← Guides

Start here

The ICP map

A company list to a segmented, tiered map. Four passes, no-credit sources first.

By Mike Hurley, founder of WarmUp. This runs our own pipeline. It is not a demo build.
Version 1.5. Last updated . About 8 minutes.

Motion
Systems
Steps
8
Tools
Clay Search, SaasyDB, DiscoLike, Freckle, Clay, Supabase, Apollo, Prospeo, AI Ark, HubSpot, OpenAI

The ICP map: the route

Left to right: the four ICP facts and the three sources, the rule, passes 1 to 4, the verdict. An orange arrow is a yes. The two lines underneath carry every no to Disqualified and every unknown to Undecided. Each pass is drawn in full beside its step under "How to build it". Go to the passes in detail

  1. The Clay query

    Three sources, one master table

  2. The rule

    Four passes, cheapest first

  3. Pass 1, exclusions

    No provider credits

    Clean?

    • Yes: On to pass 2.
    • No: Disqualified, with a reason.
  4. Pass 2, org structure

    No provider credits

    Structure read?

    • Yes: On to pass 3.
    • Unknown: Undecided, with what is missing.
  5. Pass 3, the real band

    Paid

    In band?

    • Yes: On to pass 4.
    • No: Disqualified, with a reason.
  6. Pass 4, the qualifier

    Paid

    B2B, with evidence?

    • Yes: Assigned, with a segment and a tier.
    • Unknown: Undecided, with what is missing.
    • No: Disqualified, with a reason.
    • Signal flow
    • Pass, yes
    • Fail, no: disqualified
    • Unknown: undecided

    Fail closed: an unknown never counts as a pass.

  • Signal flow
  • Pass, yes
  • Fail, no: disqualified
  • Unknown: undecided
  • Fail closed: an unknown never counts as a pass.
In:
One literal query, three sources, one table keyed on domain.
The rule:
One rule applied four times: a company is disqualified by the cheapest check that can disqualify it, and paid enrichment runs only on companies that survived every free check. Every exit carries a reason. Ours runs in Freckle and Clay. Tools we used: Freckle and Clay run it. Our own Postgres database (Supabase) holds it.
Out:
Every company with a status, a segment, a tier and a reason.

Download the boardOpen it full sizeOne sheet, free to keep and share.

Why it helps

Every other workflow reads the map. The hiring feed asks it whether a company is one we want. The social play reads its flags. The daily list reads its tiers. The map is one table of companies, keyed on domain, holding every company we have ever looked at with a status, a segment, a tier and a reason: assigned to a segment, undecided with the missing fact named, or disqualified with the word that removed it. Nothing is deleted. A status changes when a fact does.

It stays cheap because of one rule applied four times: a company leaves at the cheapest check that can remove it, and paid enrichment runs only on what the sources with no provider credits could not answer. Every exit carries a reason, and every reason is a candidate filter for the source query, so the next pull is cleaner than the last.

Who counts as ICP

Four facts, and each one needs a known answer. A fact known to fail decides, even when another fact is unknown. An unknown fact reads "needs enrichment" and waits for a source. It is never counted as ICP.

  1. Headquartered in the US.
  2. 25 to 200 employees.
  3. Privately owned. Public companies, nonprofits, schools, government bodies and investor firms are out.
  4. Sells to businesses. A company that sells to both businesses and consumers counts.

Why we built it this way

  • The ICP is four facts we can actually fill. In late September we cut it to two, country and headcount, because ownership and business model were blank on most rows. Two days later ownership was filled on 98% of companies and business model on 78%, and the other two facts went back in. A rule can only use a fact that is known on nearly every row.
  • The source query is written literally. When we gave a natural-language builder the same words, it returned nearly three times as many companies as the literal query.
  • Every source lands in one table keyed on domain. With a rule for which source wins, a re-import updates a row instead of adding a twin.
  • Sources with no provider credits go first. Country and exact headcount come from a company preview by domain. Size band, revenue band and industry come from Clay Search, which draws a result quota and not credits. Private or public comes from the company type on its LinkedIn page. Sells-to-businesses comes from labels on records we had already bought. Paid enrichment runs only on a fact those leave unknown.
  • A missing sales role is recorded as a risk and never removes a company. We had that rule the other way round for two weeks and reversed it when it removed agencies we wanted.
  • "Sells to businesses" needs evidence. A label from a source we trust, or a quoted sentence from the company's own site saying businesses pay. No headline or industry name counts.
  • A person reads fifty after every run. A random fifty per segment, beside the agent's fields. A wrong call becomes a rule, a prompt fix or a source filter, never a hand edit to one row.

How to build it

  1. Write the ICP as a short list of facts, and the source query literally. Use only facts your sources fill on nearly every row. Then read the whole result, not the first page: the first page is ranked and looks nothing like the rest.

    The ICP is four facts

    HQ in the US. 25 to 200 employees. Privately owned. Sells to businesses. A fact known to fail decides. An unknown fact reads "needs enrichment", never ICP. The source search is written as a literal query. The social play's 25-to-500 gate is a different rule, by design.

