Strategic Foundation
The Ideal Customer Profile (ICP) is the customer context layer. Same architecture as Brand & Positioning, one layer over. If Brand & Positioning is who you are, ICP is who you sell to — and who simulates them.
The framework — strategy first
Same architecture as Brand & Positioning, one layer over. If Brand & Positioning is who you are, ICP & Audience is who you sell to — Brief Section 6. The same conviction-pillars-voice discipline that grounds your customer-facing positioning grounds your ICP context: vertical fit, buying-committee personas, success metrics, objections, and the words each role uses to describe their problem.
This work produces two deliverables. First, a written ICP — your vertical scorecard, buying committee, Marketing Qualified Lead (MQL) definition, funnel definitions. Second, a population of agents — one per buying-committee role — that you stress-test every important message against before publishing. Same content, two outputs.
ICP work has the same triple-audience problem as brand. Sales reps need to know who to talk to. Marketers need to know who to write for. AI agents need to know who they're producing for. The doc has to serve all three — or it fails one of them, and fails them at scale.
THE ICP, MAPPED TO THE THREE AUDIENCES
Sales — Brief Sections 6.1–2.4: vertical scorecard, MQL definition, funnel stages. The doc that onboards a new Account Executive (AE) onboards every prospecting agent.
Marketing — Brief Sections 6.5–2.7: buying-committee personas, success metrics, objections, proof patterns. Where every prompt and ad knows who it's writing to.
Agents — same Sections 2.5–2.7, transformed into a stress-test Council via Prompt 6 below. Every important message gets reviewed by the Council before publish.
Pre-AI, ICPs were annual artifacts. Marketing built them in Q4, Sales argued about them in Q1, and they atrophied by Q3. Three shifts since:
From annual artifact → living dashboard. Agents watch firmographic signals, hiring patterns, intent data, and closed-won/closed-lost outcomes. ICPs refresh on cadence, not calendar.
From firmographic checklist → behavioral signature. AI scores accounts on job-description language, tech-stack composition, leadership posts, and event attendance — signals richer than industry + headcount.
From "who we sell to" → "who we sell to AND who simulates them." The ICP Researcher Agent pattern (Prompt 6 + Agent below) turns each buying-committee persona into a stress-test agent. A modern ICP is half data, half voices.
If you can't stress-test a message against your ICP before publish, you don't have an ICP. You have a slide.
Marketing at [COMPANY NAME] is ICP-led. Every campaign, every piece of content, and every ad impression should be traceable to a segment we have committed to winning. This work defines the marketing lens on our ICP.
THE THREE LAYERS — TOTAL MARKET → SERVICEABLE MARKET → IDEAL CUSTOMER
Before the ICP, three layers above it — the Total Addressable Market (TAM), Serviceable Addressable Market (SAM), and the ICP itself. Each one narrows the universe by an order of magnitude, and each one earns its keep by removing accounts you should not be spending money against.
| LAYER | WHAT IT MEANS | HOW TO SIZE IT | WHO USES IT |
|---|---|---|---|
| TAM — Total Addressable Market | The full universe of accounts that could ever buy your category, ignoring whether you can reach them, sell them, or serve them. | Top-down: industry reports + analyst sizings. Bottom-up: number of accounts × average Annual Contract Value (ACV) ceiling. Use both and triangulate. | Board decks, fundraising, category-creation arguments. |
| SAM — Serviceable Addressable Market | The slice of TAM that matches your motion, geography, language, regulatory posture, and integration footprint. The accounts you actually could serve today. | Filter TAM by: geography you sell in, motion you run (Product-Led Growth, or PLG, vs. sales-led), languages you support, compliance frameworks you’ve cleared, integrations you ship. | Annual planning, channel mix, headcount math, where reps actually prospect. |
| ICP — Ideal Customer Profile | The slice of SAM where your product wins consistently — highest win rate, fastest sales cycle, lowest churn, biggest expansion. The math actually works here. | Quantitative: 10–20 attributes narrowed to the 3–5 must-haves, weighted to 100 points, validated against closed-won and closed-lost data. The methodology below. | Every paid dollar, every prospecting list, every prompt and agent that downstream agents ship. |
Most ICP failures are SAM failures one layer up. A team writes “our ICP is multi-unit restaurants over 50 locations” when their SAM is actually multi-unit restaurants in the geographies they sell in, with payroll systems they integrate with. The ICP can’t be sharper than the SAM it’s drawn from.
