied to a base file and never summed on top of it.

Deduplicate on the documented key

The data dictionary names the unique transaction identifier: the contract transaction unique key, a derived, system-generated database key identifying each contract transaction record [18]. It concatenates the agency identifier, referenced IDV agency identifier, procurement instrument identifier (PIID), modification number, parent award identifier and a transaction number [18]. For IDV records only three components are used [18], and the transaction number is a tie breaker for legal, unique transactions that would otherwise share a key [18]. The award-level key is different: it identifies the award, not a transaction within it [18].

Three consequences follow:

  • Keep every modification row. One award can have many legitimate modifications (the FAR defines a modification as any written change in the terms of a contract [22]), so rows are never reduced to one per award.
  • Do not treat the API internal identifier as a transaction key. The transaction search sample repeats one internal identifier across several modifications of an award [34].
  • Validate collisions instead of dropping rows. Compare every field on colliding rows; identical rows from overlapping files are duplicates, while rows that differ in obligation or date get an exception flag.

Check the schema version

Treasury's Government-wide Spending Data Model (GSDM) page lists release 1.3 dated 6 November 2026 and release 1.2.1 dated 1 May 2026 [39], calls itself the authoritative source for terms, definitions, formats and structures [39], and notes the model reflects the migration of FPDS to SAM.gov [39]. Planning notes that treat 1.2.1 as current are out of date. Record the dictionary version beside the snapshot date. The GSDM is a schema reference, not a company-revenue dataset.

Formula Sheet: Gross, Reversals, Net and Concentration Measures

Gross, reversal and net

The glossary defines a de-obligation as a cancellation or downward adjustment of previously obligated funds [2], and the guide says negative obligations occur when agencies correct errors or reflect new information [1]. GAO's table likewise describes the action obligation as the amount obligated or deobligated [26]. For a frozen entity and period scope, let a_i be the action obligation on transaction i:

  • Gross positive obligations: P = sum of max(a_i, 0).
  • Absolute de-obligations: D = sum of abs(min(a_i, 0)).
  • Net obligations: N = P - D, which equals the plain sum of a_i.

Reconcile N to the net-obligation view the API or interface returns for the same filters, and leave original rows untouched. The spending-by-category contract says its spending level determines the data source for totals [40], so record which level the comparison view used. A mismatch becomes an exception row, not an edit.

Net shares and when they are valid

The net group share is s_g = N_g / N, reported only when total N is positive and every relevant N_g is non-negative. Top-one and top-three shares use one declared grouping (awarding top-tier agency, awarding subagency, funding label with an unknown bucket, or award). If the conditions fail, report amounts and the exception. Do not clip negative groups to zero: clipping changes the denominator and yields shares that no longer sum to a meaningful whole.

A positive-obligation share P_g / P is allowed as a labelled sensitivity measure. It ignores reversals, so it can overstate a group that later had large de-obligations, and it cannot replace the primary net view. Show both for the same grouping and explain any gap.

Why the Herfindahl index is not a stable score here

The FTC and Department of Justice merger guidelines define the HHI as the sum of the squares of the market shares [3], note that it reaches 10,000 in a single-firm market [3], and treat an HHI above 1,800 as highly concentrated [3]. Those thresholds belong to antitrust market definitions, not a contractor's customer mix, so using them as a trigger would be an invention. Squared shares also presuppose non-negative, fully allocated groups. With a negative group, an unknown bucket or a doubtful perimeter, an HHI looks precise and is not. The worksheet reports it only for fully allocated, non-negative groupings and labels it descriptive.

Award-level concentration

Top-award share uses the same N basis: the net obligations of the largest award, summed across its modification rows in the window, divided by total N. Record both parent PIID and order PIID so the vehicle view and award view do not double count. Vehicle position matters to advisers, who separate backlog already awarded with a defined payment schedule from recompete positions [12] and note that reliance on a single contract or small collection of contracts is riskier [12].

Figure 02
Build and validate the concentration worksheet
  1. 01Freeze the entity perimeter

    Freeze included legal entities and the subsidiary or group scope. Keep uncertain entity edges outside accepted totals and record them as open questions.

  2. 02Declare the extraction scope

    State transaction action dates, prime contract types A to D, the snapshot and data-as-of marker, and completeness flags for provisional periods.

  3. 03Deduplicate transaction rows

    Deduplicate on the contract transaction unique key, then compute gross positive obligations, absolute de-obligations and net obligations by awarding agency.

