🏭 Industry Classifier - NAICS & SIC Code Assignment
Pricing
Pay per event
⚡ Assign a company a NAICS and/or SIC industry code with confidence and evidence, no LLM. ✅ Walks an embedded sector-then-detail code tree, scores with TF-IDF + phrase hits, and uses negative keywords so "SaaS for dentists" lands in software, not dentistry.
The companies to classify, as an array of flat objects with a name, description/tags, and/or a url. Combine freely with companyDatasetIds.
[ { "name": "Bright Smile Dental Clinic", "description": "A friendly dental clinic offering checkups, cleanings, and orthodontist services for the whole family.", "tags": [ "dental", "healthcare" ] }, { "name": "DentaFlow", "description": "SaaS practice management software platform for dentists and dental clinics, built as a cloud app with an API.", "tags": [ "software", "saas" ] }, { "name": "Joe's Diner", "description": "A full-service, sit-down restaurant serving breakfast, lunch and dinner off a bistro menu.", "tags": [ "restaurant" ] }]Apify dataset IDs to pull companies from, in addition to (or instead of) the inline companies list.
If true, fetch the page at each company's url field and append its meta description + visible text to the description before classifying. If false (default), classification uses only the text you already supplied.
Field holding the company's name. Carries the strongest classification signal.
Field holding the company's longer description text (what fetched site text, if any, is appended to).
Field holding the company's website URL, used only when fetchSite is true.
Field on each company to use as its output ID. Defaults to an auto-generated C000001-style ID when omitted or blank.
If the best-matching sector, or the best-matching detail code within the accepted sector, scores below this confidence, classification stops there and the company is sent to the review queue instead of a forced guess. Expressed 0-100; 55 means 0.55.
How hard a matched negative keyword (e.g. "software"/"saas" on Offices of Dentists) crushes that code's score. 0 disables negative keywords entirely; 100 zeroes the score outright the moment one is found. Expressed 0-100; 80 means the score is multiplied by 0.2.
Relative weight of TF-IDF-ish cosine similarity in the score. Weights are normalized against the other three weights, so only the ratio between them matters.
Relative weight of how much of a code's own vocabulary (title + keywords) is found in the company text.
Relative weight of character-trigram Dice similarity, which tolerates typos and word-order drift that token matching misses.
Relative weight of exact multi-word keyword-phrase matches (e.g. "point of sale", "custom software development") — the strongest single signal when it fires, and what is surfaced back to you as evidence.
Timeout per company website fetch, only used when fetchSite is true.
Include the approximate crosswalked SIC code(s) alongside the NAICS detail code on each classified row.