RPS = w1 × PriceDecay + w2 × VolumeDecay + w3 × DaysInactive
 

PriceDecay   = (current_price / peak_52w_price −- 1.0), bounded [0, 1]
VolumeDecay  = 1 −- (avg_30d_volume / avg_prior_year_volume), bounded [0, 1]
DaysInactive = min(days_since_last_trade / 730, 1.0)
 

Weights (empirically calibrated): w1 = w1=0.35 (price decay)
  w2 = w2=0.30 (volume decay)
  w3 = w3=0.35
(days
 inactive)
RPS > 0.80   →  Critical distress   — highest tokenization urgency
RPS 0.60–0.80 →  Significant distress — high urgency
RPS 0.40–0.60 →  Moderate distress   — standard outreach
RPS < 0.40   →  Early stage         — nurture sequence

 
🎯9.4 8.4 Composite IDOS Score Architecture

9.4.1  Signal Weights and Refresh Cadence

The Issuer Distress and Opportunity Score (IDOS)
🔹 8.4.1 Score Architecture

The Issuer Distress and Opportunity Score (IDOS) synthesizes all EDGAR and OTC Markets signals into a single normalized composite score for each of the ~approximately 15,000 companies in the target universe, refreshed continuously as new data becomesarrives available.through real-time feed monitoring.
 


















































Signal Component



Source



Weight



Refresh Cadence


Liquidity Distress Index (LDI) — NLP scan


EDGAR 10-K / 10-Q


0.22


Per new filing


Tier Degradation Score


OTC Markets


0.18


Daily


Recovery Probability Score (RPS)


OTC Markets


0.18


Daily


Time Since Last Recorded Trade


OTC Markets


0.12


Daily


Shareholder Count (log-normalized)


EDGAR DEF 14A


0.12


Annual / on new filing


8-K Trigger Score


EDGAR RSS Feed


0.10


Real-time


Form D Urgency Score


EDGAR Form D


0.08


Per new filing


 
🔹 8.9.4.2  Outreach Pathway Multipliers
The composite IDOS score is adjusted by an Outreach Pathway Multiplier reflecting the quality of the existing relationship between OTCM Protocol and the issuer's transfer agent:

 

























Outreach Classification
Pathway
Multiplier
Criteria

Multiplier


Criteria


Direct Referral


2.0×


ESTEmpire Stock Transfer client confirmed + 10,000+ shareholders of record


Warm Lead


1.5×


ESTEmpire Stock Transfer client confirmed · any shareholder count


Cold Prospect


1.0×


No existing ESTEmpire or OTCM relationship on file


 
🔹 8.9.4.3  Dynamic Priority Queue
The IDOS engine maintains a continuously ordered priority queue across all ~15,000 target universe companies. Outreach automation consumes from the head of this queue, dispatching the highest-IDOS issuers into active sequences while maintaining a rolling view of the top 50–100 priority targets at any given time. Companies already in active outreach sequences are excluded from re-dispatch until the current sequence resolves — preventing duplicate contact and ensuring professional engagement cadence.

 
🤖9.5  AI Model Technical Specification
Version:
9.5.1 6.1 | Applies To: Layer 9 — Predictive Marketing AI Module

Model Architecture — IDOS Scoring Engine
The IDOS is computed by a gradient boosted decision tree ensemble (XGBoost v2.x). This architecture was selected for three reasons specific to the OTCM use case:
 
Interpretabilityinterpretability — each scoring decision can be traced to specific feature contributions, supporting regulatory defensibility
Performancedefensibility; performance on sparse tabular data — SEC EDGAR filings produce sparse, irregular time-series data that gradient boosting handles bettermore effectively than sequence models
Lowmodels; and low inference latency — sub-10ms per issuer scoring enables real-time refreshes across 15,000 companiescompanies.

 
9.5.2  Training Methodology
 








































Parameter
Specification

Parameter


Specification


Training data window


2015–2024 SEC EDGAR + OTC Markets historical data


Training sample size


~180,000 issuer-quarter observations


Positive class definition


Issuers that engaged with a tokenization/tokenization or liquidity solution within 12 months


Negative class definition


Issuers that remained in grey/grey or expert market without engagement


Train / validation / test split


70% / 15% / 15% (— time-based split —split, no future data leakage)leakage


Cross-validation


Walk-forward validation — 4 folds ·folds, each fold = 2 years


Held-out test period


2023–2024 data (never seen during training)


Primary evaluation metric


AUC-ROC on held-out test set (target: > 0.78)


 
9.5.3  Model Drift Detection and Retraining Policy
 
































Metric



Monitoring Frequency



Drift Threshold



Response


Score distribution shift (PSI)


Weekly


PSI > 0.20


Alert + manual review


Prediction accuracy on recent cohort


Monthly


AUC-ROC drops > 0.05 from baseline


Trigger retraining


Feature distribution shift


Weekly


KS-statistic > 0.15 for any feature


Alert + feature audit


Engagement prediction vs. actual


Quarterly


Precision/recall drops > 10%


Full model retraining


Retraining Cadence:

 Scheduled quarterly retraining with 6-month rolling training window expansion. Emergency retraining triggered if drift thresholds are exceeded. Retrained models require validation AUC-ROC ≥ 0.75 before promotion to production.
 