Picking the beachhead is the operating decision that follows from the TAM → SAM → ICP funnel and precedes everything else. A beachhead is a small, specific segment inside your SAM where you concentrate every resource — paid, sales, product, partnerships — until you dominate it. Then you expand to the adjacent segments. The alternative — spray-and-pray across the full SAM — is how most early B2B companies burn cash without ever building a reputation in a specific community.
THE CHARACTERISTICS OF A GOOD BEACHHEAD
A segment is beachhead-worthy when three things are true:
The beachhead checklist below is the 9-step pattern. Run it before the ICP construction methodology — the beachhead chooses which slice of the SAM deserves the deep ICP work.
THE BEACHHEAD CHECKLIST — 9 STEPS
Concrete example: a workforce-management SaaS targets “multi-unit restaurants 50+ locations” as the ICP, but the beachhead is sharper — “multi-unit restaurants 50+ locations headquartered in Texas or Florida, on Toast as their POS, with a VP Ops hired in the last 12 months.” Sales has fast cycles in this beachhead. Marketing’s paid CPL is half the SAM average. Customer Marketing has 4 references inside the segment by month 6. Then year 2 expands outward — to other POS integrations, then other states, then adjacent verticals.
DON'T BUILD ICP FROM SLIDES — BUILD IT FROM SIGNALS YOU ALREADY HAVE
The most common ICP & Audience failure: a CMO writes an "ICP definition" from a positioning deck before mining the signals their company already has. What comes out is the segment marketing wishes they sold to, not the one sales actually closes. The fix is upstream — before the ICP fields below, mine the two signals every B2B company already runs.
Run both in parallel. The intersection is your real Tier 1 — the segment your math actually works on.
| SALES SIGNAL — "WE LOVE LEADS LIKE THIS" | PLG SIGNAL — "THEY GOT VALUE FAST" |
|---|---|
|
Which leads do sales get most excited about? Which companies or buyers have no tradeoffs in the conversation? Who has the fastest buying cycles? |
Who activates fastest after sign-up? Who expands naturally without a CSM nudge? Who has the lowest churn? |
Then validate with customer conversations. Find the common traits. That's your right-to-win segment — not the one the deck claims.
If you haven't run them yet from Brand & Positioning's pre-brand work, run them now. Same five questions, used here to validate the easy-wins hypothesis with the words real customers use:
Listen for the recurring phrase across 4–5 customers. That phrase is the segment language — and it almost always sharpens whatever was in the ICP deck. Save it; Product Marketing & Competitive Intel (Product Marketing) and Customer Marketing both consume it.
THE FOCUS DIVIDEND — WHY THIS IS WORTH THE WEEK
Teams that lock in their ICP commonly see ~2× paid marketing efficiency. What focus enables: cut nearly half the event roadmap, tighten paid campaign targets, sharpen the PLG motion, shut down non-critical campaigns. Demand goes up across the board. The right ICP isn’t a slide — it’s the operating decision that lets you say no to non-fit work.
FROM “WE LOVE LEADS LIKE THIS” TO A 100-POINT SCORE
The signals above (sales-loves, PLG-activates, customer-language) are qualitative. They tell you which segment to focus on. The 8-step methodology below converts that into a quantitative ICP — a weighted scorecard where every account in your pipeline gets a letter grade, A through F, against the same criteria.
