Introduction

Baselayer's Industry Prediction product provides accurate business classification by analyzing multiple data sources and returning standardized industry codes (NAICS, MCC, SIC) along with confidence metrics and risk indicators.

For detailed information about industry classification systems (NAICS, MCC, and SIC codes), their structures, and how they differ, see our Industry Classification Systems guide .

This guide explains how Industry Prediction works, how the accuracy score is calculated, and how to effectively use these predictions in your verification workflows.

Why industry classification matters

Accurate industry classification is fundamental to:

Risk management: Identifying prohibited or high-risk business types
Regulatory compliance: Meeting card network requirements and AML obligations
Underwriting decisions: Assessing industry-specific risk factors
Portfolio management: Segmenting and monitoring businesses by sector

What you'll learn

How the accuracy score is calculated and what it represents
Which data sources and signals contribute to predictions
How to interpret different accuracy levels and respond appropriately
Best practices for using industry predictions in automated workflows
How to handle edge cases and low-confidence predictions

Understanding the Accuracy Score

The accuracy field represents Baselayer's confidence level in the predicted industry classification, expressed as a decimal value between 0 and 1 (or 0% to 100%).

What accuracy represents:

High accuracy (0.75–1.00): Strong, reliable prediction based on multiple consistent signals
Medium accuracy (0.50–0.74): Moderate confidence; prediction based on limited or mixed signals
Low accuracy (0.00–0.49): Weak prediction; insufficient or conflicting data

Accuracy is not a measure of how "correct" the prediction is, it's a measure of how confident Baselayer is based on the available data. A business with a rich online presence and clear operational indicators will typically receive a higher accuracy score than one with minimal digital footprint.

Data Inputs & Signals

Baselayer's Industry Prediction model analyzes multiple data sources to classify a business. The strength and consistency of these signals directly influence the accuracy score.

Primary data sources

Baselayer's Industry Prediction is based entirely on publicly available information:

1. Website content

Homepage text, product descriptions, and service offerings
Business descriptions and "about us" content
Navigation structure and site organization
Meta descriptions and page titles

When a business has a well-developed website with clear descriptions of their products or services, predictions are typically more accurate.

2. Social media profiles & online listings

Business profiles on platforms like Yelp, Google Business, and Facebook
Review platforms with business descriptions
Professional networks (LinkedIn company pages)
Online directories and local listings

The consistency of industry information across platforms strengthens prediction confidence.

3. Public records & licenses

Publicly available professional licenses (where applicable)
Business registrations in public databases
Industry-specific certifications or permits found online

These provide authoritative signals when available.

4. Business name analysis

Industry-specific keywords in the business name
Common naming patterns for certain sectors
Geographic or service-type indicators

Names like "Mike's Plumbing Co" or "Main Street Dental" provide strong classification signals.

How signals combine

Strong prediction (high accuracy):

Multiple data sources align on the same industry
Website contains detailed, industry-specific content
Business name includes clear industry indicators
Social profiles and online listings show consistent industry information

Weak prediction (low accuracy):

Limited or no website content
Generic business name without industry signals
Conflicting information across sources
Minimal online presence

Understanding the Accuracy Score

The accuracy field represents Baselayer's confidence level in the predicted industry classification, expressed as a decimal value between 0 and 1 (or 0% to 100%).

What accuracy represents:

High accuracy (0.75–1.00): Strong, reliable prediction based on multiple consistent signals

Medium accuracy (0.50–0.74): Moderate confidence; prediction based on limited or mixed signals

Low accuracy (0.00–0.49): Weak prediction; insufficient or conflicting data

How Accuracy is Calculated

Baselayer uses a proprietary machine learning model to predict industry codes and calculate accuracy scores. Here's how the process works:

Step 1: Data collection

When you request an Industry Prediction, Baselayer:

Analyzes the business name and address provided
Discovers and scrapes the business website (if available or found)
Searches for social media profiles and online listings
Looks for public records such as business licenses or certifications

Step 2: Signal extraction

The model extracts industry-relevant signals from each data source:

Text analysis: Keywords, phrases, and descriptions indicating business activity
Structural analysis: Website organization, navigation patterns, and content types
Consistency analysis: How well information aligns across different sources

Step 3: Classification & confidence scoring

The model:

Predicts the most likely NAICS code based on extracted signals
Maps to corresponding MCC and SIC codes
Identifies 4–8 keywords representing core business activities
Calculates an accuracy score based on:
- Signal strength: Quality, quantity and clarity of available data
- Signal consistency: Agreement between different data sources

Step 4: Validation & output

The prediction is validated against:

Known industry patterns and classification rules
Card network risk indicators (Mastercard, Visa)
Historical accuracy for similar business profiles

The final output includes the predicted codes, accuracy score, keywords, and risk indicators.

Interpreting Accuracy Levels

Understanding what different accuracy scores mean helps you make appropriate decisions about automation and review workflows.

High accuracy: 0.75–1.00

What it means:

Strong, reliable prediction based on multiple consistent data sources
The business has a clear digital presence with identifiable industry indicators
High likelihood that the predicted industry is correct

Typical scenarios:

Established businesses with detailed websites
Companies with industry-specific names (e.g., "Smith & Associates Law Firm")
Businesses with consistent information across multiple platforms
Organizations with clear product catalogs or service descriptions

Recommended action:

Safe for automated decisioning
Use directly in prohibited industry checks
Trust the predicted NAICS/MCC/SIC codes for risk assessment

Example: A plumbing company with:

Website showing services, service areas, and pricing
Business name "ABC Plumbing & Heating Services"
Google Business profile with reviews
LinkedIn company page describing HVAC services

→ Accuracy: 0.92 — Very confident the business is in NAICS 238220 (Plumbing, Heating, and Air-Conditioning Contractors)

Medium accuracy: 0.50–0.74

What it means:

Moderate confidence based on limited or mixed signals
Some industry indicators present but not comprehensive
Prediction is likely directionally correct but may lack specificity

Typical scenarios:

Early-stage businesses with basic websites
Generic business names without industry keywords
Limited online presence or sparse website content
Mixed signals across different data sources

Recommended action:

Use with caution in automated workflows
Consider manual review for high-risk decisions
Supplement with additional verification if needed
Focus on broader industry categories (2-digit or 4-digit NAICS) rather than specific 6-digit codes

Example: A consulting firm with:

Simple website with "About" and "Contact" pages but limited service descriptions
Generic business name "Apex Solutions LLC"
No social media presence

→ Accuracy: 0.68 — Likely consulting/professional services but unclear which specific type

Low accuracy: 0.00–0.49

What it means:

Weak prediction based on insufficient or conflicting data
The business has minimal digital footprint
Prediction should not be trusted for automated decisions

Typical scenarios:

Very new businesses without established online presence
Conflicting information across different sources
Generic holding companies or multi-industry operations

Recommended action:

Do not use for automated decisioning
Require manual review or additional documentation
Ask the applicant for clarification on business activities

Example: A startup with:

Recently registered business name "TechCo Inc."
No website or undeveloped domain
No online profiles or directory listings
SOS records with vague purpose: "technology services"

→ Accuracy: 0.42 — Insufficient data to confidently predict industry

Response Fields Explained

Industry Prediction returns several fields that work together to provide comprehensive classification information.

Core prediction fields

code (NAICS code)

The 6-digit North American Industry Classification System code
Most granular and specific industry classification
Hierarchical structure: first 2 digits = sector, 4 digits = industry group, 6 digits = specific industry
Example: 236115 = New Single-Family Housing Construction

title

Human-readable description of the NAICS code
Official NAICS title from the classification system
Example: "New Single-Family Housing Construction"

accuracy

Baselayer's confidence in the prediction (0–1 scale)
Higher values indicate stronger, more reliable predictions
Recommended threshold: ≥ 0.75 for automated use

keywords

Array of 4–8 keywords representing core business activities
Extracted from website content, business descriptions, and other sources
Useful for detecting sensitive terms within allowed industries

Alternative code formats

mcc_codes[]

Array of Merchant Category Code objects
Each object includes:
- code: 4-digit MCC (e.g., "1799")
- title: MCC description
- mastercard_risk: Boolean indicating Mastercard high-risk status
- visa_risk_tier: Visa risk classification (1, 2, 3, or null)

sic_code

Standard Industry Code (4-digit format)
International system, still used by some organizations
Example: 1711 = Plumbing, Heating, and Air-Conditioning Contractors

Risk assessment fields

risk_level

Baselayer's normalized risk assessment: low, medium, or high
Based on industry characteristics, regulatory considerations, and fraud patterns
Independent of card network risk indicators

mastercard_risk

Boolean indicating if Mastercard considers the MCC high-risk
Based on Mastercard's Brand Risk Assessment and Security Enhancement (BRAM) program
true = high-risk per Mastercard guidelines

visa_risk_tier

Visa's risk tier classification for the MCC
Possible values: "1" (high risk), "2" (standard risk), "3" (emerging high risk), or null
Based on Visa's Merchant Category Monitoring Program (MCMP)

Using Risk Indicators

Risk indicators help you align with card network requirements and implement appropriate controls.

When to use each indicator

Accuracy score

Use to determine trustworthiness of the prediction itself
Controls whether you can safely automate decisions based on the industry code
Does not assess industry risk - only prediction confidence

Risk level

Use for general risk assessment across your portfolio
Helps prioritize monitoring and review resources
Based on Baselayer's analysis of fraud patterns and industry characteristics

Mastercard risk & Visa risk tier

Use for card network compliance and payment processing decisions
Essential for payment facilitators, ISOs, and merchant acquirers
May trigger additional monitoring, reserves, or program restrictions

Best Practices

These eight practices will help you effectively implement Industry Prediction in your verification workflows:

Set appropriate accuracy thresholds

Recommended baseline: 0.75 Baselayer's internal validation shows predictions with accuracy ≥ 0.75 are reliable for automated decisioning. However, we recommend adjusting based on your risk tolerance.

