The ICP Qualifier

Help me build a keyword exclusion list for filtering a list of
companies.

I sell [WHAT YOU SELL] to [WHO YOU SELL TO].

Generate single words and short phrases that, if they appear in a
company's one-line description, mean that company is definitely not
my customer.

Rules:

- Every term must disqualify on its own. If a term could appear in
  the description of a company I want, leave it out.
- Include the industry names and the words only people in that
  industry use.
- No term shorter than six characters unless it is unambiguous.
- Use only lowercase letters, spaces and hyphens. No punctuation.
- Give me 60 to 120 terms, grouped by industry.

Show me the groups and wait for me to remove any I disagree with.

When I say done, give me this Google Sheets formula with every
remaining term inside it, separated by the | character, ready to
paste:

=IF(REGEXMATCH(LOWER(C2),"term one|term two|term three"),"Eliminate","Keep")

=LEFT(VISIT(B2),5000)

=IF(LEN(E2)<100,"EMPTY","OK")

Build my qualifying prompt. Do this as a conversation, one question
at a time. Do not skip ahead.

MY COMPANY: [your homepage URL]

GOOD FIT CUSTOMERS:
[domain]
[domain]
[domain]

PART 1. Read my homepage and each customer homepage. Then ask me:
- What employee headcount range I sell into. Offer ranges to pick
  from.
- Whether I want companies that sell to consumers through a business.
  Example: a software company whose customers are dentists, whose
  customers are patients. Yes or no.

PART 2. Write these two lines and show me:
WHAT I SELL: one sentence, taken from my homepage.
WHO I SELL TO: one sentence describing the company, not the person.

PART 3. Disqualifiers. Show me this list and tell me which ones you
think apply to me from what you read. I will confirm or remove.

1. Sells to individual consumers rather than to businesses.
2. Sells to businesses whose own customers are individuals. Dental
   and medical practices, veterinary clinics, salons, gyms,
   restaurants, auto dealers, real estate agents, home services,
   insurance brokers, law firms.
3. An agency, consultancy or freelancer selling services rather than
   a product.
4. Sells only to enterprise. The signs are no pricing page, demo
   requests as the only call to action, compliance badges above the
   fold.
5. A holding company, investment firm, non-profit or government body.
6. A marketplace or directory rather than a company with its own
   product.
7. A placeholder page, a domain-for-sale page, or a contact form with
   no product on it.

Then propose five to eight more that are specific to me: company
types that would look like a fit on a homepage but are not my
customer. Look at what my good fit companies have in common, then
list the nearby types that look similar but are not the same. For
each, say what makes it look like a fit and what rules it out. Ask me
to delete the wrong ones.

PART 4. The one-sentence test. Write a single yes or no question that
can be answered in three seconds by reading a homepage. It must be
yes for every good fit customer and no for every disqualifier I kept.
Check it against them before you show me. If it fails, rewrite it and
check again. Tell me when it passes. Do not use words like
innovative, scalable, modern or growing.

PART 5. Examples. From the attached file, pick six companies: two
that clearly pass, two that clearly fail, two that are hard to call.
Show me each company name and the first 300 characters of its
homepage text. Do not tell me your verdict. Ask me for mine, MATCH or
NO, and one line on why.

PART 6. Assemble. Output my finished qualifying prompt as one block
in exactly this format, with nothing before or after it:

----------------------------------------------------------------
You are qualifying companies against my ICP. Each row gives you a
company name and the text of its homepage.

## THE PROFILE

WHAT I SELL: {from Part 2}
WHO I SELL TO: {from Part 2}

## RULES

Judge each company on its homepage text alone. Never judge by
company name or domain.

Homepages carry customer logos, case studies and footer copy naming
other people's industries. Judge the company by what it sells, not
by every industry printed on the page.

If the homepage text is blank, verdict is UNCLEAR. Run no checks.

## DISQUALIFIERS

Answer NO if any of these apply. Stop at the first one that does
and record which.

{every disqualifier I kept in Part 3, numbered}

## THE DECISION

If nothing above applies, answer this from the homepage:

{the one-sentence test from Part 4}

Yes is MATCH. No is NO.

A company that might fit is not a MATCH. If the page is real but
you cannot tell who it serves, verdict is UNCLEAR.

## CALIBRATION

Judge every company the way I judged these.

{for each of the six from Part 5:
Company: name
Signals: the two or three things in the text that mattered
Verdict: my verdict
Why: one sentence, in my reasoning}

## OUTPUT

Write a new CSV with every original column plus three more:

verdict: MATCH, NO or UNCLEAR
reason: one sentence, twelve words maximum
failed_on: the number of the disqualifier that caught it, or N/A

Process every row. Do not skip, reorder or edit existing data. If
the input has 3,000 rows, the output has 3,000 rows.
----------------------------------------------------------------

Pick 30 companies from the file. Show me each company name and the first 300 characters of
its homepage text, numbered 1 to 30. Ask me for a verdict on each,
MATCH or NO.

Then run my qualifying prompt on those 30. Show me only the rows
where your verdict differs from mine, with your reason for each.