  4. 04Reconcile gross, reversals and net

    Compute P, D and N = P - D. Reconcile net obligations to the API or interface using the same filters; preserve original rows and record mismatches as exceptions.

  5. 05Allocate supported funding

    Allocate reconciled account lines. For partial evidence, put the remainder in unknown; with no usable lines, put the whole amount there.

  6. 06Apply the net-share rule

    Report net shares only when total net obligations are positive and every relevant group is non-negative. Otherwise report amounts and the exception.

Interpretation Traps, Timing and Data Limits

Five traps

  • Agency count is not balanced exposure. Ten awarding agencies can still mean most obligations come from one.
  • Procurement office count is not an independent customer count. Offices within one agency may share a funding line and decision chain [12].
  • Ceilings and options are not funded backlog. Potential award amount assumes all options are exercised [2]. An indefinite-quantity contract obligates the Government only to a stated minimum order [41], incrementally funded contracts obligate funds only to the amount allotted [42], and advisers review funded value and unfunded backlog separately [43].
  • Reported commitments are not recognized revenue or profit. One public contractor defines funded backlog as firm orders for which funding is contractually authorized and appropriated [17]; that definition cannot be copied onto USAspending data.
  • No observed work is not evidence of no federal business. Some smaller executive branch agencies and the legislative and judicial branches are not required to report [4], and GAO's March 2026 statement to Congress says agencies do not consistently report some spending [44].

Data limits that bound conclusions

Cadence statements differ. FAR 4.604 requires a contract action report in FPDS within three business days after award, and 30 days for urgent or emergency authorities [27]. USAspending says contract data appear within five days [37] and that the site refreshes daily after the nightly pipeline [4], while a public finance association describes a 2-week cycle [25] and CRS says agencies report within 30 days [15]. They describe different systems, but a user cannot assume one cadence. These cadence differences describe the datasets, not any contractor.

Timing at the fiscal-year boundary

FY2026 ended on 30 September 2026, but the last quarter may be incomplete. DoD procedures withhold contract action reports from public view until 90 days after the date signed [45], CRS repeats the roughly 90-day lag [15], and the archive page applies it to displayed contract award data [38]. On 8 November 2026 the DoD awarding-agency pull showed $449.8 billion for FY2026 and $541.8 billion for FY2025 [7] [8], and the latest account submission period the API listed ended on 31 August 2026 [46]. The source does not say why FY2026 is lower, so the safe reading is that the window is not mature.

The working rule: choose a common mature action-date window for all candidates. For a DoD-heavy screen, a window ending 30 June 2026 leaves more than four months of settling time at the 8 November snapshot. Label July to September 2026 provisional. Do not annualize the 39 days of FY2027 observed so far. For candidates with little DoD exposure the window can be later, but state which agencies drove the choice.

Data Analysis and Evidence

Table 2 lists quantitative anchors that set the scale of the data, the sensitivity of totals to vintage and definition, and documented quality limits.

AnchorValue and as-of dateUse in the worksheet
Federal contract obligations, FY2025About $793 billion, GAO analysis of FPDS data as of April 2026 [6]Scale check; shows a vintage lag of roughly seven months behind year end.
Federal contract obligations, FY2024About $755 billion, FPDS data as of June 2025 [47]; GAO's data-quality report cites the same scale [48]Totals shift slightly by vintage and presentation.
DoD and civilian split, FY2024$445.1 billion DoD and $310 billion civilian [49]DoD is about 58.9 percent of the total by our arithmetic.
Competition rates, FY2025Civilian agencies 89 percent, Defense about 52 percent [6]Sector context for recompete dynamics.
DoD awarding view, types A to D$541.8 billion FY2025; $449.8 billion FY2026 as pulled 8 November 2026 [8] [7]Window maturity; awarding view, not comparable to GAO's presentation, so do not use it as a scale benchmark against the GAO total. The source does not explain the FY2026 versus FY2025 gap; treat FY2026 as provisional given the 90-day DoD lag and the partly elapsed data window, and do not read it as a spending decline.
VA awarding view, types A to D, FY2025$78.3 billion [50]Example civilian awarding-agency figure; same vintage caveat.
Government-wide vendor concentration (dated)Top 10 vendors held over 27 percent of FY2015 obligations [51]Shows concentration statements must name their basis.
Procurement data quality36 agencies, or 51 percent, confirmed completing quality reports for FY2023 [48]Completeness caveat.

