Back to Frameworks

VaryOn Meridian

/ Data Quality

Data Layer

Data quality and value assessment for AI agent consumption

4Dimensions
Geo MeanAggregation
<10msCached Latency
0-100Score Scale

Purpose

Meridian evaluates the quality and value of external data sources consumed by AI agents during inference-time operations. It provides runtime gating via middleware interception, including Model Context Protocol (MCP) server integration, preventing low-quality data consumption at the OS network stack level before socket creation.

In markets where AI agents autonomously purchase and consume third-party data - APIs, feeds, datasets - Meridian provides a bounded, machine-consumable numerical index (0-100) with formal mathematical properties, delivered in real time (<10ms cached) to enable automated procurement decisions and dynamic pricing.

By intercepting tool calls prior to external API execution and denying calls to data sources scoring below configurable thresholds, the system reduces network traffic by 20-40%, eliminates downstream processing of low-quality data, and ensures deterministic agent behavior with guaranteed quality floors.

Core Formula

Score(X)=100×(Sα×Qβ×Dγ×Fδ)1W\text{Score}(X) = 100 \times \left( S'{}^{\alpha} \times Q'{}^{\beta} \times D'{}^{\gamma} \times F'{}^{\delta} \right)^{\frac{1}{W}}

Where S', Q', D', F' = max(dimension, ε) are floored dimensions, ε = 0.01 prevents multiplicative annihilation, W = α + β + γ + δ = 1.0, with default weights α=0.35, β=0.25, γ=0.25, δ=0.15.

Aggregation Rationale

The weighted geometric mean provides three critical mathematical properties: (1) Non-compensatory behavior - a dimension at floor ε = 0.01 produces a multiplicative penalty that cannot be offset by strength in others; (2) Imbalance penalty - balanced dimensions score higher than imbalanced ones with the same arithmetic mean; (3) Constant elasticity - a 1% improvement in any dimension produces a predictable percentage improvement in the composite score.

This aggregation method follows precedent from the UN Human Development Index and reflects the economic reality that data value is multiplicative. A source with S = ε (commodity), regardless of perfect quality (Q = D = F = 1.0), yields Score = 4.27, not the 75.75 that arithmetic mean would produce.

For numerical stability, the system computes in log-space: log(Score/100) = (1/W)[α·log(S') + β·log(Q') + γ·log(D') + δ·log(F')], with precomputed logarithmic values cached alongside the composite score.

Scoring Dimensions

1

Scarcity

35%

Measures inverse availability of functionally equivalent substitute sources at procurement time. The highest-weighted dimension (α=0.35) because monopoly data commands premium pricing.

S(X)=111+ek(nn0)S(X) = 1 - \frac{1}{1 + e^{-k(n - n_0)}}

Logistic function where n = count of equivalent alternatives from independent registry, k = 1.5 (steepness), n₀ = 3 (midpoint where S = 0.50).

  • n=0 (monopoly) -> S=0.989 near-maximum scarcity
  • n=3 (midpoint) -> S=0.500 with maximum rate of change
  • n>=6 (commodity) -> S~ε floor applied
  • Equivalence via schema similarity cos(embed) >= τ₁ and field overlap >= τ₂
  • Independent registry prevents provider gaming
2

Quality

25%

Arithmetic mean of sub-dimensions because partial quality is additively valuable. A dataset with high accuracy but moderate freshness remains useful.

Q(X)=waQa+wfQf+wcQc+wsQs+wvQvQ(X) = w_a \cdot Q_a + w_f \cdot Q_f + w_c \cdot Q_c + w_s \cdot Q_s + w_v \cdot Q_v

Weights: accuracy 30%, freshness 25%, completeness 20%, structure 15%, verification 10%.

  • Accuracy Q_a = 1 - error_rate from verification samples
  • Freshness Q_f = exp(-λ × age) with domain-specific half-lives
  • Threat intel t½ = 24hr, Financial t½ = 5hr, B2B contacts t½ = 30 days
  • Structure from 1.0 (typed+semantic) to 0.1 (unstructured)
  • Verification from 1.0 (regulatory-certified) to 0.2 (unverified)
3

Decision Impact

25%

Single-source marginal degradation using O(1) computation, NOT O(2ⁿ) Shapley enumeration. Measures actual influence on agent outcomes.

D(X)=E×(0.5×De+0.5×Du)D(X) = E \times (0.5 \times D_e + 0.5 \times D_u)

E = essentiality gate [0.05, 1.0], D_e = economic leverage, D_u = uniqueness via Spearman correlation.

  • Essentiality gate: soft [0.05, 1.0] or hard binary {0, 1}
  • Economic leverage D_e = log₁₀(1 + cost) / log₁₀(1 + C_max)
  • Uniqueness D_u = 1 - max correlation with alternatives
  • O(1) bounded computation vs O(2ⁿ) Shapley infeasibility
  • Leave-one-out protocol with bounded K alternatives
4

Defensibility

15%

Legal protection and competitive moat. Despite lowest weight (δ=0.15), floor penalty still applies - non-compliant data creates unbounded liability.

F(X)=wrFr+wlFl+wnFnF(X) = w_r \cdot F_r + w_l \cdot F_l + w_n \cdot F_n

Weights: exclusivity 40%, legal 40%, network effects 20%.

  • Exclusivity F_r from 1.0 (sole-source) to 0.2 (public domain)
  • Legal F_l from 1.0 (regulated+contractual) to 0.0 (public domain)
  • Network effects F_n from 1.0 (proprietary ecosystem) to 0.2 (none)
  • Composite score with inter-dimension correlation penalty
  • Floor penalty at ε=0.01 yields 68.4% score reduction

Tier System

Unrated0-19
Bronze20-39
Silver40-59
Gold60-79
Platinum80-100
77.8 / Gold

Production Tier: Transaction-Grade

Latency: <10ms cached, <500ms fresh computation

Gaming Resistance

Attack VectorDescriptionCountermeasure
Quality InflationProvider reports false accuracy or freshness metadataIndependent verification with randomized sampling, cryptographic audit trail
Alternative Source SuppressionProvider hides existence of functionally equivalent competitorsExternally maintained independent source registry, schema embedding similarity
Decision Impact ManipulationProvider engineers artificial dependency through data structureLeave-one-out testing with bounded K alternatives, objective criticality testing
Freshness SpoofingStale data presented with recent timestampsContent-based aging detection, independent timestamp verification
Selective AvailabilityProvider withholds data during scoring to inflate scarcityMulti-point availability testing, historical availability tracking

Edge Cases

Cold Start (New Source)

  • Provisional score P* using available dimensions only
  • Uncertainty factor U = 1/(1 + observations) applied
  • Score = P* × (0.5 + 0.5U) ensures conservative initial assessment

Zero Alternative Sources

  • Scarcity S -> 0.989 (near-maximum but not 1.0)
  • Sigmoid saturates gracefully with no division-by-zero
  • Monopoly premium reflected in pricing tier

Missing Dimension

  • Apply dimensional floor ε = 0.01 instead of zero
  • Flag as provisional with specific missing indicator
  • Multiplicative penalty ensures conservative scoring

Runtime Gating Failure

  • Fallback to cached score if fresh computation exceeds timeout
  • Degraded mode allows transaction with audit flag
  • Asynchronous score update for next request

Worked Example

Financial Market Data Provider (Options Chain)

S0.82
Two alternative sources (n=2)
Q0.74
High accuracy, 2-hour staleness
D0.88
Critical for options trading strategies
F0.70
Licensed data, standard contract
MERIDIAN = 100 × (0.82, 0.74, 0.88, 0.70)^(1/4) = 77.8
Gold

MCP server intercepts agent tool-call at middleware layer. Score computed in <10ms from cache. Gold tier triggers automatic procurement with usage-based pricing at $0.012 per query. Transaction logged to audit trail with dimensional breakdown for billing transparency.

Use Cases

Meridian could enable data quality scoring across 60+ enterprise applications where AI agents would need to evaluate and select data sources autonomously.