9.5.4  Model Governance Gates
 

































Governance Gate



Requirement



Approver


New model promotion


Validation AUC-ROC ≥ 0.75 on held-out test set


CTO sign-off— Frank Yglesias


Feature addition


Privacy review + SHAP impact analysis


ComplianceCLO Officer— Jeff Turner


Training data expansion


Data provenance documentation


LegalCLO Counsel— Jeff Turner


Production deployment


A/B shadow mode run ≥ 2 weeks


Engineering Lead


Emergency rollback


Triggered automatically if drift threshold exceeded


Automated


 
💼9.6 8.5 Investor-Side Predictive Intelligence
🔹9.6.1 8.5.1 Wallet Behavioral Profiling
On the investor side, the module operates a parallel intelligence layer targeting accredited investors most likely to purchaseparticipate in ST22 Digital Securities tokens during the bonding curve phaseofferings and post-graduation CPMM trading on CEDEX. Wallet behavioral profiles are constructed from on-chain analytics aggregated across Solana mainnet, supplemented by off-chain identity enrichment where publicly available.

On-Chain Behavioral Signal
Investor Profile Implication
Holdings of tokenized equity instruments
Demonstrated preference for regulated tokenized assets
Idle USDC/USDT balance > $25,000
Available dry powder for new position deployment
Historical bonding curve participation
Experience with launch-phase entry mechanics
Governance token holdings across protocols
Protocol-engaged sophisticated investor profile
Average position hold time > 90 days
Long-horizon investor · lower churn risk
LP provision behavior in compliant pools
Activesecondary market participant · yield-seeking behavior


🔹 8.5.2 Rule 506(c) Compliance Enforcement

All investor-side outreach operates under Rule 506(c) of Regulation D, which permits general solicitation and advertising provided that issuers take reasonable steps to verify that all purchasers are accredited investors. The module enforces this constraint programmatically at the targeting layer — no outreach sequence is triggered for any wallet or identity that has not cleared the accreditation proxy filter.

Verified accreditation records from the Issuers Portal Compliance Gateway are shared with the investor targeting module via the internal compliance data bus, eliminating duplicative verification workflows for investors who have previously undergone accreditation checks on the platform.


🔹 8.5.3 Launch Timing Optimization Engine (LTOE)

Post-deployment analysis identified that suboptimal launch timing created measurable drag on bonding curve performance during the critical price discovery window. The Launch Timing Optimization Engine (LTOE) generates a Launch Readiness Score (LRS) for each pending ST22 deployment by modeling five environmental factors:

Competing token launch schedules on Solana within the preceding and following 7-day window
Solana mainnet RPC congestion index and historical throughput patterns
Macroeconomic sentiment index (−1.0 risk-off to +1.0 risk-on) derived from market data feeds
Community engagement readiness score from social media and wallet pre-registration metrics
Qualified investor pool depth — count of accredited investor wallets confirmed ready to participate

LRS < 0.55  →  Launch deferral recommended
              Engine proposes alternative window within 7–14 days

LRS > 0.80  →  Expedited pre-launch investor notification sequence triggered
              Maximize pool participation in first 24 hours of bonding curve


🛡️ 8.6 OTCM Security Token (OTCM STO) Integration

🔹 8.6.1 AI Module as Staking Tier Capability

The Predictive Marketing AI Module integrates directly into the OTCM Security Token staking tier architecture as a gated platform capability. All benefits require active platform engagement — staking alone, without corresponding module usage, confers no incremental value.

Staking Tier
OTCM Staked
AI Module Access Unlocked
Bronze
1,000 OTCM
IDOS dashboard read-only access · top 500 issuers by score
Silver
10,000 OTCM
Full IDOS access + weekly AI-generated prospect report
Gold
50,000 OTCM
Full EDGAR NLP Engine + OTC Markets tier alerts + investor pool analytics
Platinum
100,000 OTCM
Complete suite: real-time feeds · IDOS priority queue · LTOE · wallet profiling · outreach automation


🔹 8.6.2 Per-Operation Token Burn Mechanics

Individual AI module operations carry token burn costs, generating deflationary pressure through genuine platform utilization. Burns occur at the smart contract level, are irreversible, and are publicly auditable on-chain.

AI Module Operation
OTCM Burn Cost
EDGAR batch query execution (per 500 records)
1,000 OTCM burned
OTC Markets feed refresh subscription (monthly)
50 OTCM burned
Investor wallet behavioral report (per ST22 launch)
250 OTCM burned
Launch Readiness Score analysis (per deployment)
500 OTCM burned
Full IDOS universe refresh cycle
1,000 OTCM burned
Automated outreach sequence launch (per issuer)
750 OTCM burned


🔹 8.6.3 Governance Integration

OTCM Security Token holders at Gold tier and above may submit governance proposals governing AI module operating parameters. Governable parameters include:

IDOS component weighting ratios
Tier degradation alert thresholds
LRS deferral thresholds
Outreach sequence cadence parameters
Composition of the EDGAR NLP distress phrase corpus

Governance over AI module parameters is operationally meaningful — it determines how the protocol allocates commercial attention — and is structurally consistent with the governance model, as it governs platform function rather than profit distribution.


🏰 8.7 Data Moat and Competitive Defensibility

🔹 8.7.1 Proprietary Dataset Accumulation

Every campaign executed, every issuer conversion or non-conversion, every investor wallet interaction, and every ST22 launch outcome feeds back into the module's training data. The module improves continuously with platform scale. After 500 ST22 launches, Groovy Company, Inc. dba OTCM Protocol will have accumulated a dataset on tokenized Digital Securities investor behavior, issuer conversion patterns, and launch timing correlations that no competitor can replicate without operating at equivalent scale and regulatory depth.

