A Quantitative Methodology for Classifying Federal Contractor Competitive Positioning
L. Auctor FedComp Index March 2026, v1.1
Abstract
Two contracting officers conducting market research on the same requirement today will produce different contractor sets. There is no mechanism for reconciliation. Sources sought responses depend on which firms respond. SAM.gov searches depend on which terms the officer selects. Industry day attendance depends on which firms are available. Each method is participation-dependent: the output is a function of who shows up, not of who competes. This is not a tool limitation. It is a structural property of registration-derived contractor intelligence.
The FedComp Index resolves this by constructing contractor sets from obligation data rather than registration data. The classification is deterministic: identical input data produces identical output. Any independent party querying the same USASpending.gov data for the same state and time window will produce the same classifications and the same Proximity Map outputs. The methodology is state-scoped and reproducible from public data alone. No proprietary or purchased data is used at any stage.
1. Introduction
The United States federal government obligates over $700 billion annually in contract awards to private firms. The transactional data underlying these obligations has been published through USASpending.gov [1] and machine-queryable since the Federal Funding Accountability and Transparency Act (FFATA) of 2006. Entity registration data is maintained by SAM.gov [2]. Market research under FAR Part 10 [6] does not use this data to construct contractor sets. It queries the registration population. The consequences are structural.
1.1 The Registration-Obligation Gap
In Nevada, 3,794 entities are registered in SAM.gov with federal contracting NAICS codes. Of these, 779 (20.5%) have won a federal base contract award in the trailing 5-year window. The remaining 79.5% are Registration-Inferred: present in the registration population but absent from the obligation population.
A contractor set derived from registration data is 79.5% noise. This cannot be reduced by better search terms or more careful filtering, because the noise is structural: registered entities that have never won a contract are indistinguishable from active award recipients in the registration data. A set derived from award data is Obligation-Verified, conditioned on demonstrated federal contract obligations. The FedComp Index queries the obligation population. The output addresses FAR 10.002(b)(1) [8], FAR 19.202-1 [9], FAR 7.105(b)(4) [10], and FAR 15.304 [11].
1.2 Three Structural Impossibilities
Registration-derived contractor intelligence has three properties that cannot be fixed by better tools, only by switching the data source:
Non-reproducibility. Two officers querying SAM.gov for the same requirement will produce different source lists depending on search terms, filters, and timing. The output is not auditable under FAR 33.103 [12] because a second officer cannot reproduce the first officer's contractor set.
Fragility blindness. A contractor with $12M concentrated in a single contract expiring in 6 months and a contractor with $12M distributed across 11 active contracts appear identical in any one-dimensional ranking. Revenue rankings cannot distinguish current position from durable position.
Concentration opacity. The distribution of federal contract dollars follows extreme concentration (Gini 0.955 in Nevada, exceeding U.S. household wealth inequality at 0.85). 7 contractors (1%) hold 63.0% of all obligated dollars. This concentration is unmeasurable from registration data because registration data contains no dollar values. The structural shape of the obligation landscape is invisible to the tools currently used to analyze it.
2. Vocabulary
Posture Class. An obligation type assigned by the intersection of a contractor's volume tier (total obligated dollars) and frequency tier (base contract count) over the trailing 5-year window.
Registration-Obligation Gap (ROG). The structural divergence between the set of firms registered under a procurement code and the set of firms that have been awarded contracts under that code. In the Nevada population, 79.5% of registered entities have zero obligation history.
Obligation-Verified. A contractor set derived from actual contract award data, conditioned on demonstrated federal obligations.
Registration-Inferred. A contractor set derived from SAM.gov registration data, reflecting self-reported capability without demonstrated obligation history.
Obligation Density. The ratio of total obligated dollars to base contract count. Measures revenue concentration per competitive win.
Class Migration. The movement of a contractor from one Posture Class to another across scoring periods.