$2.5T+Total Market
60Use Cases
180+Companies

Find a use case for your industry

Showing 60 of 60 use cases

Active Agent Markets

Mature agent ecosystems with immediate adoption potential

DEX Aggregator Optimization

DeFi & Crypto

Score liquidity pool data quality from Uniswap, Curve, Balancer, PancakeSwap. Evaluate slippage predictions and gas cost estimates.

Potential Users:

MEV Bot Coordination

DeFi & Crypto

Assess mempool data quality from multiple Ethereum nodes. Score flashloan opportunity data from lending protocols.

Cross-Chain Bridge Intelligence

DeFi & Crypto

Evaluate oracle price feeds (Chainlink, Band Protocol, API3). Score bridge liquidity and security audit data.

Yield Farming Automation

DeFi & Crypto

Assess APY data accuracy across DeFi protocols. Score impermanent loss predictions from analytics providers.

Programmatic Ad Bid Optimization

Advertising Technology

Real-time scoring of impression quality from SSPs. Evaluate viewability predictions from multiple vendors.

Market:$500B

Ad Fraud Detection Networks

Advertising Technology

Score click/impression authenticity from verification services. Assess bot traffic patterns from multiple detection systems.

Audience Data Marketplace

Advertising Technology

Evaluate first-party vs third-party segment quality. Score cookie match rates and identity graph accuracy.

High-Frequency Trading Execution

Capital Markets

Assess order book depth data from multiple exchanges. Score market microstructure signals for alpha generation.

Market:$1T daily volume

Crypto Arbitrage Networks

Capital Markets

Evaluate price feed latency across CEX and DEX platforms. Score cross-exchange transfer time estimates.

Smart Order Routing

Capital Markets

Assess venue liquidity and execution quality. Score payment for order flow arrangements.

NFT Trading Automation

Web3 & Metaverse

Evaluate metadata accuracy and rarity calculations. Score floor price predictions from analytics platforms.

Potential Users:

GameFi Asset Optimization

Web3 & Metaverse

Assess in-game economy data and token rewards. Score guild performance and scholarship opportunities.

Metaverse Real Estate Valuation

Web3 & Metaverse

Evaluate location traffic data across virtual worlds. Score development potential and rental yields.

API Gateway Intelligence

Cloud Infrastructure

Assess endpoint reliability, latency, and rate limits. Score API documentation quality and versioning.

Serverless Function Orchestration

Cloud Infrastructure

Evaluate cold start times and resource availability. Score function performance across cloud providers.

Enterprise Automation

Enterprise systems ready for agent integration

Freight Capacity Matching

Supply Chain

Score carrier reliability and on-time performance. Evaluate real-time capacity from load boards.

Procurement Bot Networks

Supply Chain

Assess supplier catalog accuracy and pricing. Score vendor compliance and sustainability data.

Potential Users:

Dropshipping Inventory Management

E-commerce

Evaluate stock levels across multiple wholesalers. Score shipping time estimates and costs.

Potential Users:

Commodity Price Discovery

Commodities

Assess spot and futures price data from exchanges. Score weather and crop yield predictions.

Potential Users:
Market:$5T annual

Container Tracking Intelligence

Logistics

Evaluate AIS vessel data and port congestion. Score ETA predictions from shipping lines.

Potential Users:

Alternative Data Alpha Generation

Asset Management

Score satellite imagery for economic indicators. Evaluate social sentiment from multiple NLP providers.

Market:$100T AUM

Credit Risk Data Aggregation

Credit & Lending

Assess bureau data quality and coverage. Score alternative credit signals (utility, rental, telecom).

RegTech Compliance Scoring

Regulatory Technology

Evaluate KYC/AML data from screening providers. Score transaction monitoring alerts accuracy.

Insurance Claims Automation

Insurance

Assess damage estimates from photo AI analysis. Score weather data for catastrophe claims.

Robo-Advisory Data Fusion

Wealth Management

Evaluate market data for portfolio rebalancing. Score ESG ratings from multiple providers.

Clinical Trial Site Selection

Pharmaceuticals

Score patient recruitment potential by geography. Evaluate investigator performance history.

Real-World Evidence Aggregation

Healthcare Analytics

Assess EHR data quality for outcomes research. Score patient registry completeness.