    Three sources, one master table

    SaasyDB export

    Software companies with founding, funding and headcount columns. SaasyDB beats Clay when both name a company.

    DiscoLike agencies pull

    Agencies in six English-speaking markets, similarity-scored, with customers and the existing map excluded at the source.

    Keyed on domain

    Every source lands in one companies table keyed on the registrable domain, so a re-import updates instead of duplicating.

    In: One literal query, three sources, one table keyed on domain. Tools we used: Clay Search, SaasyDB, DiscoLike
  2. Land every source in one table keyed on the company's main domain. Add a column for the source and a rule for which source wins on a duplicate. Never keep one table per cut. Cut with views.
  3. Pass 1, no provider credits. Dedupe, probe for dead and redirected domains, run a word list over name and description, and apply the segment's industry exclusions. Write a reason word on every exit.

    Dedupe and precedence

    Is this domain already here from a stronger source?

    Source: the company's main domain; a fixed order of sources when two name the same company

    Pass
    First sighting, or the winning source.
    Fail
    The losing duplicate: precedence_over on record.

    Dead, redirected, excluded

    Does the site answer, and could we ever sell to them?

    Source: DNS and HTTP probe; a word list on name and description; the segment's industry exclusions

    See note A

    Pass
    Live site, no excluded word or industry.
    Fail
    dead_domain, excluded_word, excluded_industry: the three biggest exits, each with the word that removed it.
    Unknown
    A missing DNS answer is unresolved evidence, not a dead domain: kept.
    Clean?
    Yes:On to pass 2.
    No:Disqualified, with a reason.
    Tools we used: A code step in Freckle. No provider is called.
  4. Pass 2, no provider credits. A people search restricted to the domain for sales, marketing-leader and owner titles. Record what you find. Disqualify nobody on it.

    Who works there

    A sales role, a marketing leader, an owner or delivery title?

    Source: A people search restricted to the domain, two calls, free

    Pass
    Titles found and recorded, not just booleans.
    Fail
    Never on its own. A missing sales role is a recorded risk, not a disqualifier.
    Unknown
    No Apollo record for the domain: a second source, then undecided.

    "Roles never disqualify an agency."

    Our rule book, after we had it wrong once

    The org check records facts the tier rules read later. It removes nobody by itself; that was a rule we got wrong once and reversed.

    See note B

    Structure read?
    Yes:On to pass 3.
    Unknown:Undecided, with what is missing.
    Tools we used: Apollo basic people search, as a Freckle step. Prospeo as the second source.
  5. Pass 3, fill the facts. Country and exact headcount from a company preview by domain (ask ten at a time, then retry each miss alone). Size band and revenue band from Clay Search. Ownership from the LinkedIn company type. Pay for one company enrichment only where a fact is still unknown. Revenue is recorded and never filtered.

    Before paying: four sources, no provider credits

    HQ country, exact headcount
    A company preview by domain (AI Ark). Batch ten, then retry each miss alone.
    Size band, revenue band, industry
    Clay Search. It draws a result quota, not credits.
    Private or public
    The company type on its LinkedIn page.
    Sells to businesses
    Labels on records already bought, then the site read in pass 4.

    The CRM's own enrichment is the cross-check. An empty answer is not a no.

    Enrich once, check the band

    HQ in the US, 25 to 200 employees, privately owned?

    Source: The no-credit sources above first. One paid company enrichment only where a fact is still unknown.

    See note C

    Pass
    All three known and true. Revenue is recorded, never a filter.
    Fail
    HQ outside the US, headcount outside the band, or a public company, nonprofit, school, government body or investor firm.
    Unknown
    A fact still unknown reads "needs enrichment" and waits for a source. Funding goes to the timing signals, never to membership.
    In band?
    Yes:On to pass 4.
    No:Disqualified, with a reason.
    Tools we used: No provider credits: AI Ark previews, Clay Search, the LinkedIn company type, DiscoLike records, HubSpot enrichment. Paid: Apollo organization enrichment.
  6. Pass 4, the paid read. One research call over the home, about and product pages only, filling a fixed set of fields: sells to businesses (with a quoted sentence), what they sell in one sentence, segment. Unknown means undecided, never a pass.

    Read the site, answer a schema

    Does the company sell to businesses, and what does it sell?

    Source: One research call over the home, about and product pages only; a locked schema; a quoted sentence as evidence

    See note D

    Pass
    sells_b2b yes, only if businesses are the paying customer; segment confirmed or reassigned.
    Fail
    not_b2b, not_services: a known no with the sentence that says so.
    Unknown
    No sentence found: undecided, with what is missing on record. Never a pass.
    B2B, with evidence?
    Yes:Assigned, with a segment and a tier.
    Unknown:Undecided, with what is missing.
    No:Disqualified, with a reason.
    Tools we used: A research agent in Freckle. For agencies, a site-reading script on our own OpenAI key.

