Event Oracle

TEST 1

Aggregation Query - Most Common Events

"Show me the top 10 most common event types"

SELECT event_type, COUNT(*) as count
FROM events
GROUP BY event_type
ORDER BY count DESC
LIMIT 10;

Rank	Event Type	Count
1	complied_regulatory_compliance_moderate	531,486
2	developed_product_development_moderate	475,810
3	developed_technology_development_moderate	395,604
4	authorized_other_material_small	305,332
5	complied_financial_compliance_moderate	189,940
6	impaired_asset_impairment_medium	183,231
7	developed_drug_development_moderate	175,118
8	reorganized_corporate_restructuring_substantial	173,055
9	approved_regulatory_approval_small	159,153
10	granted_stock_option_none	151,001

💡 Insight

Regulatory compliance (721K events) and innovation/development (1.05M events) dominate, representing 15% of all 11.9M events. High frequency of routine compliance validates our compression strategy.

TEST 2

Pattern Matching - Red Flag Detection

"Find companies with workforce reductions in 2024"

SELECT cik, company_name, event_type, event_date, sentiment
FROM events
WHERE (event_type LIKE 'reduced_%workforce%' OR event_type LIKE 'terminated_%')
AND event_date >= '2024-01-01'
AND event_date < '2025-01-01'
ORDER BY event_date DESC
LIMIT 20;

💡 Insight

Mix of minor adjustments and major restructuring efforts. Most terminations reported on December 31 (fiscal year-end timing). Claude correctly used LIKE patterns to match compound event_type format (verb_object_magnitude).

TEST 3

Predictive Analysis - Returns by Event Type

"Which event types have the best average 6-month returns with at least 100 occurrences?"

SELECT event_type, AVG(return_6m) as avg_return_6m, COUNT(*) as count
FROM events
WHERE return_6m IS NOT NULL
GROUP BY event_type
HAVING COUNT(*) >= 100
ORDER BY avg_return_6m DESC
LIMIT 20;

Rank	Event Type	Avg 6M Return	Count
1	discontinued_operation_cessation_major	2,010.87%	414
2	entered_consulting_agreement_minor	1,165.69%	108
3	covenant_violation_default_major	702.46%	157
4	filed_bankruptcy_filed_critical	648.72%	454
5	impaired_goodwill_impairment_critical	412.33%	127

💡 Insight

Massive returns associated with distress events - likely survivorship bias. Only companies that recovered from bankruptcy/discontinuation still report returns 6 months later. Failed companies have no return data.

Alpha Opportunity: Market initially overreacts to distress signals. Predictive models could identify which distressed companies will survive.

TEST 4

Red Flag Detection - Auditor Dismissals

"Show me companies with auditor dismissals in the last 5 years"

SELECT cik, company_name, event_type, event_date
FROM events
WHERE (event_type LIKE 'dismissed_auditor%' OR event_type LIKE 'changed_auditor%')
AND event_date >= CURRENT_DATE - INTERVAL '5 years'
ORDER BY event_date DESC
LIMIT 20;

💡 Insight

Most auditor changes are routine, but CRITICAL and MAJOR dismissals indicate serious issues:

Disagreements over accounting treatment
Discovery of material weaknesses
Precursor to restatements or investigations
Potential fraud or going concern issues

Use Case: Run daily to generate red flag alerts for investment risk systems. Automatic notifications when CRITICAL or MAJOR dismissals occur.

TEST 5

Temporal Pattern Detection - Partnership → Development

"Find companies that had a partnership event followed by a product development event within 90 days"

SELECT p.cik, p.company_name,
       p.event_type as partnership_event, p.event_date as partnership_date,
       d.event_type as development_event, d.event_date as development_date
FROM events p
JOIN events d ON p.cik = d.cik
                 AND d.event_date > p.event_date
                 AND d.event_date <= p.event_date + INTERVAL '90 days'
WHERE p.event_type LIKE 'partnered_%'
  AND d.event_type LIKE 'developed_%'
ORDER BY p.event_date DESC
LIMIT 20;

Company	Partnership Event	Development Event	Gap
CUMMINS INC	partnered_joint_venture_substantial	developed_technology_development_significant	⚡ 4 days
BioAtla, Inc.	partnered_licensing_agreement_moderate	developed_tax_law_development_moderate	⚡ 4 days
Nukkleus Inc.	partnered_strategic_alliance_substantial	developed_exclusive_distribution_moderate	⚡ 9 days
Blue Gold Ltd	partnered_strategic_alliance_substantial	developed_product_development_moderate	19 days
PALATIN TECHNOLOGIES	partnered_strategic_alliance_substantial	developed_technology_development_moderate	32 days

💡 Insight

Strategic partnerships lead to rapid product development within 1-3 months. Three patterns emerge:

Lightning-fast (⚡ 4-9 days): Partnership was FOR the development
Rapid (19-39 days): Development initiated immediately after partnership
Planned (61-80 days): Development rolled out over 2-3 months

🚫 This Query is IMPOSSIBLE with Text-Based RAG (Fintool)

Fintool searches text chunks and returns passages. It cannot JOIN events by company and calculate date differences. Vector search has no concept of temporal ordering. SQL's date arithmetic (INTERVAL '90 days') has no equivalent in embedding space.

This is Event Oracle's unique advantage: Structured events + SQL = temporal pattern discovery impossible with text search.

Capability	Fintool (RAG)	Event Oracle (Structured)
Basic Q&A	✓ Excellent	✓ Excellent
Temporal Patterns	✗ Impossible	✓ Native (SQL JOINs)
Aggregations	✗ Limited	✓ Native (GROUP BY)
Predictive Analysis	✗ Not available	✓ Returns pre-calculated
Pattern Discovery	✗ Cannot do	✓ Complex SQL queries
Cost (1K queries/day)	$143K/month	$450/month
Query Speed	2-5 seconds	2-3 seconds

🎯 Key Innovation: PostgreSQL as Structured RAG

Architecture

Cost Comparison: Event Oracle vs Fintool

Fintool (Traditional RAG)

Event Oracle (Structured RAG)

Test Results: 5 Query Types Demonstrated

Aggregation Query - Most Common Events

💡 Insight

Pattern Matching - Red Flag Detection

💡 Insight

Predictive Analysis - Returns by Event Type

💡 Insight

Red Flag Detection - Auditor Dismissals

💡 Insight

Temporal Pattern Detection - Partnership → Development

💡 Insight

🚫 This Query is IMPOSSIBLE with Text-Based RAG (Fintool)

Capability Comparison: RAG vs Structured

Key Insight