This dataset constitutes a structural moat: the AI's accuracy and efficiency improve as the platform grows, which improves commercial outcomes, which accelerates platform growth — compounding the data advantage with each issuer onboarded.


🔹 8.7.2 Moat Reinforcement Dynamics

Successful issuer conversion
         ↓
ST22 Digital Securities trading volume
         ↓
Transaction fee revenue
         ↓
AI module development investment
         ↓
Improved IDOS accuracy → faster conversion at lower cost
         ↓
Better unit economics → reinvestment in data infrastructure
         ↓
(Loop repeats — self-funding and self-reinforcing)

The competitive advantage is therefore self-funding and self-reinforcing — requiring no incremental investment beyond normal platform operations to maintain and grow.


⚡ 8.8 Performance Specifications

Performance Metric
Specification
IDOS refresh latency
 (8-K trigger event)
< 60 seconds from EDGAR RSS publication
IDOS full universe refresh cadence
Every 24 hours · continuous partial refresh
EDGAR full-text search throughput
1,000 CIKs per batch cycle
OTC Markets tier change detection
Near real-time · 15-minute polling interval
Investor wallet profile refresh
Every 6 hours · event-driven on large on-chain movements
Launch Readiness Score update frequency
Every 4 hours · real-time on Solana congestion events
Priority queue depth
15,000+ issuers · memory-resident with persistent backing
EST cross-reference match latency
< 5 seconds on new issuer ingestion
Outreach sequence trigger delay
< 2 minutes from IDOS threshold breach


▪️ 8.9 Regulatory and Privacy Considerations

All data sources consumed by the Predictive Marketing AI Module are publicly available. SEC EDGAR is a public database operated for the express purpose of providing universal investor access to issuer filings. OTC Markets Group publishes tier data, issuer profiles, and trading statistics as a market infrastructure function. No proprietary, non-public, or personally identifiable data is accessed without explicit consent in the issuer prospecting pipeline.

Investor-side wallet behavioral profiling operates on on-chain transaction data, which is inherently public on the Solana blockchain. Off-chain identity enrichment is limited to publicly available professional data sources.

Compliance framework:

CAN-SPAM Act requirements for all outreach communications
GDPR Article 6(1)(f) legitimate interest basis where applicable
Rule 506(c) general solicitation standards for all investor outreach
No storage or processing of payment card data, SSNs, or health information
All data handling practices documented in the OTCM Protocol Privacy Policy
Subject to the annual compliance audit framework described in Section 6


Groovy Company, Inc. dba OTCM Protocol · Wyoming Corporation · invest@otcm.io · otcm.io

🤖 Layer 9 — The commercial intelligence engine monitoring ~15,000 U.S. OTC companies in real time, scoring issuer distress, targeting verified accredited investors, and optimizing ST22 launch timing.


▪️ 8.1 Module Overview and Strategic Purpose

🔹 8.1.1 The Commercial Velocity Problem

OTCM Protocol's technical infrastructure — ST22 tokens, Transfer Hook enforcement, the Federated Liquidity Protocol, and Empire Stock Transfer custody — constitutes a best-in-class architecture for tokenizing illiquid securities. However, technical excellence alone does not generate issuer pipeline or trading volume. The protocol faces a structural commercial challenge that no amount of infrastructure investment resolves: issuers and investors must find each other before tokenization can occur.

The target universe of approximately 11,000+ OTC companies with degraded or eliminated market maker eligibility represents approximately $50 billion in trapped shareholder value. These companies do not systematically self-identify as tokenization candidates. Their boards are not monitoring blockchain infrastructure developments. Their shareholders are experiencing liquidity crises in silence. Without a systematic, data-driven mechanism to identify the highest-urgency candidates and route them into the onboarding framework, pipeline growth remains dependent on manual outreach — a fundamentally unscalable approach for a platform designed to operate at the scale of the entire OTC market.

On the investor side, ST22 tokens face an equivalent discovery problem. Unlike exchange-listed securities with Bloomberg terminal coverage, broker-dealer distribution networks, and institutional research coverage, ST22 tokens are invisible to their highest-probability buyers — accredited investors, family offices, and tokenized asset specialists — absent a systematic mechanism to surface launches at the precise moment of investor intent. The Predictive Marketing AI Module resolves both sides of this equation simultaneously, functioning as the commercial engine that converts OTCM's technical moat into revenue velocity.

🔹 8.1.2 Module Architecture Philosophy

The Predictive Marketing AI Module is designed around three core architectural principles:

Data Primacy: All prospecting and targeting decisions are derived from structured public data sources — SEC EDGAR filings and OTC Markets Group datasets — rather than purchased contact lists or broad-spectrum advertising. This approach produces higher conversion rates, lower acquisition costs, and full regulatory defensibility.
Mathematical Scoring over Human Judgment: The module generates an Issuer Distress and Opportunity Score (IDOS) for every OTC-listed company in the target universe, refreshed on a continuous basis. Outreach prioritization is driven by algorithmic scoring rather than relationship-dependent sales cycles, enabling the platform to process hundreds of issuers in parallel without proportional headcount growth.
Compliance-Constrained Targeting: All investor outreach operates within Rule 506(c) general solicitation parameters — targeting only verified or verifiable accredited investors. The module enforces this constraint programmatically, filtering wallet behavioral profiles and off-chain identity signals through accreditation criteria before any outreach sequence is triggered.

🔹 8.1.3 Module Position in the Protocol Stack

The Predictive Marketing AI Module operates at Layer 4 (Application Services) of the OTCM nine-layer architecture, consuming data from Layer 6 (Oracle Network and Custody Integration) and feeding qualified issuer and investor records into the Issuers Portal Compliance Gateway (Section 7) and CEDEX trading infrastructure (Section 4). The module does not alter or bypass any Transfer Hook security controls; it operates exclusively at the pre-onboarding discovery and targeting layer.

🏗️ 8.2 SEC EDGAR Integration Architecture

🔹 8.2.1 EDGAR Data Feed Overview

The U.S. Securities and Exchange Commission's Electronic Data Gathering, Analysis, and Retrieval system (EDGAR) constitutes the authoritative public record for all U.S. registered and reporting companies. For the Predictive Marketing AI Module, EDGAR functions as a real-time issuer distress signal monitor. All data consumed from EDGAR is publicly available and carries no data licensing restrictions or privacy implications.

EDGAR Endpoint
AI Module Application
efts.sec.gov (Full-Text Search API)
Distress language NLP scan — real-time
data.sec.gov/submissions/ (Filing History)
Form D, 8-K, 10-K indexing per CIK
data.sec.gov/api/xbrl/ (Structured Financials)
Financial health scoring — quarterly
EDGAR RSS Feed (New Filing Notifications)
8-K trigger alert pipeline — real-time

🔹 8.2.2 Form D Filing Intelligence

Every company raising capital under Regulation D Rule 506(b) or Rule 506(c) must file a Form D with the SEC within 15 days of the first sale. The module monitors Form D filings continuously to identify companies that are actively capital-raising but lack public trading market liquidity — a compound signal that strongly predicts tokenization readiness. A Form D filed 12-24 months prior with no subsequent SEC filings and no reported active trading market produces the highest urgency score in this category, indicating a company that has exhausted its private raise capacity and has no public liquidity alternative.

🔹 8.2.3 Natural Language Processing — Liquidity Distress Index

Annual reports (Form 10-K) and quarterly reports (Form 10-Q) contain Management Discussion and Analysis (MD&A) sections where management is required to disclose material conditions affecting the company's operations and securities. The module applies NLP-based sentiment and keyword extraction to these filings, constructing a Liquidity Distress Index (LDI) for each issuer. LDI scores are normalized to a 0-100 scale and updated each time a new filing is detected in the EDGAR RSS feed for a given CIK.

Target phrase corpus (weighted by distress signal strength):

HIGH WEIGHT (2.0x): 'no established trading market exists' | 'shareholders may have difficulty selling'
HIGH WEIGHT (2.0x): 'limited or no market maker activity' | 'no assurance that a liquid market will develop'
MEDIUM WEIGHT (1.5x): 'thin trading volume' | 'limited trading market' | 'no broker-dealer has agreed to make a market'
INDICATOR WEIGHT (1.0x): 'OTC Markets' | 'Pink Sheets' | '15c2-11' | 'delisted' | 'trading was suspended'

🔹 8.2.4 8-K Trigger Alert System

Form 8-K current reports disclose material events at public companies in near-real-time. The module monitors the EDGAR RSS feed for 8-K filings from companies within the target universe and scores each filing against a trigger classification matrix. An 8-K trigger alert routes the issuer to an expedited outreach sequence, bypassing the standard IDOS queue, on the basis that companies disclosing material distress events have boards actively seeking solutions and represent high-conversion windows typically lasting 30-90 days.

8-K Item Code / Event Type
Trigger Score & Outreach Response
4.01 — Changes in Certifying Accountant
0.85 — High priority: leadership / audit uncertainty
4.02 — Non-Reliance on Prior Financial Statements
0.90 — High priority: financial restatement distress
5.02 — Departure / Appointment of Officers
0.65 — Medium: board restructuring signal
1.03 — Bankruptcy or Receivership Filing
0.95 — URGENT: shareholder protection window open
2.04 — Triggering Events for Debt Acceleration
0.80 — High: debt covenant / liquidity stress

🔹 8.2.5 Proxy Statement Shareholder Count Intelligence

Annual proxy statements (Form DEF 14A) and Form 10-K annual reports contain the number of holders of record for each class of common equity, machine-readable through the EDGAR XBRL API. The module extracts and indexes shareholder count data for all target universe companies, enabling prioritization by trapped holder volume — a direct proxy for the social and investor relations impact of OTCM's intervention and a key determinant of issuer willingness to bear onboarding costs.

Shareholder Count Tier
Outreach Priority & Response
10,000+ holders of record
TIER 1 — Immediate outreach; direct referral pathway if EST client
2,500 – 9,999 holders of record
TIER 2 — Scheduled outreach sequence; standard IDOS queue
500 – 2,499 holders of record
TIER 3 — Nurture sequence; periodic re-score
< 500 holders of record
TIER 4 — Monitoring only; flag for future re-evaluation

🔌 8.3 OTC Markets Group Intelligence Integration

🔹 8.3.1 OTC Markets Data Architecture

OTC Markets Group operates the electronic marketplace for approximately 12,000 U.S. and international securities across three market tiers: OTCQX (highest standards), OTCQB (venture stage), and OTC Pink (open market). OTC Markets publishes structured issuer profile data, trading history, tier status, news feeds, and transfer agent information through publicly accessible interfaces. The module integrates OTC Markets data across five distinct intelligence categories for continuous issuer universe scoring.