Two points emerge. Totals depend on definition and vintage: GAO's figures use FPDS data frozen at a stated month, the live API figures move with each pull, and awarding-agency views will not match funding-agency views. And the federal market is itself concentrated, with DoD holding a majority share of FY2024 contract dollars, so a DoD-heavy candidate inherits both the sector's weight and its publication lag. The FY2015 vendor figure is dated and describes vendors, not agencies.

Competition data are not part of the arithmetic, but they inform the follow-up. One adviser recommends separate sensitivity analysis for backlog tied to affected agencies or vehicles rather than a blended haircut [52]. The worksheet supplies the agency and vehicle cuts such an analysis needs and leaves the haircut to specialists.

Never spread a transaction equally across visible funding agencies, because the label names the largest funder and says nothing about proportions.

Case Study: Fictional Worked Example (Hypothetical Example)

This is a hypothetical example with invented figures and no real contractor. It uses one frozen entity, Example Services LLC (fictional), a window of 1 October 2025 to 30 June 2026 action dates, award types A to D and a snapshot of 8 November 2026. One administering agency (Agency A, with subagencies A1 and A2), two funders (Agency B and Agency C), one de-obligation and one transaction with an unsupported allocation. Figures are in millions of dollars. Table 3 lists six transactions.

TxnAwardAwarding agency / subagencyFunding labelAllocation statusAction obligation ($M)
T1Award 1A / A1BSupported5.0
T2Award 2A / A1BSupported3.0
T3Award 2A / A1CSupported2.5
T4Award 3A / A2CSupported1.5
T5Award 1A / A1BSupported (de-obligation)-1.0
T6Award 3A / A2BUnsupported (multi-funder flag, no matching account lines)2.0

Each row has its own transaction key. Award 1 has two rows because one is a modification; collapsing it to one row per award would hide the reversal. T6's label lists only the largest funder, so its two million dollars cannot be assigned to B with confidence. The arithmetic:

  • Gross and net: P = 5.0 + 3.0 + 2.5 + 1.5 + 2.0 = 14.0. D = 1.0. N = 13.0, positive and equal to the plain sum.
  • Awarding view: Agency A holds N = 13.0, a 100 percent share. Administering concentration is complete by construction and says nothing about funding diversity.
  • Subagency view: A1 has N = 9.5 (73.1 percent) and A2 has N = 3.5 (26.9 percent). A1's positive share is 10.5 of 14.0 (75.0 percent), 1.9 points above its net share.
  • Funding view with unknown bucket: confirmed B has N = 7.0 (53.8 percent), C has 4.0 (30.8 percent) and unknown holds 2.0 (15.4 percent). B's share is bounded between 53.8 and 69.2 percent; the upper bound assumes all of T6 is B's.
  • Positive-obligation sensitivity: B is 8.0 of 14.0 (57.1 percent), C is 4.0 (28.6 percent), unknown is 2.0 (14.3 percent). B reads higher than on the net view because the de-obligation lowered B's net total. Net describes exposure after reversals; positive describes gross activity.
  • Award view: Award 1 N = 4.0 (30.8 percent), Award 2 N = 5.5 (42.3 percent), Award 3 N = 3.5 (26.9 percent). Top-award net share is 42.3 percent, against 39.3 percent on a positive basis.
  • Exception demonstration: if T5 were -9.0, confirmed B would be N = 5.0 + 3.0 - 9.0 = -1.0. Total N stays positive at 5.0, but the net share rule fails, so the worksheet reports group amounts and the exception, triggers review, and presents neither shares nor an HHI.

The accepted observations are: administering concentration with Agency A is complete; subagency A1 carries a majority of net obligations; confirmed funding splits between B and C with 15.4 percent unknown; no award exceeds 42.3 percent of net obligations. The open-questions register records what public data cannot settle:

  • T6 funding allocation: which agencies funded the two million dollars.
  • T5 reason: why the reversal occurred.
  • Revenue denominator: recognized revenue for the same period, from management financials.
  • Award structure: whether awards are ceiling-limited or option-bearing, from contract documents.
  • Perimeter: whether any affiliate belongs in the scope.

Company-wide federal dependence stays unknown, and no risk score is produced.

