Sector Quantum Stress Index · UK SME risk intelligence
Forward-looking UK SME distress,
by sector × region, every quarter.
A hybrid quantum-AI risk index applying IBM Research, Zurich’s published credit-risk methodology to the UK’s largest unified SME data asset (4.8M enriched companies, 12M-edge director graph, 18 regulator-grade signal feeds, Insolvency Service outcome labels). The first quarterly publication covers Construction (SIC 41-43) × West Midlands (NUTS-1) from Q3 2026.
What the index measures
Per-sector distress probability
The probability that a representative cohort of UK SMEs in a given SIC × region experiences a Gazette-recorded insolvency event within 90, 180, or 365 days. Computed via a hybrid quantum-classical classifier (Phase 2) and aggregated quarterly.
Tail-risk concentration (top decile)
The 90th-percentile distress score across the cohort. Identifies how concentrated the bad outcomes are. Sectors where p90 is far above the mean are showing tail-risk clustering, typically pre-cascade warning signal.
Cohort-size + composition
Number of companies in the cohort and stratification by size band (turnover quartile). Smaller cohorts get wider confidence intervals; these are surfaced honestly rather than hidden.
How it is computed
- Data source. Five-year iXBRL filing trajectories for ~50,000 stratified UK SMEs, joined to Gazette insolvency outcomes via
outcomes.insolvency_events. Stratified to balance positive (dissolved) vs negative (survived) labels. - Classical baseline. RBF-kernel Support Vector Machine on the same feature set + train/test split, the benchmark the quantum classifier must match or beat. Reproduction script:
scripts/quantum/03_train_classical_baseline.py. - Quantum classifier. Variational Quantum Classifier (Havlíček et al., Nature 2019) and Quantum Kernel SVM (Schuld, arXiv:2101.11020) trained on identical data via the second-order ZZ feature map. Implemented on Qiskit Machine Learning.
- Sector aggregation. The trained classifier scores every company in a SIC × region cohort. The aggregate produces the Sector Quantum Stress Index (mean, p50, p90, top-stressed-count). Updated quarterly.
- Hardware validation. A subset of inference runs is executed on IBM Heron r2 (
ibm_kingston) via Qiskit Runtime with T-REx readout error mitigation (Stamatopoulos et al. 2020). Reported alongside the simulator-based result for noise-margin transparency.
Full methodology + paper citations: /methodology#quantum. Implementation lives at scripts/quantum/; research notes at docs/strategy/quantum-papers/.
Publication roadmap
The first issue covers a single pilot cohort. Subsequent issues expand sector × region coverage on a published schedule.
Q3 2026
PilotPilot issue
Construction (SIC 41-43) × West Midlands (NUTS-1). Methodology defence + first SQSI value + 4-year historical back-fit.
Q4 2026
PlannedSector expansion
Add 3 sectors: Retail & wholesale, Hospitality, Transport & logistics.
Q1-Q2 2027
PlannedRegion expansion
Full 9 NUTS-1 region coverage per priority sector.
Year 2
ResearchEmpirical recalibration
Post-IP-final-report ingestion: switch LGD from rules-based to empirically-fitted. Re-train quantum classifier on richer outcome labels.
Honest limits
- This is not a quantum-advantage claim. At current NISQ hardware capacity, the simulator and the QPU produce equivalent answers in equivalent wall-clock time at the problem sizes RecoupIQ operates at. The index is a quantum-classical hybrid methodology layer, not a quantum-only product.
- Sector context, not a per-company score. The SQSI describes the cohort. Per-company risk uses the classical pricing engine documented at /methodology.
- Index lag. Quarterly retraining means the index reflects the most recent fully-elapsed quarter. Real-time monitoring is a different product surface (forensic report + watchlist).
- Confidence bands surfaced, not hidden. Smaller cohorts have wider confidence intervals. We report them on every published index value rather than dropping the row.
- Not a credit rating. Under the UK Credit Rating Agencies Regulations 2010 / EU CRA Regulation, this index does not constitute a credit rating. It is decision-support intelligence for sector-allocation, portfolio monitoring, and underwriting context.
Who consumes the index
Specialist SME lenders & invoice-finance
Sector concentration limits, lending-book stress monitoring, portfolio-level VaR/CVaR inputs.
Trade credit insurers
Premium pricing context, sector capacity decisions, underwriting overlay.
Insolvency practitioners, restructuring advisers, litigation funders
Quarter-ahead sector focus: which sectors will produce the most new engagements next quarter.
Accountants and business advisors
Client-advisory context, “is this client’s sector deteriorating faster than the wider economy?”, backed by transparent methodology.
UK SMEs and entrepreneurs
Free public quarterly publication, sector-entry, supplier-vetting, and where-to-trade context for the segment most exposed to late payment.
Policy, press, academics
Citable, version-stamped, methodology-published index of UK SME stress by sector × region.
Notified when the Q3 2026 pilot lands?
Free notification, no obligation. The pilot issue is published as a downloadable PDF and an interactive table on this page.