Create tiered review policies

Don't treat all accuracy levels the same way:

Auto-approve zone (≥ 0.75):

Trust the predicted industry code
Apply prohibited industry checks automatically
Use risk indicators for decisioning

Review zone (0.50–0.74):

Flag for quick analyst review
Cross-reference with application data
Request clarification if industry is unclear
Use broader NAICS categories (2-digit or 4-digit) for policy checks

Manual verification zone (< 0.50):

Request additional information about business activities
Do not auto-decline based on prediction
Consider alternative verification methods

Establish prohibited industry lists at multiple levels

Use NAICS hierarchy strategically:

2-digit codes (broad sectors): Ban entire economic sectors when appropriate (e.g., 71 = Arts, Entertainment, and Recreation)

4-digit codes (industry groups): Target specific industry groups while allowing related activities (e.g., 7132 = Gambling Industries)

6-digit codes (specific industries): Precision targeting for niche restrictions (e.g., 713210 = Casinos but allowing 713290 = Other Gambling Industries)

Example policy:

Prohibited 2-digit NAICS: ["71", "92"]
Prohibited 4-digit NAICS: ["4539", "8129"] 
Prohibited 6-digit NAICS: ["453998", "812990", "812199"]

This allows flexible risk management without overly broad restrictions.

Monitor keywords for nuanced risk

Even within allowed industries, certain keywords may indicate prohibited or high-risk activities:

Common sensitive keywords:

Pain (opioid risk in healthcare)
CBD, cannabis, hemp (regulated substances)
Vape, tobacco, nicotine (age-restricted products)
Escort, adult, massage (potential TOSA risk)
Crypto, coin, token (virtual currency)
Gambling, betting, casino (gaming activities)

Implementation approach: Create a keyword watchlist and flag businesses if a prohibited keyword is found.

Baselayer has a complete list of recommended keywords that can be shared upon request.

Leverage network risk indicators

For payment processing use cases:

Mastercard compliance: If mastercard_risk: true, the MCC is on Mastercard's Brand Risk Assessment and Security Enhancement (BRAM) program. This may require:

Additional underwriting and monitoring
Higher reserves or transaction limits
Specialized high-risk processing programs

Visa compliance: Use visa_risk_tier to apply appropriate controls:

Tier 1: High-risk, stringent requirements
Tier 2: Standard monitoring
Tier 3: Emerging high-risk, enhanced oversight

Always consult current card network rules, as these change periodically.

Troubleshooting & Edge Cases

When accuracy is unexpectedly low

Possible causes:

Website is new or under development
- Newly launched sites may not have enough content indexed
- "Coming soon" pages provide no classification signals
Business operates primarily offline
- Local service businesses without strong online presence
- Traditional brick-and-mortar operations
Generic or misleading business name
- Names like "ABC Solutions" or "Main Street Enterprises" provide no industry clues
Conflicting signals across sources
- Website describes one type of business, online listings indicate another
- Multiple lines of business with no clear primary activity

When the predicted industry seems wrong

Verification steps:

Review the keywords field: often provides insight into why a particular industry was predicted
Check if accuracy is low: predictions below 0.75 should be treated with skepticism
Look for mixed signals: website may describe multiple activities, model chose the most prominent
Consider franchise vs. corporate distinctions: corporate websites may not reflect individual franchisee operations

When to request Order.WebsiteAnalysis with Industry Prediction

For best accuracy, request Industry Prediction together with Website Analysis:

{
  "options": ["Order.NaicsPrediction", "Order.WebsiteAnalysis"]
}

Website Analysis provides additional content that strengthens industry predictions, especially for:

Businesses with detailed websites
Service-based companies where operations aren't obvious from the name
Companies in ambiguous or overlapping industry categories

Key Takeaways

Accuracy ≥ 0.75 is reliable for automated decisioning in most use cases
Accuracy represents confidence, not correctness - it measures data quality and signal strength
Low accuracy doesn't mean wrong, it means insufficient data to be confident
Combine accuracy with risk indicators (keywords, Mastercard risk, Visa tier) for comprehensive screening
Use NAICS hierarchy strategically, 2-digit for broad bans, 4-digit for targeted restrictions, 6-digit for precision

Additional Resources

API Documentation

Industry Prediction API Reference — Complete endpoint details and request/response schemas
Web Presence API Guide — How to request Industry Prediction alongside other orderables

Related Guides

Web Presence & Orderables: Best Practices — Using Website Analysis, Social Profiles, and Reviews alongside Industry Prediction
Business Search: Best Practices — Core KYB verification and entity matching

Support For questions about Industry Prediction or to report systematic accuracy issues, contact Baselayer support or your account team.