The method

Read the whole vertical. Keep the publishers that fit your ICP.

We don't search directories — we start from 120M+ classified domains, pull every domain in your industry, and have an LLM read each site against your ideal publisher profile. Every signal gets a confidence score; every claim is calibrated on a reviewed sample first.

120M+
classified domains as the source
20
publisher signals per domain
30
affiliate industries covered
The pipeline

Six steps, from the whole web to your recruitment list

Each step is explicit and inspectable. Nothing is a black box: you see the universe size, the ICP definition, the calibration sample, and the confidence on every field.

1

Universe pull by category

Your vertical maps to categories in our 120M+ domain classification database. We pull every active domain in scope — including the long tail no directory or outreach tool has ever profiled.

2

ICP as a signal set

With you, we translate "our ideal publisher" into required, preferred, and disqualifying signals from a 20-signal vocabulary. The definition is written down before anything runs.

3

Reviewed-sample calibration

We screen a sample, you review it domain by domain, and we tune the signal definitions until measured precision meets the bar. Only then does the full run start.

4

Full run with confidence

The LLM reads every domain in the universe and extracts each signal with its own confidence score. ICP fit is computed from the signal set — not guessed from a keyword match.

5

Dedup & canonicalization

Mirrors, redirects, and country variants collapse to one canonical domain per publisher. Your outreach team never contacts the same site twice under different names.

6

Delivery & refresh

A structured CSV, one row per canonical domain, ready for your partner CRM. Optional refresh runs catch new publishers, dead domains, and pivots on a set cadence.

The instrument

The 20-signal vocabulary

Every domain is scored against the same fixed vocabulary of publisher signals, each extracted from what the site actually publishes and shipped with a confidence score. Your ICP is a selection from this list — nothing vague, nothing hidden.

Relevant to industry Blog-like structure Review / comparison content Coupon / deals content Run by an individual or duo Small editorial team Affiliate links present Display ads Newsletter Active social presence Fresh content (6 months) Evergreen guides Price comparisons Tools / calculators Video content Community / UGC Niche focus Consumer audience Own commerce (disqualifier) English language

Need a signal that isn't here — "already works with affiliate networks", "accepts sponsored posts"? Custom signals are defined per engagement and calibrated the same way.

ICP = signal set

A worked example: flight-metasearch advertiser, travel vertical

An ICP is not a prose description — it is a precise set of required, preferred, and disqualifying signals. Here is a typical profile for a flight-metasearch platform recruiting content publishers.

ICP: Independent travel content publisherfor a flight-metasearch affiliate program
Required — all must hold
Relevant to travel Blog-like structure Individual or duo Evergreen guides
Preferred — boost the fit score
Affiliate links present Newsletter Fresh content
Disqualifying — excluded on match
Own commerce (sells inventory itself)

Why the disqualifier matters

An OTA that sells flights itself looks travel-relevant to a keyword filter, but it is a competitor, not a publisher. The own-commerce signal removes that whole class of false positives structurally.

How fit is computed

Each signal is extracted with a confidence score, and the ICP verdict is derived from the required / preferred / disqualifying logic. You can slice deliveries by confidence — highest-certainty domains first.

Calibration

Precision is measured before it is promised

Every engagement runs the same loop: screen a sample, have your team review it, adjust the signal definitions, repeat. The full run only starts once the reviewed sample clears the precision bar you set.

1. Sample screen

A representative slice of the vertical is screened with the draft ICP. Typical first-pass errors surface immediately — sites that match words but not intent.

2. Your review

Your affiliate team marks each sampled domain: fit or not fit. Disagreements are the valuable part — they tell us which signal definitions to sharpen.

3. Tune and repeat

Signal definitions and thresholds are adjusted and the sample is re-screened. The loop runs until measured precision holds — then, and only then, the full universe runs.

Why this order matters: tuning on a reviewed sample costs hours; discovering errors after a million-domain run costs the engagement. We never skip the loop.
Proven at scale

One engagement, in numbers

For the affiliate program of a leading airline metasearch platform, the pipeline processed the travel vertical end to end. The client reviewed a sample of the output against their own judgment.

0
domains in the source dataset
0
travel-related domains identified
0
domains ICP-screened in one run
0
precision on the client's sample review
30M
source domains
1.73M
travel-related
1.2M
ICP-screened
~96%
precision confirmed
Coverage

30 affiliate industries, 5 advertiser ICPs each

Every industry ships with five pre-built ICPs for its typical advertiser clients — travel alone covers flight metasearch, hotel booking, tours & activities, travel insurance, and travel gear. Your ICP starts from a template and is calibrated to your program.

TravelPersonal financeInsuranceInvestingCryptoSoftware / SaaSVPN & securityWeb hostingEducationFashionBeautyHealth & wellnessFitnessFood & cookingHome & gardenDIY & toolsPetsGamingConsumer techAutomotiveOutdoor & adventureBaby & parentingWeddingsPhotographyBooks & mediaJobs & careersReal estateTelecomGifts & occasionsLuxury & jewelry
The deliverable

What lands in your inbox

A deduplicated CSV, one row per canonical domain, importable into any partner CRM or outreach stack. Traffic is reported through reproducible proxies — never invented visitor counts.

Column groupContentsNotes
IdentityCanonical domain, site name, language, country focusOne row per publisher after dedup
ICP verdictFit result, fit score, per-signal valuesComputed from your required / preferred / disqualifying set
ConfidenceA confidence score on every extracted signalSlice deliveries by certainty
Traffic proxiesCrUX presence, rank groupReproducible signals only — no invented visitor numbers
Custom fieldsAnything defined for your engagementCalibrated on the same reviewed sample
Refresh: vertical universes age — new sites launch, blogs die or pivot. Optional quarterly or monthly refresh runs deliver the deltas in the same format. See pricing.

See the method run on your vertical

The free pilot configures the full pipeline for your ICP and delivers the first 500 qualified publisher domains — plus a projection of the complete run. No cost, no obligation.

Request a Free Pilot