ICP Account Scoring Runs

Outcome: one central job producing a scored, contactable account list for about 250 credits per 50 accounts — instead of fifteen reps building fifteen versions of it.

Surface: MCP server
Level: Advanced
Uses: find_companies · head_count_growth_rate · job_openings_growth_rate · find_company_techstack · find_people_within_company
Credits: ~250 per 50 accounts
Prerequisite: Lesson 03's guardrails

The gate is the whole point

Score accounts on cheap company-level data first; find people only at the survivors.

Without that gate, a 50-account run becomes a 150-person run through the most expensive tools on the server. With it, the same output costs a fraction.

This is why the job runs centrally. A rep asked to build a scored list will reasonably start by finding contacts, because contacts are what they need — and that ordering costs several times more for a worse list. The gate is a design decision that belongs with whoever owns the spend.

The chain

find_companies

Sourced on your firmographic ICP. Capped.

head_count_growth_rate

Is the company growing? Cheap.

job_openings_growth_rate

Is it hiring into your function? Cheap, and the sharpest timing signal available.

find_company_techstack

Competitor or complement present? Cheap, and positive detections only.

Score and gate

Free. This is where the run gets cheap.

find_people_within_company

Survivors only, capped per company.

The prompt


Using the SyncGTM MCP server, produce a scored account list.

Step 1 — source:
  find_companies: software, 50-500 employees, country in GB/IE/NL/DE.
  Cap at 50.

Step 2 — qualify (all 50):
  head_count_growth_rate
  job_openings_growth_rate for sales and revenue-operations roles
  find_company_techstack

Step 3 — score, and show me the table. Do not find any people yet.
  Score out of 10:
    +3 headcount growth above 10% over 6 months
    +3 two or more relevant roles posted in last 30 days
    +3 competitor tool detected
    +1 country is GB or IE

Step 4 — STOP. I will tell you which accounts to take further.

Then, after review:


For accounts scoring 6 or above:
  find_people_within_company — RevOps or Sales Ops, director and above.
  Exclude recruiting titles. Cap at 2 per company.

Return: company, domain, score, person, title, profile URL.

The explicit stop between scoring and people-finding is the guardrail that makes the cost estimate hold.

What it costs

Step	Rows	Cost	Total
Source	50	~0.1	5
Headcount growth	50	~1	50
Openings growth	50	~1	50
Tech stack	50	~1	50
Score and gate	—	0	0
Find people (15 survivors × 2)	30	~0.3	9
Contact data if needed	30	~1.3	39
			~203–250

Ungated, the same run would take 150 people through the contact tools and cost roughly twice as much for a less useful list.

Distributing the output

The list only creates value if it reaches reps in a form they will use.

Route	When
CRM, with score on the account record	Best — it appears where they already work
A shared sheet, refreshed monthly	Workable
Slack digest of newly-qualified accounts	Good complement, poor as the only route
A table only RevOps opens	Not a distribution channel

Include the reason alongside the score. “Score 8” tells a rep nothing; “Score 8 — posted 3 RevOps roles this month, running a competitor” tells them what to say.

Re-running it

Monthly is the right default for account scoring — firmographics move slowly and the timing signals are refreshed by the weekly digest in lesson 06.

On each re-run:

Append newly matching companies; do not re-source the whole universe
Re-score everything, since scoring is free
Report newly-crossed accounts, not the full list
Only find people for accounts that newly crossed

Do this now

Write your ICP as filters

Enumerated countries, size bands, industries.

Write the scoring model

Four or five criteria, whole numbers.

Run the two-stage prompt

With the stop after scoring.

Review the scored table yourself

Before any people are found.

Take the survivors through people-finding

Capped.

Distribute with reasons attached

Not just scores.

Schedule the monthly re-run

With append-only sourcing.

Check your work

The prompt stops between scoring and people-finding
Every scoring input is cheap and company-level
Reps receive the reason, not just the number
The re-run appends rather than re-sourcing
Actual cost matched the estimate

Where this breaks

A scored list distributed without reasons gets ignored. A rep looking at “Acme — 8” has no idea what to say, so they either research it themselves (duplicating the work you just did centrally) or skip it. Carrying the two or three facts behind the score costs nothing and is the difference between a list that gets worked and one that does not.

Further automation

Sync the score and its reasons to the CRM account record so it is visible without anyone opening another tool. A score that lives only in a shared sheet changes nobody’s behaviour; one on the account record changes what gets worked first.

Next lesson

05 — Inbound lead triage, the second central job.

Reference for this lesson: find_companies, head_count_growth_rate, job_openings_growth_rate, TAM Sourcing.