Figure 03
Hypothetical funding view with an unknown bucketMillions of dollars, net obligations
Source: Case Study: Fictional Worked Example (Hypothetical Example)

Implications and Future Directions

For origination teams, the consequence is a three-way status rather than pass or fail. The three statuses are:

  • Accepted: observations support continued screening (stable perimeter, mature window, non-negative groups, reconciled net view).
  • Unresolved: items need deeper diligence (negative groups, large unknown buckets, provisional quarters).
  • Unknown: items stay unknown until specialists see management financials (revenue dependence, subcontract work, funding mix beyond the label).

Publisher note (not evidence for any claim above): Docket describes itself as a platform that produces evidence-backed target research for private equity firms, and related guides are listed at docket.capital/research.

Several developments could change the worksheet. GSDM 1.3, released 6 November 2026, reflects the FPDS migration to SAM.gov and may alter field availability [39]. Because the FPDS migration may change field availability, schema stability should not be assumed. The 90-day rule is a DoD procedure rather than a USAspending feature [5], so a procedural change would change which windows are mature. Rechecking these items before each use is part of the method.

Accounting and legal questions remain outside scope. How funded versus unfunded contract value becomes recognized revenue [53], whether a contract can transfer in a transaction, and how a novation would work require specialist review. This report cites official definitions and has not been reviewed by a professional; readers should consult a qualified accountant or government-contracts attorney.

Frequently Asked Questions (FAQs)

How do you analyze government contractor revenue by agency?

Treat it as obligations by agency, not revenue. Freeze the entities, pick an action-date window, restrict to types A to D, deduplicate on the contract transaction unique key and compute P, D and N by awarding agency. Revenue by agency needs company records, because USAspending reports commitments [15].

What does a USAspending recipient profile show for awards by agency?

The worksheet uses the USAspending API and bulk downloads rather than recipient profile pages. The API serves the same cuts: the spending-by-category contract returns top results sorted by total amount [40]. Check that level, date type and award types match the declared scope.

How do you download prime award data by awarding agency?

Use the Custom Award Data Download page, documented in the API contract [54], or the monthly archive files. Download transaction-level contract files, filter by action date and awarding agency code, and record snapshot and data-as-of dates.

Is the Herfindahl index appropriate for agency concentration?

Only as a descriptive figure on fully allocated, non-negative groupings. Its 1,800 threshold is an antitrust market benchmark [3]. With negative groups or unknown allocations, report amounts and exceptions.

How should private equity teams screen a contractor for diversification?

Name the claim, freeze the perimeter, declare the window, build awarding and funding views, compute net and positive shares, and log open questions. Advisers recommend mapping customer and contract concentration with mitigation plans for top exposures [10] and weighing backlog, option years and recompete risk together [10].

What is govcon customer concentration risk?

It is dependence of revenue on few customers, contracts or vehicles. One firm's commercial diligence looks beyond contract face value to recompete exposure and customer concentration [55]. Public obligation data inform part of it; company revenue data are needed for the rest.

Why can one agency appear as the customer when others funded the work?

The awarding agency can administer awards for other agencies, and the transaction row lists only its largest funder [26]. Federal Account Funding shows other funders.

Conclusion

A defensible agency-concentration screen begins with a claim and a denominator, not a download. Public USAspending obligations support agency diversity, program exposure and award concentration inside a frozen perimeter and a declared mature window. They do not support a statement about the share of company revenue, because obligations are commitments and not recognized revenue. Awarding and funding agency are different roles, and a multi-funded transaction lists only its largest funder, so agency counts and funding labels mislead if read as complete.

The controls are simple and reproducible:

  • Net shares only when valid: total N positive and every group non-negative.
  • Labelled sensitivity: positive-obligation shares shown beside, never instead of, the net view.
  • Funding completeness: a column with an unknown bucket.
  • Common mature window: with provisional quarters labelled.
  • Open-questions register: for everything public data cannot settle.

Reversals are shown, not hidden. At the fiscal-year boundary the worksheet guards against two immediate errors: treating incomplete recent quarters as final, and annualizing a few weeks of FY2027.

The result is a research status telling a team whether to continue screening, escalate to deeper diligence, or hold the question as unknown. It remains a public-data screen. Management financials, contract documents and specialist review decide the rest.

External Sources (55)

Published by

Docket Research

Docket provides deal-origination research software and managed research for private equity teams, screening companies against a mandate and showing the evidence behind each finding.

Our platform · Evidence standards · Contact

Get your first 10 targets free

Editorial note

Researched and written by Docket Research with AI-assisted research tools; sources are cited inline and checked before publication. Report an error: https://docket.capital/evidence-standards#corrections.