🔹 8.3.2 Tier Degradation Monitoring

Market tier assignment reflects issuer compliance with disclosure requirements and OTC Markets standards. Downward tier movement — from OTCQB to Pink, from Pink Current Information to Pink Limited Information, or from Pink Limited to No Information or Expert Market designation — represents a measurable, timestamped signal of issuer deterioration that correlates directly with tokenization urgency. The module monitors tier transitions daily, assigning distress delta scores that feed into the composite IDOS calculation.

OTC Market Tier
Distress Weight in IDOS Calculation
Dark / Expert Market (retail trading blocked)
1.00 — Maximum distress; immediate outreach
Pink No Information (zero disclosure)
0.85 — Critical; shareholder rights severely impaired
Pink Limited Information (lapsed disclosure)
0.65 — High; recovery probability declining
Pink Current Information (meets disclosure)
0.40 — Moderate; monitor for further degradation
OTCQB / OTCQX (active market tiers)
0.10 — Low; secondary prospects only

🔹 8.3.3 Transfer Agent Cross-Reference Protocol

OTC Markets publishes transfer agent identity on each issuer's public profile. OTCM Protocol's COO Patrick Mokros serves as President of Empire Stock Transfer (EST), which manages shareholder accounts for hundreds of public companies. The module maintains a continuously updated cross-reference index mapping OTC Markets transfer agent disclosures against the EST client roster.

Issuers identified as current EST clients via this cross-reference are reclassified from cold outreach targets to warm relationship leads, routed to a dedicated internal pathway. This bypass mechanism transforms what would otherwise be a multi-month cold sales cycle into a direct executive-to-executive engagement, leveraging the existing EST relationship to open the tokenization conversation at the board level without initiating contact through generic marketing channels. EST client status applies a 1.5x-2.0x multiplier to the base IDOS score for prioritization purposes.

🔹 8.3.4 Trading Volume and Recovery Probability Modeling

OTC Markets provides historical trading data for each listed security including daily volume, closing price, and last-trade date. The module constructs a Recovery Probability Score (RPS) for each issuer using price trajectory, volume trend, and time-since-last-trade metrics. Companies with low RPS — indicating minimal likelihood of organic market revival — are scored as highest tokenization urgency on the basis that their boards have fewer viable alternatives and are more likely to engage with OTCM's permanent liquidity solution.

Recovery Probability Score formula:

RPS = w1 * PriceDecay + w2 * VolumeDecay + w3 * DaysInactive

PriceDecay = (current_price / peak_52w_price - 1.0), bounded [0, 1]

VolumeDecay = 1 - (avg_30d_volume / avg_prior_year_volume), bounded [0, 1]

DaysInactive = min(days_since_last_trade / 730, 1.0)

Weights (empirically calibrated): w1=0.35, w2=0.30, w3=0.35

RPS > 0.80 => Critical distress — highest tokenization urgency

RPS 0.60-0.80 => Significant distress — high urgency

RPS 0.40-0.60 => Moderate distress — standard outreach

RPS < 0.40 => Early stage — nurture sequence

🎯 8.4 Composite Issuer Distress and Opportunity Score (IDOS)

🔹 8.4.1 Score Architecture

The Issuer Distress and Opportunity Score (IDOS) is the module's primary output for the issuer pipeline function. It synthesizes all EDGAR and OTC Markets signals into a single normalized score for each company in the approximately 15,000 company target universe, refreshed on a continuous basis as new data becomes available through real-time feed monitoring. The score drives automatic prioritization and outreach sequencing without requiring manual sales team intervention.

| Signal Component | Source | Weight | Refresh Cadence | | --- | --- | | Liquidity Distress Index (LDI) — NLP scan | EDGAR 10-K/10-Q | 0.22 | Per new filing | | Tier Degradation Score | OTC Markets | 0.18 | Daily | | Recovery Probability Score (RPS) | OTC Markets | 0.18 | Daily | | Time Since Last Recorded Trade | OTC Markets | 0.12 | Daily | | Shareholder Count (log-normalized) | EDGAR DEF 14A | 0.12 | Annual / on new filing | | 8-K Trigger Score | EDGAR RSS Feed | 0.10 | Real-time | | Form D Urgency Score | EDGAR Form D | 0.08 | Per new filing |

🔹 8.4.2 Pathway Multipliers

The composite IDOS score is further adjusted by an Outreach Pathway Multiplier that reflects the quality of the existing relationship between OTCM Protocol and the issuer's transfer agent. This multiplier transforms the scoring system from a purely algorithmic model into one that captures real-world relationship equity:

| Outreach Classification | Pathway Multiplier | Criteria | | --- | --- | | Direct Referral | 2.0x | EST client confirmed + 10,000+ shareholders of record | | Warm Lead | 1.5x | EST client confirmed; any shareholder count | | Cold Prospect | 1.0x | No existing EST or OTCM relationship on file |

🔹 8.4.3 Dynamic Priority Queue

The IDOS engine maintains a continuously ordered priority queue across all target universe companies. Outreach automation consumes from the head of this queue, dispatching the highest-IDOS issuers into active sequences while maintaining a rolling view of the top 50-100 priority targets at any given time. Companies already in active outreach sequences are excluded from re-dispatch until the current sequence resolves, preventing duplicate contact and ensuring professional engagement cadence.