Velocity. The rate and direction of change in a contractor's position on the obligation plane, computed by comparing early-period and recent-period award activity within the 5-year window.
Proximity Pressure. The net directional force on a contractor from its Proximity Map neighbors' velocity vectors, modulated by the local density of the procurement neighborhood.
3. Data Sources and Scope
3.1 USASpending.gov
The primary data source is the USASpending.gov Award Spending API [1]. The FedComp Index queries awards where the contractor's registered address is in the target state, capturing firms headquartered in the state regardless of where work is performed.
| Field | Use |
|---|---|
| Recipient UEI | Primary entity identifier |
| Recipient Name | Entity identification |
| Contract Award Type | Base contract filtering for frequency axis (definitive contracts, purchase orders) |
| Total obligated amount | Volume axis input |
| Award start date | 5-year window filter; frequency counting |
| Award end date | Active contract determination |
| NAICS code | Proximity Map input |
| PSC code | Proximity Map input (weighted 2x relative to NAICS) |
| Awarding agency | Profile display |
Awards are filtered to a rolling 5-year window based on start date. The FedComp Index filters on start date rather than action date, ensuring that contract modifications do not create false recency.
3.2 Certification Enrichment
Certification data is sourced from the SBA Dynamic Small Business Search API, which queries SAM.gov [2] registration records. This data enriches contractor records but does not gate inclusion in the index. A contractor with USASpending award records is classified regardless of whether a matching SBA record exists. Certifications (8(a), HUBZone, WOSB, SDVOSB, VOSB) are recorded for display but do not factor into the classification.
3.3 Scope
The FedComp Index is scoped by state. The entity population consists of all contractors with awards in the target state who have at least one base contract (definitive contract or purchase order) within the 5-year window. The population is derived from award data, not registration data. A contractor that has won awards but is not currently registered in SAM.gov is still classified.
The initial implementation covers Nevada.
3.4 Data Pipeline
The scoring pipeline runs on a monthly cycle. USASpending.gov is queried via its public API for all contract awards in the target state within the trailing 5-year window. Records are deduplicated by UEI and aggregated to contractor-level totals. SBA certification data is enriched where available. NAICS and PSC codes are extracted from award records for Proximity Map computation.
The pipeline produces a scored dataset containing every contractor's two-axis position, Posture Class assignment, Obligation Density, certification flags, and top-6 Proximity Map neighbors. Pre-scored datasets are published to HuggingFace and Kaggle on each monthly cycle. The Python implementation is installable via pip install fedcomp-index. The live index is hosted at fedcompindex.org.
All intermediate data is reproducible from the public API. No manual curation, filtering, or judgment enters the pipeline at any stage.
4. Classification
4.1 Two-Axis Rationale
The FedComp Index classifies contractors by their position on a two-dimensional plane: total obligated dollars and base contract count. Two numbers. The entire classification reduces to whether each number is above or below a fixed threshold. There is no weighted sum, no composite score, no normalization.
Volume measures total obligated dollars from all contract types including delivery orders, because delivery order work is real economic activity. Frequency measures base contracts only, because a delivery order is not a new competitive win. This separation is the core design: volume captures what a contractor does, frequency captures how they got there.
A contractor winning 200 contracts at $500,000 each and a contractor winning 2 contracts at $50,000,000 each have identical total dollars. Any one-dimensional ranking conflates them. The two-axis classification separates them because they represent fundamentally different obligation structures: one has distributed revenue across many wins, the other has concentrated revenue in few.
Both axes use base-10 logarithmic scaling. Federal contract dollars are heavily right-skewed; logarithmic scaling prevents large primes from compressing the entire small business population into the bottom of the range.
| Dollar amount | log10 value |
|---|---|
| $10,000 | 4.0 |
| $100,000 | 5.0 |
| $1,000,000 | 6.0 |
| $5,000,000 | 6.7 |
| $10,000,000 | 7.0 |
| $100,000,000 | 8.0 |
| $1,000,000,000 | 9.0 |
A base contract is a definitive contract or purchase order, excluding delivery orders, BPA calls, and modifications. A contractor with one IDIQ vehicle generating 500 delivery orders has won one contract, not 500. A contractor with one BPA generating 200 calls has won one contract, not 200. Both delivery order and BPA call dollars count toward volume, because they represent real obligated work. The base contract filter applies to frequency only.
