Detecting High-Potential SMEs
Leveraging Graph AI for Strategic Innovation Funding
Our SME-HGT framework utilizes heterogeneous graph neural networks on public data to predict which SBIR Phase I awardees will advance to Phase II funding, significantly improving precision and efficiency in identifying high-growth potential small businesses.
Executive Impact
Our advanced AI model provides unprecedented clarity in identifying promising Small and Medium Enterprises (SMEs), leading to more effective resource allocation and strategic investment decisions.
0.000
AUPRC Achieved (±0.003)
0
Precision @ Top 100 Firms
0.00
Lift Over Random Selection
0
Phase II Progression Base Rate
Deep Analysis & Enterprise Applications
Select a topic to dive deeper, then explore the specific findings from the research, rebuilt as interactive, enterprise-focused modules.
Overview: Identifying High-Growth SMEs
Small and Medium Enterprises (SMEs) are vital for the U.S. economy, yet systematically identifying those with high growth potential remains challenging. The SBIR program, a major source of early-stage technology funding, serves as a crucial indicator, with progression from Phase I to Phase II signaling significant potential. Our research addresses this challenge by introducing SME-HGT, a novel approach leveraging relational data to improve prediction accuracy.
Our Heterogeneous Graph Approach
We developed SME-HGT, a Heterogeneous Graph Transformer framework, using exclusively public data from sbir.gov. The model operates on a heterogeneous graph comprising 32,268 company nodes, 124 research topics, and 13 government agency nodes, connected by approximately 99,000 edges across three semantic relation types. This architecture processes input features through type-specific linear projections and multiple HGT layers to learn rich, context-aware embeddings. A strict temporal evaluation protocol ensures realistic deployment conditions by preventing information leakage, with Phase I award data prior to Jan 1, 2018, used for features, and Phase II progression labels defined within a 5-year horizon.
Performance & Relational Insights
SME-HGT significantly outperforms baseline models, achieving an AUPRC of 0.621 ± 0.003, a 3.1 percentage point improvement over MLP and 1.3 pp over R-GCN. At a screening depth of 100 companies, SME-HGT attains 89.6% precision, representing a 2.14x lift over random selection. This demonstrates that the relational structure among firms, research topics, and funding agencies provides meaningful signals for assessing SME potential. While overall discrimination is moderate, the ranking metrics highlight substantial practical value for screening.
Policy & Future Directions
Our findings have direct implications for SBIR program administration, enabling more efficient allocation of expert review resources and earlier identification of high-potential firms. The public-data-only design ensures replicability across various jurisdictions with structured grant data. Future work includes incorporating textual features from award abstracts, adding temporal dynamics, conducting ablation studies to quantify contributions of different node/edge types, and extending prediction to other targets like patent filing or acquisition events.
2.14x
Lift in identifying high-potential SMEs over random selection
SME-HGT Prediction Workflow
Heterogeneous Graph G
→
Type-Specific Linear Projection
→
HGT Layer + Residual + LayerNorm + Dropout (x3)
→
MLP Classifier (Company nodes)
→
Phase II Prediction
| Model | AUPRC | Key Advantages |
|---|---|---|
| SME-HGT | 0.621 ±0.003 |
|
| R-GCN | 0.608 ±0.013 |
|
| MLP Baseline | 0.590 ±0.002 |
|
Transforming Public Policy with Predictive AI
Summary: For SBIR program administrators, identifying high-potential Phase I awardees is crucial. Our SME-HGT model achieves 89.6% precision at the top 100, meaning it can identify ~90 successful firms out of 100 reviewed, compared to ~42 by random selection. This 2.14x lift translates to significantly more efficient allocation of expert review resources, enabling targeted technical assistance and fostering innovation.
The Challenge: Systematically identifying high-growth SMEs from a vast pool of applicants.
The Solution: Leveraging heterogeneous graph neural networks on public SBIR data to predict Phase II progression.
The Outcome: A 2.14x lift in identifying successful firms, dramatically improving efficiency and impact for policymakers.
Calculate Your Potential ROI
Estimate the impact of AI-driven SME identification on your operational efficiency and strategic investment returns.
Estimated Annual Savings
$0
Analyst Hours Reclaimed Annually
0
Your AI Implementation Roadmap
A structured approach to integrating heterogeneous graph neural networks for predictive analytics within your organization.
Phase 1: Data Integration & Graph Construction
Aggregate and clean public SBIR data, perform entity resolution, and construct the heterogeneous graph with company, topic, and agency nodes. Ensure data integrity and prepare features for model input.
Phase 2: Model Training & Validation
Train the SME-HGT model using historical Phase I awards, applying our strict temporal evaluation protocol. Tune hyperparameters for optimal performance and ensure robust validation against real-world conditions.
Phase 3: Deployment & Iteration
Deploy the trained model as a screening tool for policymakers and investors. Continuously monitor predictions, gather feedback, and iterate on graph structure, features, and model architecture for ongoing improvement and expanded applications.
Ready to Transform Your Enterprise with AI?
Connect with our experts to explore how heterogeneous graph neural networks can unlock new levels of insight and efficiency for your organization.
Ready to Get Started?
Book Your Free Consultation.
Let's Discuss Your AI Strategy!
Lets Discuss Your Needs
Select a Date
Sun
Mon
Tue
Wed
Thu
Fri
Sat