Try OnsaMost founders spend weeks on their ICP. Some hire consultants. Others fill out 30-field templates that gather dust in Google Drive.
We watched 47 real conversations where B2B founders built their Ideal Customer Profile from scratch. The pattern was consistent: the best ICPs take 15 minutes to draft and 3 iterations to nail down.
The first version is never right. That’s the whole point. An ICP isn’t a document you write once — it’s a filter you sharpen by seeing who it catches.
Here’s the exact framework that emerged from those conversations.
An ICP is not a persona. It’s not a TAM analysis. It’s not a 40-slide deck about market segments.
It’s a filter. A set of criteria that tells your sales team — or your AI prospecting tool — exactly who to call first and who to skip.
A fintech API startup came in saying “we sell to payment companies.” That’s not an ICP. That’s a market. Ten minutes later, their ICP looked like this:
European fintech platforms, under 500 employees, integrating 3+ payment providers, with a Head of Payments or VP Engineering.
Notice what changed: industry became specific (fintech platforms, not “payment companies”), size got bounded, a technical signal was added (3+ integrations — indicating complexity they solve), and the buyer role was defined.
That’s the difference between “who could buy” and “who will buy.”
The best ICPs we’ve seen share three traits: - Specific enough to search. You should be able to type it into LinkedIn Sales Navigator and get under 5,000 results. - Narrow enough to message. Your outreach should reference something specific about why they fit. - Broad enough to sustain. You need at least 200-500 companies in your addressable list, or you’ll run out of pipeline in a month.
Every ICP we’ve seen work follows a three-column structure. This isn’t theoretical — it’s what emerged from watching real founders iterate.
These are non-negotiable. If a company doesn’t match all of these, they’re out.
Industry/Vertical — SaaS, Fintech, Logistics — Your product solves industry-specific problems
Company Size — 50-500 employees — Below 50 = no budget. Above 500 = enterprise sales cycle
Geography — US, Western Europe — You need to support their timezone and compliance
Buyer Role — VP Sales, Head of Growth — This person has the problem AND the budget
Stage/Funding — Series A-C — Pre-seed can’t afford you. Post-IPO has internal tools
These don’t disqualify, but they predict higher conversion.
Recent funding round — They have budget and pressure to grow
Hiring for your buyer role — The function is a priority
Using complementary tools — CRM, marketing automation = sales-mature
Growing headcount (20%+ YoY) — Active growth = active pain
Competitor mentioned on site — They’re in the market
This column prevents your biggest time waste.
Competitors — They’re building the same thing
Agencies/Consultancies — Different buying pattern, high churn
Companies with internal tools — They’ve already solved this
Wrong geography (no office) — Can’t support them
Pre-revenue startups — No budget, long sales cycle
One of the most common mistakes we see: founders skip Column 3 entirely. They spend weeks chasing agencies that look like perfect fits on paper but never convert. Define who you’re NOT selling to. It saves more time than defining who you are.
Here’s a real example of how ICP refinement works in practice. The details are anonymized, but the pattern is exact.
Iteration 1: Too broad. A logistics technology company started with: “We sell to automotive companies shipping between Mexico and the US.”
That returned 2,000+ companies. Way too many to prioritize. The founder couldn’t tell which ones to call first.
Iteration 2: Geographic focus. Narrowed to: “Automotive companies in Texas with cross-border shipping operations.”
Better. Down to ~400 companies. But still mixed — some were manufacturers, some were parts distributors, some were dealerships. Very different buyers.
Iteration 3: Role filter. Added: “…with a VP of Finance or Supply Chain Director.”
Now they’re targeting the person who actually feels the pain of cross-border logistics costs. The list dropped to ~150 companies. Every one was a real prospect.
Iteration 4: Overcorrection. Tried to narrow further: “…who attended specific trade shows.”
List dropped to 12 companies. Too narrow. Backed up to Iteration 3 and added funding/growth signals instead.
The lesson: your ICP emerges from seeing wrong leads. You don’t figure it out by thinking harder. You figure it out by running a search, looking at the results, and asking “why are half of these wrong?” This is the same iterative pattern that drove $3B in market cap for C.H. Robinson — they didn’t automate perfectly on day one, they refined through real data.
This loop — search, review, refine — typically takes 3 rounds and about 15 minutes total. The founders who spend weeks on it are usually stuck in Iteration 1, trying to get it perfect before testing.
An ICP that lives in a document is useless. It needs to translate directly into search filters.
Here’s how each ICP criterion maps to LinkedIn Sales Navigator (the most common B2B prospecting tool):
Industry — Industry filter — Use LinkedIn’s taxonomy, not your own
Company Size — Company headcount — Ranges: 11-50, 51-200, 201-500, etc.
Geography — HQ location — Can filter by country, state, or metro
Buyer Role — Job title + Seniority — “VP” + “Sales” captures VP of Sales
Funding Stage — — — Not in Sales Nav. Use Crunchbase or enrichment tools
Tech Stack — — — Use tools like BuiltWith or enrichment APIs
Hiring Signals — — — Check job boards or use intent data
Notice the gaps. Sales Navigator doesn’t have filters for funding stage, tech stack, or hiring signals. These are Column 2 criteria — you need enrichment tools or AI sales assistants to layer them on.
This is exactly why many teams are moving from manual Sales Navigator searches to AI-powered prospecting. The AI can cross-reference LinkedIn data with Crunchbase funding, job postings, and technographic data in a single query. Tools like Clay focus on waterfall enrichment across 150+ providers, while others take a conversational approach where you describe your ICP in plain English.
These come directly from real conversations. Every single one has happened multiple times.
No, you don’t. If your ICP is “any company with a sales team,” you have a market, not an ICP. The whole point is to narrow. Start with your last 5 closed deals. What did they have in common?
Startups targeting Fortune 500 companies with a $50/month tool. Enterprise vendors going after 5-person startups. Match your price point and sales motion to your target size. If you don’t have a dedicated enterprise sales team, don’t target enterprise.
“We sell globally” is a strategy for Salesforce, not for a 10-person startup. Geography affects language, timezone, compliance, payment methods, and cultural sales norms. Pick 1-2 geographies and dominate them first.
Without exclusions, you’ll waste 30-40% of your prospecting time on companies that were never going to buy. Agencies, competitors, companies with internal tools — define these upfront.
An ICP with 15 must-have criteria produces a list of 3 companies. That’s not an ICP, that’s a named account list. Keep Column 1 to 4-5 criteria max. Use Column 2 for everything else.
Your ICP should change every quarter. As you close more deals (or lose them), new patterns emerge. The company you thought was a perfect fit 6 months ago might have churned. The vertical you ignored might be your fastest-growing segment.
Here’s the exact template we recommend. Fill this out, run a search, and refine. Total time: 15 minutes for V1.
Must-Haves (all required): - Industry: ___ - Company size: ___ to ___ employees - Geography: ___ - Buyer role: ___ - One qualifying signal: ___
Amplifiers (increases priority): - ___ - ___ - ___
Exclusions (always skip): - ___ - ___ - ___
Validation check: Run this in Sales Navigator or your prospecting tool. You should get 200-2,000 results. If more, add criteria. If fewer, remove criteria.
An ICP without execution is just a document. Here’s how it connects to the rest of your sales motion:
1. ICP → Lead list. Use your ICP criteria to build a targeted prospect list. Tools like Apollo, ZoomInfo, or AI prospecting platforms can automate this.
2. Lead list → Qualification. Not every company on your list is ready to buy. Score and qualify based on timing signals and intent data.
3. Qualification → Outreach. Personalize your messaging using ICP-specific pain points. A VP of Sales at a 200-person SaaS company has different problems than a Head of Growth at a Series A startup.
4. Outreach → Feedback loop. Track which ICP segments convert. Double down on what works. Refine the criteria that aren’t predicting success.
The founders who build pipeline fastest aren’t the ones with the most sophisticated ICP. They’re the ones who go from V1 to V3 the fastest. In 36 interviews with salespeople using AI, the top performers weren’t the ones with better ICPs on paper — they were the ones who iterated faster on targeting and messaging.
How is an ICP different from a buyer persona?
An ICP defines the company. A buyer persona defines the person at that company. You need both, but start with the ICP. There’s no point crafting a detailed persona for a VP of Sales at companies that will never buy your product.
How many ICPs should I have?
Start with one. If you’re selling to genuinely different markets (e.g., SMB SaaS and enterprise manufacturing), you might need 2-3. But most early-stage companies should focus on one ICP until they’ve validated it with 20+ closed deals.
What if I don’t have enough data to define an ICP?
Use your best guess based on the problem you solve. Who has this problem the worst? Start there. Your first 10 conversations will tell you if you’re right. If you want to validate assumptions before committing, synthetic market research can simulate buyer responses to test whether your ICP resonates — Colgate used this approach to replace $50K consumer studies.
How often should I update my ICP?
Review quarterly. Update when you see patterns in wins or losses that your current ICP doesn’t explain. If you’re closing deals outside your ICP, it might need expanding. If you’re losing deals inside your ICP, it might need narrowing.
Can AI build my ICP for me?
AI can accelerate the process — analyzing your existing customers, identifying patterns, and suggesting criteria. But the strategic decisions (which market to focus on, which segments to prioritize) still need human judgment. Tools like Onsa.ai can draft an ICP from a company URL in minutes, but the refinement loop still requires your input.
This article is based on patterns from 47 real ICP-building conversations. All examples are anonymized. Want to build your ICP in 15 minutes? Try Onsa.ai — our AI agents create ICPs from just a URL.