🤖 AI MODEL TECHNICAL SPECIFICATION

Version: 6.0 | Applies To: Layer 9 — Predictive Marketing AI Module

Model Architecture — IDOS Scoring Engine

The Issuer Distress and Opportunity Score (IDOS) is computed by a gradient boosted decision tree ensemble (XGBoost v2.x). This architecture was selected over neural networks and linear models for three reasons specific to the OTCM use case: (1) interpretability — each scoring decision can be traced to specific feature contributions, supporting regulatory defensibility; (2) performance on sparse tabular data — SEC EDGAR filings produce sparse, irregular time-series data that gradient boosting handles better than sequence models; (3) low inference latency — sub-10ms per issuer scoring enables real-time refreshes across 15,000 companies.

python



# IDOS Model Configuration
model_config = {
    "framework": "XGBoost",
    "version": "2.0.x",
    "estimators": 500,
    "max_depth": 6,
    "learning_rate": 0.05,
    "subsample": 0.8,
    "colsample_bytree": 0.7,
    "objective": "reg:squarederror",   # Continuous score 0.0–1.0
    "eval_metric": ["rmse", "mae"],
    "early_stopping_rounds": 50,
}

# Output: IDOS score ∈ [0.0, 1.0]
# 0.0 = Minimal distress / Low opportunity
# 1.0 = Maximum distress / Highest tokenization opportunity

Training Methodology

Parameter
Specification
Training data window
2015–2024 SEC EDGAR + OTC Markets historical data
Training sample size
~180,000 issuer-quarter observations
Positive class definition
Issuers that engaged with a tokenization/liquidity solution within 12 months
Negative class definition
Issuers that remained in grey/expert market without engagement
Train / validation / test split
70% / 15% / 15% (time-based split — no future data leakage)
Cross-validation
Walk-forward validation — 4 folds, each fold = 2 years
Held-out test period
2023–2024 data (never seen during training)
Primary evaluation metric
AUC-ROC on held-out test set (target: > 0.78)

Feature Engineering

python



# Feature categories and their contribution weights (SHAP values — approximate)
feature_groups = {
    "filing_recency": {
        "days_since_last_10k": 0.18,        # Most predictive single feature
        "days_since_last_8k": 0.09,
        "filing_frequency_12m": 0.07,
    },
    "otc_market_signals": {
        "tier_degradation_score": 0.18,     # Pink → Expert → Grey movement
        "days_since_last_trade": 0.12,
        "market_maker_count": 0.08,
        "bid_ask_spread_trend": 0.06,
    },
    "financial_health": {
        "revenue_trend_4q": 0.11,
        "cash_position_normalized": 0.08,
        "debt_to_equity_trend": 0.05,
    },
    "shareholder_signals": {
        "holder_count_record": 0.10,         # DEF 14A extraction
        "insider_ownership_pct": 0.04,
        "institutional_ownership_pct": 0.04,
    },
}

Model Drift Detection and Retraining Policy

Metric
Monitoring Frequency
Drift Threshold
Response
Score distribution shift (PSI)
Weekly
PSI > 0.20
Alert + manual review
Prediction accuracy on recent cohort
Monthly
AUC-ROC drops > 0.05 from baseline
Trigger retraining
Feature distribution shift
Weekly
KS-statistic > 0.15 for any feature
Alert + feature audit
Engagement prediction vs actual
Quarterly
Precision/recall drops > 10%
Full model retraining

Retraining Cadence: Scheduled quarterly retraining with 6-month rolling training window expansion. Emergency retraining triggered if drift thresholds exceeded. Retrained models require validation AUC-ROC ≥ 0.75 before promotion to production.

EDGAR NLP Pipeline — Failure Handling

python



class EDGARPipelineConfig:
    # EDGAR XBRL schema version dependencies
    SUPPORTED_XBRL_VERSIONS = ["2003", "2009", "2013", "us-gaap-2023"]

    # Failure handling per feed type
    FAILURE_MODES = {
        "edgar_api_timeout": {
            "retry_attempts": 3,
            "retry_backoff_seconds": [5, 30, 120],
            "fallback": "use_last_cached_filing",
            "max_cache_age_hours": 72,
            "score_impact": "mark_filing_recency_features_as_null",
        },
        "xbrl_parse_failure": {
            "fallback": "attempt_html_extraction",
            "secondary_fallback": "use_prior_period_values",
            "score_impact": "apply_0.85_confidence_discount_to_financial_features",
        },
        "schema_version_unknown": {
            "action": "log_to_schema_monitor",
            "fallback": "attempt_best_effort_mapping",
            "alert": "notify_engineering_within_24h",
        },
        "edgar_rate_limit": {
            "action": "exponential_backoff",
            "max_delay_seconds": 300,
            "daily_api_budget": 10000,  # EDGAR allows ~10 req/sec
        },
    }

    # EDGAR schema change monitoring
    SCHEMA_MONITOR = {
        "source": "https://xbrl.fasb.org/us-gaap/",
        "check_frequency": "weekly",
        "alert_on_new_version": True,
        "migration_sla_days": 30,
    }

Investor-Side Behavioral Profiling — Data Boundaries



ON-CHAIN DATA (public — no consent required):
  ✓ Wallet transaction history on Solana
  ✓ Token holdings (ST22 and OTCM)
  ✓ Staking activity and epoch participation
  ✓ CEDEX trading history (public on-chain)

OFF-CHAIN DATA (public professional sources only):
  ✓ Public LinkedIn professional data (job title, firm name)
  ✓ SEC Form D investor lists (accredited investor filings)
  ✓ Public FINRA BrokerCheck data