    Verdict

    Assigned, with a segment and a tier

    Companies across manufacturing, professional services, agencies, startups and SaaS, each with a tier the plays read and the facts that set it.

    Undecided, with what is missing

    Companies waiting on one fact, with the fact named. A missing fact is never a disqualification; the next source or the next signal fills it.

    See note A

    Disqualified, with a reason

    Each with a reason word a person can read and a filter can act on: excluded industry, excluded word, dead domain, not B2B, out of band.

    Pass 5, a person reads fifty

    After a run a person reads a random fifty per segment beside the agent's fields. A wrong call becomes a rule, a prompt fix or a source-filter fix, never a hand edit.

    Contacts, off until the estimate row

    Off until approved

    The map holds companies. People are pulled for tiers 1 and 2 only, under a daily cap, after a written estimate a person has approved.

    Out: Every company with a status, a segment, a tier and a reason. Tools we used: Supabase holds the map. HubSpot shows it to the rep.
  7. Run ten rows and read the spend forecast. Then a hundred. Then the rest. Queued work cannot be recalled, so admit only what you are willing to wait out.
  8. Feed back what you learned. The top exit reasons become filters on the source query, the wrong calls from the fifty become prompt or rule fixes, and the counts go into the ledger.

One row

  1. acme-agency.com from the search
  2. live site, no excluded word or industry
  3. an ops title found, recorded
  4. 140 people, private
  5. "we help B2B brands…" quoted from its own site
  6. agencies, tier 1
Illustrative.

What it returned, and why it is built this way

The map as it stands, and the rules that shape it.

AWhat the map holds
Every company we have ever looked at, in one table keyed on domain: assigned to a segment with a tier, undecided with the missing fact named, or disqualified with the reason word. Nothing is deleted; a status changes when a fact does. (How it reads.)
BTiers the plays read
Three tiers set by the facts, not by a score a person tunes. The daily list reads them every morning to order the calls; the hiring feed ranks a signalled account inside its tier; the social play reads no tier at all. (How it reads.)
CNo-credit sources first, paid on what is left
Passes 1 and 2 use no provider credits. In pass 3, four sources that cost no provider credits fill most of the ICP facts, and paid enrichment runs only on a fact they leave unknown. The shape is the same whatever builder you use. (Design rule.)
DWhy every exit has a reason
The biggest exits are excluded industry, excluded word, dead domain and not B2B. Each reason is a possible fix to the source query, so the next pull is cleaner and cheaper than the last. (How it reads.)

Copy this

The ICP rule, the sources, the qualifier's answer and the exit reasons

# the ICP rule: four facts, decided in this order
disqualified:      the map already says so, with its reason
not_icp:           any fact KNOWN to fail
                   (HQ not US | headcount outside 25-200 | public, nonprofit, school, government or investor firm | sells only to consumers or government)
needs_enrichment:  country, headcount, ownership or business model unknown, and nothing known fails
icp:               HQ US, 25 to 200 employees, privately owned, sells B2B   # B2B and B2C together counts as B2B

# where each fact comes from: sources with no provider credits first, paid last
hq_country:   company preview by domain -> the CRM's own enrichment -> Clay Search
headcount:    company preview by domain (exact) -> Clay Search (a band, stored as its midpoint)
ownership:    the company type on its LinkedIn page
sells_b2b:    labels on records already bought -> the site read in pass 4
rules:        accept a preview only when the returned domain equals yours; an empty answer is not a no

# the qualifier's answer, one fixed shape (unknown is allowed; a guess is not)
sells_b2b:       yes | no | unknown     # yes only when businesses are the paying customer for the main offer
evidence_url:    the page on the company's own site that shows who pays
evidence_text:   the sentence on that page that shows it (quoted, not paraphrased)
what_they_sell:  one plain sentence naming the offer and who buys it
segment:         saas | manufacturing | agencies | professional_services | startup | unknown

# the exit reasons we write, one word per row, so a filter can read them back
duplicate  dead_domain  excluded_word  excluded_industry  headcount_out_of_band  public_company  too_young  not_b2b
undecided reasons name the missing fact: unknown_headcount, unknown_ownership, unknown_b2b, segment_unknown

The same guide as plain Markdown, for an agent or a notes app: /library/icp-map.md

Words we use

The map:
One table of companies, keyed on domain, with a status, a segment, a tier and a reason for each. Nothing is deleted from it. A status changes when a fact does.
Tier:
The rank a company holds on the map because of its fit. A signal moves a company up inside its tier and never changes the tier.
Unknown:
The answer a check gives when it cannot tell. An unknown is never counted as a pass. It waits for a person or for a missing fact.
No provider credits:
A source that answers without drawing down a data provider's paid credits. It is not the same as free: it can still use a search quota or an allowance.
Ledger:
The place the receipts are read. One row per run, read by a person every morning.

Want your list mapped before you buy data?

We build it in your tools and run it with your team.

30 minutes to scope it. Engagements are contracted through WarmUp, Inc.