4.2 Posture Class Assignment
Boundaries are fixed at $5,000,000 in total obligated dollars and 3 distinct base contracts over the 5-year window. These correspond to $1,000,000 in average annual obligation and three independent competitive wins. Thresholds are inclusive: a contractor with exactly $5,000,000 or exactly 3 base contracts meets the high tier.
| Class | Volume | Frequency | Structural property |
|---|---|---|---|
| 1 | High | High | Distributed revenue, repeatable wins, durable position |
| 2 | High | Low | Concentrated revenue, exposed to contract expiration |
| 3 | Low | High | Many small wins, demonstrated repeatability, growth trajectory |
| 4 | Low | Low | Limited activity, entry position or expired pipeline |
4.3 Empirical Findings
The classification was applied to the Nevada population of 779 Obligation-Verified contractors.
Registration-Obligation Gap. Of 3,794 SAM-registered entities with federal contracting NAICS codes, 779 (20.5%) hold at least one base contract award in the 5-year window. The remaining 79.5% are Registration-Inferred only.
Contractor Comparison.
| Contractor | 5yr Base Volume | Base Contracts | Obligation Density | Class |
|---|---|---|---|---|
| Trax International Corp | $167.1M | 3 | $55.7M/contract | 1 |
| UBC National Job Corps | $59.3M | 1 | $59.3M/contract | 2 |
| Spacecraft Components Corp | $3.2M | 484 | $6.6K/contract | 3 |
| Sarah Hummel DVM LLC | $2.4M | 1 | $2.4M/contract | 4 |
Trax at $167M across 3 base contracts and UBC at $59M on a single contract represent completely different structural positions despite both being high-volume. Trax won 3 independent competitive actions (Class 1); UBC holds one (Class 2). Spacecraft Components wins frequently at small scale with 484 base contracts (Class 3); Sarah Hummel has one contract (Class 4).
Figure 1. Nevada contractor population (n=779) plotted in log-transformed obligation space. Each point represents one Obligation-Verified contractor. Dashed lines indicate classification thresholds (tau_v = $5M, tau_f = 3 contracts).
Obligation Density. The ratio of total obligated dollars to base contract count spans 160,000x across the Nevada population, over five orders of magnitude:
| Statistic | Value |
|---|---|
| Minimum | $492/contract |
| 25th percentile | $26,000/contract |
| Median | $62,000/contract |
| 75th percentile | $239,000/contract |
| Maximum | $78,900,000/contract |
Class Distribution. The Nevada population distributes as follows:
| Class | Count | % | Median Volume | Median Contracts | Median OD |
|---|---|---|---|---|---|
| Class 1 | 62 | 8.0% | $17.3M | 11 | $1.7M/contract |
| Class 2 | 18 | 2.3% | $15.2M | 1 | $11.5M/contract |
| Class 3 | 219 | 28.1% | $400K | 5 | $56K/contract |
| Class 4 | 480 | 61.6% | $100K | 1 | $48K/contract |
The distribution forms a pyramid: 8.0% systematic winners, 2.3% concentrated risk, 28.1% growth pipeline, 61.6% entry level. The four classes are empirically distinct populations, not arbitrary quadrants. Class 1 and Class 2 have similar median volume ($17.3M vs $15.2M) but Class 2's median Obligation Density is 7x ($11.5M vs $1.7M per contract), confirming that the frequency axis separates structurally different obligation positions at the same dollar scale. Class 1 and Class 3 median Obligation Density differs by 30x.
Structural Fragility.
| Class | Base contracts expiring within 12 months |
|---|---|
| Class 1 | 11.8% |
| Class 2 | 25.0% |
| Class 3 | 4.9% |
| Class 4 | 13.8% |
Class 2 contractors carry 5x the near-term expiration exposure of Class 3 contractors. Three Class 2 contractors in the Nevada population have 100% of their base contracts expiring within 12 months, representing $24.5M in obligated volume in the top 2% by dollar ranking that is structurally exposed to contract termination without replacement pipeline.
Concentration. The distribution of total obligated dollars across the Nevada population is more concentrated than U.S. household wealth inequality.
| Population segment | Share of total obligated dollars |
|---|---|
| Top 1% (7 contractors) | 63.0% |
| Top 5% (38 contractors) | 88.6% |
| Top 10% (77 contractors) | 94.8% |
| Top 20% (155 contractors) | 98.1% |
The Gini coefficient for the Nevada contractor population is 0.955 (U.S. household wealth: 0.85, U.S. household income: 0.41). Class 1 constitutes 8.0% of the scored population and holds 86.7% of total obligated dollars. Class 4 constitutes 61.6% and holds 2.2%.
Figure 2. Lorenz curve for Nevada obligated dollar distribution. The shaded area between the curve and the equality line corresponds to the Gini coefficient of 0.955.
4.4 Sensitivity Analysis
The classification is stable under threshold variation. Sensitivity analysis on the Nevada population (779 contractors):
| Threshold variation | Unchanged |
|---|---|
| +/-20% | 83-88% |
| +/-50% | 81-82% |
| +100% (double both thresholds) | 77% |
Doubling both thresholds preserves 77% of all classifications. The diminishing decay rate indicates that the majority of contractors are classified by structural position rather than proximity to the boundary. The underlying continuous values (total obligated dollars, base contract count, Obligation Density) are published in every scored record, enabling independent reproduction under any alternative threshold choice.
4.5 Class Migration
Class Migration describes the movement of a contractor from one Posture Class to another across scoring periods. The four primary paths:
- Class 3 to Class 1: the growth path. More contracts at increasing dollar values.
- Class 1 to Class 2: the concentration path. No new wins, large contracts retained.
- Class 2 to Class 4: the expiration path. Large contracts end without replacement.
- Class 4 to Class 3: the entry path. New contractor begins winning contracts.
4.6 Velocity
The 5-year award window contains temporal structure. Splitting the window into an early period and a recent period and computing each contractor's volume and frequency in each produces a velocity vector: the direction and magnitude of change in log-transformed obligation space. Velocity measures the rate of new contract wins, not the performance status of existing contracts. A contractor performing on a large contract won in the early period with no new wins in the recent period is correctly classified as declining in new award activity. The continuous velocity vector (dv, df) is published for each contractor; directional labels are a convenience summary.
The structural finding is invariant to split point. At every tested split (1.5, 2.0, 2.5, 3.0, and 3.5 years), Class 1 contractors are more likely to be declining than growing, and Class 3 contractors are more likely to be growing than declining. At the 2-year split: 58% of Class 1 contractors are declining. The top of the pyramid is eroding. 26% of Class 3 contractors are growing, representing the active pipeline of firms moving toward Class 1.
Figure 3. Velocity vector field. Each arrow shows a contractor's direction and magnitude of movement in log-transformed obligation space.
The aggregate velocity described above (fixed 2-year split across the full population) is used for population-level trend detection. A second velocity measure, cadence velocity, is computed at the individual contractor level by comparing time since last award against that contractor's historical average gap between awards and whether recent award pace is faster or slower than the contractor's own baseline.