The discipline matters because it’s the only way to keep the ICP honest as the company grows. Without the score, “ICP” becomes an opinion war between sales, marketing, and product. With the score, it becomes a fact the whole company defers to.
| STEP | WHAT YOU DO | DELIVERABLE |
|---|---|---|
| 1. Gather data on your best customers | Pull the top 20–30 customers by ACV, retention, expansion, and reference willingness. Capture every firmographic, technographic, and behavioral attribute you can. | A spreadsheet of “best customer” rows with 20+ columns. |
| 2. Analyze the success patterns | Look for what’s common across the best. Industry, company size, revenue, geography, age, growth rate, satisfaction score, retention rate, expansion rate, tech stack, leadership tenure. | A pattern list — the recurring attributes that show up in 70%+ of the best customers. |
| 3. Define 10–20 candidate attributes | Convert the patterns into measurable criteria. “Multi-location” becomes “5+ operational locations.” “Established” becomes “10+ years in market.” “Growing” becomes “15%+ YoY revenue growth.” | A list of 10–20 quantitative criteria. |
| 4. Segment if you have to | If your product genuinely serves multiple distinct segments (e.g., enterprise + mid-market, or two unrelated industries), build two ICPs. One per segment. Do not combine into one mushy ICP. | 1–3 distinct ICPs, each with its own list of criteria. |
| 5. Narrow to the 3–5 must-haves | Workshop with sales, marketing, and CS. The same question for each criterion: “If a prospect failed this one, would we still pursue them?” If yes, it’s a nice-to-have. If no, it’s a must-have. The list shrinks fast. | The 3–5 criteria that genuinely predict best-customer behavior. |
| 6. Assign weights to 100 points | Not all must-haves are equal. Revenue might be 40 points, employee count 30, industry 20, geography 10. The total has to be 100. Use closed-won and closed-lost analysis to ground the weighting in data, not in instinct. | A 100-point weighted scorecard. |
| 7. Grade every account A through F | Score each account against the weighted criteria, sum the points, convert to letter grade. A: 90+ · B: 80–89 · C: 70–79 · D: 60–69 · F: under 60. Sales prospects A and B only. Marketing funds A and B campaigns. C–F respond to inbound only. | A graded prospect list usable by sales today. |
| 8. Validate, refine, biannual review | Score your existing customer base against the new ICP. Best customers should land A/B. Underperforming customers should land C/D/F. Where the grade doesn’t match the reality, the criteria or weights are wrong — fix them. Re-run every six months as the market shifts. | An ICP that holds up under real-world data and stays accurate over time. |
A workforce-management SaaS settles on four criteria after step 5 — annual revenue (40 points), employee count (30 points), number of operational locations (20 points), YoY growth rate (10 points). They evaluate one account, “Acme Inc”:
| CRITERION | WEIGHT | ACME’S VALUE | POINTS EARNED |
|---|---|---|---|
| Annual revenue ($10M–$100M) | 40 | $1.2B — over the range | 40 (still in the right zip code for ACV) |
| Employee count (50–500) | 30 | 1,200 — over the range | 30 (above range but still serviceable) |
| Number of operational locations (5+) | 20 | 3 locations | 0 — fails the must-have |
| YoY growth rate (10–15%) | 10 | 12% | 10 |
| Total | 100 | 80 → Grade B |
Acme scores a B. Sales will work the account but won’t lead a campaign with it. The “number of locations” failure is the tell — Acme is a single-site giant, not a multi-site operator, which means the product solves a smaller part of their problem than it does for a 50-location buyer. We’d still close the deal if they came inbound, but we wouldn’t pay to acquire more accounts like Acme.
EXAMPLE QUANTITATIVE CRITERIA — PICK FROM THIS MENU IN STEP 3
Common attributes that show up in B2B SaaS ICP scoring. Use these as a starting list; your own success patterns dictate which ones make the final 3–5.
The ICP that lives in a slide deck is dead. The ICP that lives in the marketing-automation lead score, the Sales Development Representative (SDR) cadence selector, the paid-media bid modifiers, and the prompts every downstream agent inherits is alive. Once the 100-point scorecard exists, automate it. Inject it into the same Brief that grounds every other area of this playbook.
Set a biannual review on the calendar. Markets shift, your product expands, your best customers change. The ICP that worked in Q1 last year is probably wrong in Q3 this year by 10 points.
Now — with the funnel sized, the beachhead picked, the upstream signals mined, and the methodology understood — define the verticals you’re committing to below. The brief-builder captures your Tier 1/2/3 verticals and the buying committee that grounds every downstream prompt.
