Skip to main content
Industry Prediction classifies a business’s industry based on its online presence: website content, business name, address, and any other signals Baselayer discovers. It returns a 6-digit NAICS code, SIC and MCC codes, confidence score, risk level, keywords, and card network risk indicators.
Note: Baselayer uses 2017 NAICS codes across all industry prediction and classification endpoints. The 2017 revision is the basis for all returned codes, filters, and industry-related fields. If you’re cross-referencing against another system, confirm it also uses the 2017 standard.

​
When to use it

  • You need to screen applicants against prohibited or restricted industry lists
  • You want to validate user-submitted industry classification
  • You are running card network compliance checks (Mastercard BRAM, Visa risk tiers)
  • You want to classify industry for underwriting or risk-based decisioning
  • You need early industry classification before proceeding with a full Business Search

​
How to request it

Industry Prediction is available on both integration paths.

​
Via POST /web_presence_requests

Include Order.NaicsPrediction in the options array. Results are returned inline in the response at industry_prediction.

​
Via POST /searches

Include Order.NaicsPrediction in the options array. Results are not inline - Baselayer returns a NAICSPredictionRequest tracking object in orderables[]. Fetch the result using the URL in that object, or listen for the NaicsPredictionRequest.completed webhook.
The search response will contain:
Fetch the result:

​
Via Order.Enhanced

Order.Enhanced on POST /searches includes Industry Prediction automatically alongside Website Analysis, Social Profiles, Reviews, and expanded officer/address discovery. See Online Presence: Basics for when to use Enhanced vs. individual orderables.

​
Response fields

​
Core fields

​
mcc_codes[] entry fields

A single industry prediction may return multiple MCC codes. Check all of them for mastercard_risk and visa_risk_tier - a single flagged entry is sufficient to trigger compliance controls.

​
sic_codes[] entry fields

​
Interpreting results

Industry Prediction produces two types of signals: industry classification (what does this business do?) and compliance screening (is this business in a permitted sector?). The accuracy score determines how much confidence to place in the result. Importantly, accuracy is not a measure of whether the prediction is correct - it measures how much data was available and how consistently it pointed in the same direction. A low score means insufficient signal, not a wrong answer. Never auto-decline on a low accuracy score; route to manual review instead. What drives accuracy: The model draws on website content, social media profiles and online listings, public records and licenses, and business name analysis. A business with a detailed website, consistent information across platforms, and an industry-specific name will typically score higher than one with minimal online presence. The more signals available and the more they agree, the higher the accuracy. For the full decisioning framework - including confidence thresholds, prohibited industry list structure, keyword scanning, and card network compliance - see Industry Prediction section in the Best Practices guide: tiered review policy, NAICS hierarchy strategy, keyword watchlists, and MCC handling. The key fields to evaluate in every application:
  • accuracy - apply automated decisioning at ≥ 0.75; require manual review below that threshold
  • code - check against your prohibited and restricted NAICS lists at 2-digit, 4-digit, and 6-digit levels
  • keywords[] - scan against your sensitive keyword watchlist for restricted activities within otherwise permitted sectors; a complete recommended keyword list is available from your account representative
  • mcc_codes[].mastercard_risk and mcc_codes[].visa_risk_tier - check all entries; any flagged MCC triggers card network compliance controls
  • reasoning - read when a prediction looks surprising, when a reviewer needs justification, or when debugging why a borderline classification went one way over another. It surfaces the evidence Baselayer relied on and occasionally runner-up codes that were considered.