Cadence velocity distribution across contractors with 2+ base contracts:
| Class | Accelerating | Steady | Slowing | Stalled | Declining | Measured |
|---|---|---|---|---|---|---|
| Class 1 | 10% | 17% | 43% | 0% | 30% | 62/62 |
| Class 2 | 0% | 33% | 33% | 8% | 25% | 12/18 |
| Class 3 | 18% | 13% | 43% | 0% | 25% | 219/219 |
| Class 4 | 7% | 44% | 22% | 0% | 28% | 141/480 |
Cadence velocity requires 2+ contracts to compute a rhythm. All Class 1 and Class 3 contractors have cadence. 6 of 18 Class 2 contractors and 339 of 480 Class 4 contractors have only a single contract and no cadence to measure.
43% of Class 1 contractors are slowing relative to their own historical cadence, and 30% are declining. Only 10% are accelerating. The systematic winners are not maintaining their rhythm. Zero Class 2 contractors are accelerating: concentrated revenue with no growth momentum.
5. Proximity Map
The Proximity Map identifies which contractors in the same state win similar types of contracts at similar scale. For each scored contractor, it outputs the six firms in the index with the highest procurement overlap based on shared procurement codes and comparable dollar volume.
5.1 Computation
For each NAICS code [4] held by the target contractor (derived from award records, not SAM registration):
members = set of all contractors in the index holding that NAICS code
weight = 1.0 / (|members|^2)
For each PSC code [5]:
members = set of all contractors in the index holding that PSC code
weight = 2.0 / (|members|^2)
Each non-target contractor accumulates overlap weight across all shared codes. Rare code overlaps dominate: a NAICS code held by 2 contractors produces a weight of 0.25; a code held by 20 produces 0.0025. Squaring the frequency denominator makes the weighting convex: common codes decay faster than linearly.
PSC codes carry a base weight of 2.0 because they describe what is being delivered, while NAICS codes describe the industry of the provider.
5.2 Scale Filter
Two firms may share a rare PSC code but operate at vastly different dollar volumes. The Proximity Map applies a scale filter:
ratio = min(target_vol, candidate_vol) / max(target_vol, candidate_vol)
proximity_score = overlap_score * (ratio^2)
| Volume ratio | Scale factor |
|---|---|
| 1:1 (identical scale) | 1.00 |
| 1:2 | 0.25 |
| 1:5 | 0.04 |
| 1:10 | 0.01 |
| 1:100 | 0.0001 |
Scale-disjoint pairs converge to zero, ensuring the Proximity Map reflects procurement overlap rather than taxonomic similarity.
The Proximity Map top-6 output shares 39.5% average overlap with cosine similarity on the same procurement code vectors. The divergence is structural: cosine similarity has no scale filter, producing matches between contractors at 1000:1 dollar ratios who share common codes. The Proximity Map's inverse-frequency-squared weighting and scale filter produce a different contractor set, concentrated on firms that share rare codes at comparable scale.
5.3 Neighborhood Density
The Proximity Map produces a graph. Each contractor's six nearest neighbors may or may not be neighbors of each other. The fraction of neighbor pairs that are mutually connected is the Neighborhood Density: a measure of whether a contractor operates in a crowded niche (high density, neighbors win similar contracts to each other) or at a crossroads between separate markets (low density, neighbors are independent).
| Class | Avg Density | Dense (>0.3) | Sparse (<0.1) |
|---|---|---|---|
| Class 1 | 0.289 | 42% | 4% |
| Class 2 | 0.311 | 60% | 0% |
| Class 3 | 0.336 | 47% | 9% |
| Class 4 | 0.423 | 62% | 4% |
Class 4 has the highest average density: their neighbors are tightly clustered in narrow markets. Class 1 has the lowest: systematic winners operate in more diverse, less crowded procurement neighborhoods. Class 2's density (0.311) sits closer to Class 1 (0.289) than to Class 4 (0.423), indicating that high-volume concentrated contractors are not necessarily in crowded niches.