Three to four verticals. Be specific. Save to Brief Section 2.1 — fills into every ICP, ABM, and content prompt on the site.
Saved to Brief Section 2.1. Available as [VERTICAL] in every prompt.
Four roles. Specific titles. The Brief stores who you sell to so every downstream prompt — Content, Demand, Lifecycle, ABM, Customer Marketing — knows who it's writing for.
Saved to Brief Sections 6.2–2.5. Each role fills the matching [ECONOMIC BUYER], [CHAMPION], etc. across the site.
What buyers say in their own words — and the events that make them shop for your category.
Saved to Brief Sections 6.6–2.7. Available as [TOP 3 PAINS] and [TRIGGERS] across the site.
PERSONAS MOVED OUT OF ICP — FOR GOOD REASON
The four-role buying committee captured above tells you who sits in the room. The persona deep dive — the seven-dimension profile per role with goals, priorities, challenges, objections, metrics, watering holes, trigger events, plus the propensity-to-buy signal layer — lives in Buyer Personas & Propensity.
Fill in the buying committee here first (titles, reporting lines, pains, triggers). Then take each role to Buyer Personas & Propensity to build the full persona profile + propensity scoring model. The two are designed to be consumed sequentially.
| STAGE | BUYER STATUS | MARKETING GOAL | PRIMARY TACTICS |
|---|---|---|---|
| Unaware | Has problem, doesn't know you | Awareness and problem framing | Paid social, content syndication, events, PR, SEO |
| Aware | Knows you exist | Education and intent building | Content, webinars, LinkedIn organic, email newsletter |
| Considering | Actively researching | Preference creation | Demos, case studies, comparison content, review sites |
| Decision | Evaluating you vs. 1–2 others | Remove friction and build confidence | Customer references, review site badges, proof-of-value |
| Customer | Post-sale | Expansion and advocacy | Customer content, community, review requests |
The prompt pack
Each prompt is a named, named-by-what-it-does deliverable. Click any card to expand the paste-able body. Run against your Operator Brief.
Five copy-paste prompts. Open ChatGPT, Claude, or Gemini. Paste a prompt. Run it. The output of one prompt feeds into the next.
READ THIS ONCE BEFORE ANY PROMPT IN THIS BOOK
These prompts assume you've populated your Operator Brief (the worksheet that lives in /Operator-Brief-Worksheet.docx). When a prompt asks for OPERATOR BRIEF, paste the relevant Brief sections rather than typing context from scratch.
Your output then arrives in your voice, against your buyers, using your differentiators. Not [BRACKETED] generics. The Brief is the difference between an LLM helper and a tool that sounds like you.
Prompt 1
A ranked table of 6–10 verticals scored on a 5-factor rubric, with Tier 1/2/3 recommendation.
Prompt 2
Four-row committee map (Economic Buyer, Champion, Technical Evaluator, End User) with success metric, objection, proof asset.
Prompt 3
Four-stage funnel definitions with entry, exit, and leading indicator for each stage.
Prompt 4
Three hero variants, one per Tier 1 vertical's economic buyer.
Prompt 5
A one-page ICP brief consolidating outputs of prompts 1–4.
Prompt 6
Four reusable system-prompt blocks — one per buying-committee role — that you can drop into Claude Projects, custom GPTs, or feed to the ICP Researcher Agent Agent. Each block makes that persona available as a stress-test agent for every important message.
The agent spec
How to install this agent
Owns the ICP definition + the 100-point scorecard + the A-through-F account grading. Refreshes from closed-won + Win/Loss patterns. The agent that prevents ICP drift from becoming ICP rewrite.