PROHIBITED DATA SOURCES:
  ✗ Private health, financial, or social data
  ✗ Social media personal accounts
  ✗ Purchased third-party consumer data profiles
  ✗ Any data requiring user consent not yet obtained

Model Governance

Governance Gate
Requirement
Approver
New model promotion
Validation AUC-ROC ≥ 0.75 on held-out test set
CTO sign-off
Feature addition
Privacy review + SHAP impact analysis
Compliance Officer
Training data expansion
Data provenance documentation
Legal Counsel
Production deployment
A/B shadow mode run ≥ 2 weeks
Engineering Lead
Emergency rollback
Triggered automatically if drift threshold exceeded
Automated

💼 8.5 Investor-Side Predictive Intelligence

🔹 8.5.1 Wallet Behavioral Profiling

On the investor side, the module operates a parallel intelligence layer targeting accredited investors most likely to purchase ST22 tokens during the bonding curve phase and post-graduation AMM trading on CEDEX. Wallet behavioral profiles are constructed from on-chain analytics aggregated across the Solana mainnet, supplemented by off-chain identity enrichment where publicly available. All on-chain data is inherently public on the Solana blockchain.

 































On-Chain Behavioral Signal



Investor Profile Implication


Holdings of tokenized equity instruments


Demonstrated preference for regulated tokenized assets


Idle USDC/USDT balance > $25,000


Available dry powdercapital for new position deployment


Historical bondingparticipation curvein participationregulated offerings


Experience with launch-phasecompliant entryDigital Securities mechanics


Governance token holdings across protocols


Protocol-engaged sophisticated investor profile


Average position hold time > 90 days


Long-horizon investor;investor — lower churn riskrisk, better fit for Rule 144 holding period


LP provision behavior in compliant pools


Active market participant;participant — yield-seeking behavior compatible with CEDEX economics


 

🔹9.6.2 8.5.2 Rule 506(c) Compliance Enforcement
All investor-side outreach operates under Rule 506(c) of Regulation D, which permits general solicitation and advertising provided that issuers take reasonable steps to verify that all purchasers are accredited investors. The module enforces this constraint programmatically at the targeting layer — no outreach sequence is triggered for any wallet or identity that has not cleared the accreditation proxy filter.
Verified accreditation records obtained throughfrom the IssuersEmpire PortalStock ComplianceTransfer Gatewayonboarding (Section 7)registry are shared with the investor targeting module via the internal compliance data bus, eliminating duplicative verification workflows for investors who have previously undergonebeen accreditationcertified checksas accredited through prior ST22 participation on the platform.

 
🔹9.6.3 8.5.3 Launch Timing Optimization Engine (LTOE)
Post-deployment analysis of ST22 launch events identified that suboptimal launchoffering timing — competing launches on Solana,launches, adverse macroeconomic conditions, RPC network congestion, and insufficient communityinvestor pre-warming — created measurable drag on bondingST22 curveoffering performance during the critical priceinitial discoverydistribution window. The Launch Timing Optimization Engine (LTOE) generates a Launch Readiness Score (LRS) for each pending ST22 deploymentoffering by modeling five environmental factors:
 

•       Competing token launchoffering schedules — Competing ST22 offerings on SolanaCEDEX within the preceding and following 7-day window
 — investor attention is finite

•       Solana mainnetnetwork conditions — Mainnet RPC congestion index and historical throughput patterns
 — high congestion increases transaction latency and cost

•       Macroeconomic sentiment — Sentiment index (-−1.0 risk-off to +1.0 risk-on) derived from market data feeds
Community engagement— readinessrisk-off scoreconditions from social media and wallet pre-registration metrics
Qualifiedreduce investor appetite for new positions

•       Investor pool depth — countCount of Empire-verified accredited investor wallets confirmed as available and ready to participate — a direct demand-side readiness signal

•       Community engagement readiness — Wallet pre-registration metrics and investor notification open rates from prior outreach sequences

 

LRS outputs below 0.55 trigger a launch deferral recommendation, with the engine proposing an alternative launch window within the following 7-7–14 day period.days. LRS outputs above 0.80 trigger an expedited pre-launch investor notification sequence to maximize pool participation in the offering's first 2472 hourshours.
of
 bonding curve activity.

🛡️9.7 8.6 OTCM Security Token (OTCM STO) Integration
🔹9.7.1 8.6.1 AI Module as Staking Tier Capability
The Predictive Marketing AI Module integrates directly into the OTCM Security Token staking tier architecture as a gated platform capability. Access to AI module features is unlocked through active token staking, consistent with the consumption-based utility model.staking. All benefits require active platform engagement — staking alone, without corresponding module usage, confers no incremental value.
| 




|---|||
|
|




Staking Tier
|

OTCM Staked
|

AI Module Access Unlocked
|



---
Bronze
|

Bronze | 
1,000 OTCM
|

IDOS dashboard read-only access;· top 500 issuers by score
|


Silver
|

10,000 OTCM
|

Full IDOS access + weekly AI-generated prospect report
|


Gold
|

50,000 OTCM
|

Full EDGAR NLP Engine + OTC Markets tier alerts + investor pool analytics
|


Platinum
|

100,000 OTCM
|

Complete suite: real-time feeds,feeds · IDOS priority queue,queue LTOE,· LTOE · wallet profiling,profiling · outreach automation
|

 
🔹9.7.2 8.6.2 Per-Operation Token Burn Mechanics
Individual AI module operations carry OTCM token burn costs, generating deflationary pressure through genuine platform utilization rather than artificial supply reduction.utilization. Burns occurexecute at the smart contract levellevel, are irreversible, and are irreversible. All burn events arepermanently recorded on-chainchain.
and
 publicly auditable.