5.4 Proximity Pressure
Combining the Proximity Map with the velocity field produces a measure of directional proximity pressure. For each contractor, the velocity of its neighbors is computed only in shared procurement codes. A neighbor growing in codes the target also holds exerts incoming pressure (compression). A neighbor declining in shared codes creates expanding opportunity. Growth in unrelated codes is excluded.
| Class | Compression | Expansion | Neutral |
|---|---|---|---|
| Class 1 | 19% | 76% | 4% |
| Class 2 | 24% | 65% | 12% |
| Class 3 | 20% | 73% | 5% |
| Class 4 | 22% | 69% | 9% |
76% of Class 1 contractors are in expansion: their neighbors are retreating in shared procurement codes. 24% of Class 2 contractors face compression in their specific codes despite holding high dollar volume.
5.5 Output
The Proximity Map outputs three properties per contractor: the six nearest firms by proximity score, the neighborhood density of the local neighborhood, and the net proximity pressure from neighbor velocity. The proximity output is directional: Contractor A having Contractor B in its map does not guarantee that B has A.
6. Limitations
Subcontracting is not captured. USASpending.gov reports prime awards. A contractor whose primary business model is subcontracting may be underclassified. The Proximity Map identifies which Class 1 and Class 2 contractors operate in the same codes as smaller firms, making natural subcontracting relationships visible under FAR 19.702 [7].
NAICS code accuracy depends on contractor self-classification. The FedComp Index mitigates this by deriving codes from actual award records rather than registration data alone.
The Proximity Map is observational. It identifies contractors who win similar types of contracts at similar scale based on public award data. It does not predict solicitation outcomes or model bidding behavior.
Single-state implementation. The current implementation covers Nevada. The methodology supports additional states without modification; the limitation is operational, not architectural.
7. Falsification Criteria
The methodology would be falsified if:
- The Registration-Obligation Gap were shown to be negligible, rendering Obligation-Verified sets equivalent to registration-derived sets.
- The two-axis classification failed to separate contractors with meaningfully different obligation structures.
- The Proximity Map weighting failed to distinguish contractors who win similar contracts from contractors who merely share a NAICS code.
8. Conclusion
The implementation is open source under the MIT License and maintained publicly on GitHub. The source code, classification engine, and data pipeline are open to inspection, replication, and contribution. It is provided as public infrastructure.
The classification updates continuously as new obligation data becomes available.
References
- USASpending.gov. Award Spending API Documentation. U.S. Department of the Treasury. https://api.usaspending.gov
- System for Award Management (SAM.gov). Entity Registration Data. U.S. General Services Administration. https://sam.gov
- Contract Award Data. SAM.gov Procurement Data. U.S. General Services Administration. https://sam.gov/content/contract-data
- North American Industry Classification System (NAICS). U.S. Census Bureau. https://www.census.gov/naics
- Product and Service Codes (PSC). U.S. General Services Administration. https://www.acquisition.gov/PSC_Manual
- Federal Acquisition Regulation Part 10, Market Research. U.S. General Services Administration. https://www.acquisition.gov/far/part-10
- Federal Acquisition Regulation 19.702, Statutory Requirements for Subcontracting Plans. U.S. General Services Administration. https://www.acquisition.gov/far/19.702
- Federal Acquisition Regulation 10.002, Procedures. U.S. General Services Administration. https://www.acquisition.gov/far/10.002
- Federal Acquisition Regulation 19.202-1, Encouraging Small Business Participation. U.S. General Services Administration. https://www.acquisition.gov/far/19.202-1
- Federal Acquisition Regulation 7.105, Contents of Written Acquisition Plans. U.S. General Services Administration. https://www.acquisition.gov/far/7.105
- Federal Acquisition Regulation 15.304, Evaluation Factors and Significant Subfactors. U.S. General Services Administration. https://www.acquisition.gov/far/15.304
- Federal Acquisition Regulation 33.103, Protests to the Agency. U.S. General Services Administration. https://www.acquisition.gov/far/33.103