Drug Supply Chain Verification

Pharmaceutical Distribution

Evaluate serialization data for DSCSA compliance. Score temperature excursion data for cold chain.

Medical Device IoT Monitoring

Medical Devices

Assess sensor data reliability from implantables. Score predictive maintenance signals.

Virtual Power Plant Orchestration

Renewable Energy

Score distributed energy resource availability. Evaluate demand response capability predictions.

Grid Balancing Automation

Grid Management

Assess renewable generation forecasts accuracy. Score real-time pricing signals from ISOs.

Oil & Gas Exploration Data

Oil & Gas

Assess seismic survey quality and resolution. Score well log data from drilling operations.

Dynamic Pricing Intelligence

Retail

Score competitor price scraping accuracy. Evaluate demand elasticity predictions.

Potential Users:
Market:$5T annual

Inventory Optimization Networks

Retail Operations

Assess stock level data across channels. Score demand forecast accuracy by SKU.

Customer Data Platforms

Marketing Technology

Evaluate identity resolution accuracy. Score behavioral prediction model outputs.

Product Review Authenticity

E-commerce

Assess review verification from multiple sources. Score sentiment analysis accuracy.

GDS Data Aggregation

Travel

Score flight availability from Amadeus, Sabre. Evaluate pricing accuracy and fare rules.

Revenue Management Systems

Hospitality

Assess competitor rate shopping data. Score demand forecast for dynamic pricing.

Potential Users:

Mobility-as-a-Service Platforms

Transportation

Evaluate multimodal routing options. Score real-time transit data accuracy.

Potential Users:

Logistics Network Optimization

Logistics

Evaluate last-mile delivery partner performance. Score route optimization predictions.

Potential Users:

MLS Data Syndication

Real Estate

Score listing accuracy across platforms. Evaluate comparable sales data quality.

Potential Users:

Automated Valuation Models

Real Estate

Assess property characteristic data. Score neighborhood trend predictions.

Construction Progress Monitoring

Construction

Evaluate drone imagery and site sensors. Score subcontractor performance data.

Property Management Automation

Real Estate

Assess maintenance request prioritization. Score tenant screening data sources.

Potential Users:

Smart Building Optimization

Building Management

Evaluate HVAC sensor data quality. Score occupancy predictions for energy management.

Emergency Response Coordination

Public Safety

Assess multi-agency communication quality. Score incident prediction accuracy.

Defense Intelligence Fusion

Defense

Evaluate OSINT source credibility. Score threat assessment data.

Border Security Systems

Border Control

Assess biometric matching accuracy. Score cargo screening data quality.

Digital Identity Verification

Identity Verification

Evaluate document authenticity scores. Score liveness detection accuracy.

Potential Users:

Future Markets

Next-generation markets developing agent capabilities

Precision Medicine Data Markets

Genomics

Evaluate genomic data quality and coverage. Score phenotype-genotype association data.

Carbon Credit Verification

Sustainability

Evaluate offset project data and additionality. Score third-party verification reports.

EV Charging Network Optimization

Electric Vehicles

Evaluate charger availability and pricing data. Score grid capacity at charging locations.

Social Commerce Analytics

Social Media

Evaluate influencer performance metrics. Score viral trend predictions.

Autonomous Vehicle Networks

Automotive

Assess V2V communication data quality. Score HD map updates and road conditions.

Potential Users:

LLM Training Data Markets

Artificial Intelligence

Score dataset quality, bias, and licensing. Evaluate synthetic data generation quality.

Autonomous AI Agent Coordination

Artificial Intelligence

Assess tool/API reliability for agent selection. Score agent reputation in decentralized networks.

Edge Computing Resource Markets

Cloud Infrastructure

Evaluate edge node performance metrics. Score workload placement optimization data.

Quantum Computing Access

Quantum

Assess quantum processor availability. Score algorithm performance predictions.

Satellite Data Markets

Space Technology

Evaluate earth observation data quality. Score analytics provider accuracy.

Smart City Sensor Networks

Government

Score IoT device data reliability. Evaluate crowd-sourced city data.