| Task | Frequency | Duration | Output goes to |
|---|---|---|---|
| Daily account scoring | Daily 04:00 | ~10 min | CRM + ABM Account Researcher |
| Weekly closed-won audit | Weekly Mon 11:00 | ~30 min | Sales Director + VP Marketing |
| Weekly ICP digest | Weekly Fri 14:00 | ~20 min | VP Marketing + Sales Director |
| Monthly score-vs-outcome analysis | Monthly 1st | ~90 min | VP Marketing + CRO |
| Monthly Win/Loss theme review | Monthly 15th | ~60 min | VP Marketing + Win/Loss Agent owner |
| Quarterly ICP refresh | Quarterly Q+10 days | ~6 hours | VP Marketing + CRO + Sales Director |
| Quarterly beachhead recommendation | Quarterly Q+12 days | ~2 hours | Sales Director + Field Marketing |
Scheduled (cron-style):
| Schedule | What it runs |
|---|---|
0 4 * * * | Daily account scoring |
0 11 * * 1 | Weekly closed-won audit |
0 14 * * 5 | Weekly ICP digest |
0 9 1 * * | Monthly score-vs-outcome analysis |
0 9 10 1,4,7,10 * | Quarterly ICP refresh |
Event-driven:
| Event | What it runs |
|---|---|
| New account appears in CRM | Score within 24 hours; route tier-1 to ABM Account Researcher |
| Closed-won at C or D grade | Flag for scorecard miss analysis at next weekly review |
| Win/Loss Agent adds a new theme | Map theme to scorecard criteria; add to refresh queue if material |
| Market Sizing flags ICP drift | Audit affected accounts; decide ICP scorecard vs. SAM definition needs update |
| Source | Type | Cadence | Required? |
|---|---|---|---|
| Operator Brief (Sections 1, 2, 3, 4) | Markdown | Read every run | Required — ICP starts here |
| CRM account + opportunity full history | API | Real-time + daily bulk | Required |
| Firmographic data (ZoomInfo / Clearbit / LinkedIn Sales Nav) | API | Daily | Required — scorecard criteria inputs |
| Win/Loss Agent themes | Markdown | Per-interview | Required |
| Revenue Attribution Engine win-rate data | JSON | Weekly | Required for score-vs-outcome correlation |
| Account Intel Hub records | JSON | Live query | Required for grading enrichment |
| ICP scorecard config (criteria + weights) | YAML | Quarterly refresh | Required — core config |
| Output | Format | Target path | Audience |
|---|---|---|---|
| ICP definition document | Markdown | /icp/definition.md | VP Marketing + CRO + every AE |
| Per-account grade + score breakdown | JSON | /icp/scorecard/<account-id>.json | CRM (synced field) + AEs + ABM Account Researcher |
| Weekly ICP digest | Markdown + Slack message | /icp/digest/YYYY-WW.md | VP Marketing + Sales Director |
| Quarterly ICP refresh document | Markdown + diff | /icp/quarterly-refresh/Q<n>.md | VP Marketing + CRO + Board (if material change) |
| Beachhead recommendation | Markdown | /icp/beachhead/Q<n>.md | Sales Director + Field Marketing |
| Scorecard miss alerts | Slack DM + ticket | Slack DM + Linear | VP Marketing + Sales Director |
| Trigger condition | Escalate to | Within |
|---|---|---|
| < 70% of closed-won at A or B grade in a quarter | VP Marketing + CRO + Sales Director | < 7 days (scorecard isn’t working) |
| A-grade win rate not ≥ 2× F-grade | VP Marketing + CRO | Same quarter (no discrimination) |
| Win/Loss surfaces a new must-have criterion not in scorecard | VP Marketing + CRO | < 14 days (refresh queue priority) |
| Scorecard refresh missed Q+14-day deadline | VP Marketing + CRO | Immediate |
| Brand & Positioning misalignment with scorecard | VP Marketing + Head of Brand | Same quarter |
| Platform / tool | Used for | Required? |
|---|---|---|
| Claude Project + Postgres (scorecard table) | Scoring + history | Required |
| Salesforce / HubSpot API | CRM read + grade write-back | Required |
| Firmographic data API | Scorecard criteria inputs | Required |
| Python + pandas | Score-vs-outcome correlation analysis | Required |
| Slack API | Weekly digest + scorecard miss alerts | Required |
| Looker / Mode / Tableau | Grade distribution visualization | Optional but recommended |
Evals — output quality checks:
Hallucination defense — specific checkpoints:
First-run checklist — 5 steps from spec to running agent: