Your program already has a category tree, a partner-tiering scheme, and fields your CRM expects. Instead of forcing your universe into our standard signal set, we engineer bespoke fields to your exact taxonomy, calibrate them on a reviewed sample, and classify 120M+ classified domains straight into the schema your team already works in.
A generic "is_icp + confidence" row is a fine start, but mature programs run on their own vocabulary. When the deliverable already speaks your schema, it drops into your stack with no re-mapping and no lossy translation.
Multi-level taxonomies, sub-verticals, and content-type buckets specific to how you organize partners — not a flat industry label. Each publisher is placed in your tree, with confidence.
Your "premium publisher," your "editorial vs. deals" line, your tier thresholds — encoded as extraction rules and calibrated to match your team's judgment, not a generic default.
Output columns named and typed to match your CRM or PIM. The delivered CSV imports without a mapping spreadsheet — the fields already line up.
Custom classification is field engineering, not a checkbox. Each of your fields becomes an extraction task with a written definition, a confidence model, and a calibration pass before anything runs at scale.
We take your category tree, tier rules, and field list — a spreadsheet, a data dictionary, or a schema export — and map every field you want populated.
Each field becomes a precise, written definition of what evidence on a site makes it true — so "premium" or "enterprise-focused" means something specific and repeatable, not a vibe.
We classify a sample against your fields, your team reviews it row by row, and we tune the definitions until the calls match your judgment at the precision you set.
The calibrated field set runs over the whole vertical from 120M+ classified domains — each domain read once and populated across every one of your custom fields with confidence scores.
Mirrors and country variants collapse to one canonical domain, keyed the way your CRM expects, so records join cleanly on import.
The CSV arrives with your field names, your value sets, and a confidence column per field — ready to load without a translation layer.
Every engagement keeps the standard signal set as a foundation; custom fields sit on top. The teal chips are always available; the indigo chips are examples of bespoke fields defined per engagement.
The indigo fields are illustrative. Whatever your team tracks — a compliance flag, a brand-safety tier, a regional bucket — can be defined as a field and calibrated the same way as any standard signal.
The mapped example below is drawn from real classification work. Because the publisher is an independent site, we follow standard confidentiality practice on public pages: the domain is withheld and details generalized, so the profile is intentionally not traceable to a specific site — including through a web search. That protects the publisher without changing the underlying data. Your pilot and client deliverables carry the actual domains populated across every custom field, so each classification can be verified directly.
A generalized publisher mapped to an illustrative custom taxonomy — the standard fields plus the client's bespoke ones, each with confidence. Domain withheld; values illustrative.
Instead of "home & garden," the publisher lands three levels deep in the client's own category tree — the granularity their routing and reporting actually run on.
The lower-confidence fields — team-size band, premium flag — are exactly the ones to spot-check first. Confidence makes a custom field auditable, just like a standard one.
A custom field isn't a looser field. Every one you define gets a written extraction rule, a confidence model, and a reviewed-sample calibration — the same rigor as the standard signal set that reached ~96% precision in the reference engagement.
Reference engagement (a large-scale production run on our own classification platform): a 30M dataset → 1.73M travel-related domains → 1.2M ICP-screened → ~96% precision on a reviewed sample. Custom-taxonomy figures are illustrative.
One row per canonical domain, columns named and typed to your schema, with a confidence score beside every custom field so your team can trust and audit each classification.
| Column | Example value | Origin |
|---|---|---|
canonical_domain | your CRM key | Standard — dedup join key |
category_l1 / l2 / l3 | Home / Smart-home / Security | Custom — your category tree |
content_type | editorial_reviews | Custom — your value set |
premium_flag + confidence | true · 0.80 | Custom — your definition |
team_size_band + confidence | 2–5 · 0.71 | Custom — calibrated bands |
is_icp + confidence | fit · 0.90 | Standard — kept alongside custom |
The free pilot configures the full pipeline for your ICP and returns your first 20 qualified publishers plus a full-run projection — the fastest way to scope a bespoke custom-taxonomy build. No cost, no obligation.
Request a Free Pilot