AI Module Operation



OTCM Burn Cost


EDGAR batch query execution (per 500 records)


1,000 OTCM burned


OTC Markets feed refresh subscription (monthly)


50 OTCM burned


Investor wallet behavioral report (per ST22 launch)offering)


250 OTCM burned


Launch Readiness Score analysis (per deployment)


500 OTCM burned


Full IDOS universe refresh cycle


1,000 OTCM burned


Automated outreach sequence launch (per issuer)


750 OTCM burned


🔹
 8.6.3 Governance Integration

OTCM Security Token holders at Gold tier and above may submit governance proposals governing AI module operating parameters. Governable parameters include: IDOS component weighting ratios; tier degradation alert thresholds; LRS deferral thresholds; outreach sequence cadence parameters; and the composition of the EDGAR NLP distress phrase corpus. Governance over AI module parameters is operationally meaningful — it determines how the protocol allocates commercial attention — and is structurally consistent with the governance model, as it governs platform function rather than profit distribution.
🏰9.8 8.7 Data Moat and Competitive Defensibility
🔹 8.7.1 Proprietary Dataset Accumulation

Every campaign executed, every issuer conversion or non-conversion, every investor wallet interaction, and every ST22 launchoffering outcome feeds back into the module's training data. The module improves continuously with platform scale. After 500 ST22 launches, Groovy Company, Inc. dba OTCM Protocol will have accumulated a dataset on tokenizedDigital securitiesSecurities investor behavior, issuer conversion patterns, and launchoffering timing correlations that no competitor can replicate without operating at equivalent scale and regulatory depth.
This dataset constitutes a structural moat: the AI's accuracy and efficiency improve as the platform grows, which improves commercial outcomes, which accelerates platform growth,growth — compounding the data advantage with each issuer onboarded. Competitors entering the issuer tokenization market without equivalent historical data will face systematically lower conversion rates and higher acquisition costs for an indefinite period.
🔹
 8.7.2
MoatReinforcementDynamics






TheSelf-Reinforcing feedbackCompetitive loopAdvantage
is self-reinforcing and directionally aligned with the platform's core commercial objectives. 
Each successful issuer conversion generates ST22 trading volume. Trading volumeVolume generates transaction fee revenue. Fee revenue fundsdeepens continuedthe AIGlobal moduleUnified development.CEDEX Liquidity Pool. Deeper pool improves secondary market quality, making the platform more attractive to the next issuer and the next investor. Improved AI module accuracy drives faster issuer conversion at lower cost. Lower acquisition cost improves unit economics, enabling reinvestment inThe data infrastructure and model improvement.

Critically: the data moat deepens not as a byproduct of time but as a direct function of commercial success. The competitive advantage is therefore self-funding and self-reinforcing — requiring no incremental investment beyond normal platform operations to maintain and grow.


 
⚡9.9 8.8 Performance Specifications
 






























tracked)













Performance Metric



Specification


IDOS refresh latency (— 8-K trigger event)event


< 60 seconds from EDGAR RSS publication to score update


IDOS full universe refresh cadence


Every 24 hours;hours · continuous partial refresh on real-time events


EDGAR full-text search querythroughput
throughput

1,000 CIKs per batch cycle


OTC Markets tier change detection


Near real-time;time · 15-minute polling interval


Investor wallet profile refresh
cadence

Every 6 hours;hours · event-driven on large on-chain movements


Launch Readiness Score update frequency


Every 4 hours;hours · real-time on Solana congestion events


Priority queue depth
(issuers

15,000+ issuers;issuers · memory-resident with persistent backing store


ESTEmpire cross-reference match latency


< 5 seconds on new issuer ingestion


Outreach sequence trigger delay


< 2 minutes from IDOS threshold breach


 

▪️9.10 8.9 Regulatory and Privacy Considerations
All data sources consumed by the Predictive Marketing AI Module are publicly available. SEC EDGAR is a public database operated for the express purpose of providing universal investor access to issuer filings. OTC Markets Group publishes tier data, issuer profiles, and trading statistics as a market infrastructure function. No proprietary, non-public, or personally identifiable data is accessed without explicit consent in the issuer prospecting pipeline.
Investor-side wallet behavioral profiling operates on on-chain transaction data, which is inherently public on the Solana blockchain. Off-chain identity enrichment is limited to publicly available professional data sources. All investor outreach is compliant with CAN-SPAM Act requirements, GDPR Article 6(1)(f) legitimate interest basis where applicable, and Rule 506(c) general solicitation standards. The module does not store or process payment card data, social security numbers, or health information. All data handling practices are documented in the OTCM Protocol Privacy Policy and subject to the annual compliance audit frameworkframework.
described
 in
Section6.






Rule 506(c) Compliance Gate

The AI module enforces Rule 506(c) compliance programmatically. No investor outreach sequence is triggered for any wallet or identity that has not cleared the module's accreditation proxy filter. Where the wallet has a verified accreditation record on file with Empire Stock Transfer, that record is used directly. Where it does not, the module routes the prospect through the Empire onboarding flow before any offering-specific communication is triggered. This is not a policy — it is a code-level constraint on the outreach automation pipeline.


 








Groovy Company, Inc. dba OTCM Protocol  |  CIK: 1499275  |  Version 7.0  |  March 2026  |  Confidential




 
            
    

            
        
    

    

    

    

        

             Back to top