How to Define an ICP for B2B Outbound: Start With Who Actually Bought

Aakash Sinha

KEY SUMMARY

There are two ICPs. The account ICP, industry, size and stage, is what most teams write, and the standard advice to build it from closed-won deals is right. The title ICP, the job titles that actually bought, is what outbound sends to, and almost nobody builds it. Read it off your own data: every contact on a won deal, the people your customer success team emails, and the contacts who sign renewals. It's usually one level below the title on the slide. Build your outbound universe as those titles multiplied by your target accounts, decide who not to email as carefully as who to, and expect about half of any raw export to fail the cut.

On one LinkedIn outbound campaign I ran this year, a raw export of 1,990 founders came down to 1,036 before a single message went out. Not because the copy changed. Because 954 of the people on it should never have been on it.

Cleaning an outbound list: a raw export of 1,990 founders narrowed to business founders, then an experience gate, leaving 1,036 ready to send

Half of a raw export usually shouldn't be messaged.

That's about normal. And it's why, when a campaign comes back with poor reply rates and everyone starts rewriting the copy, I ask to see the list first. In my experience most of what fixes outbound is who's on the list. The copy comes a distant second.

The list goes wrong at the definition, not the export. Most ideal customer profiles are written in a workshop, and most of them describe companies. Outbound doesn't email companies. It emails people, and the people are where the slide and the CRM disagree.


Two ICPs, and outbound sends to the second one

Two ICPs: the account ICP of industry, size and stage that decides which companies, and the title ICP of the job titles that bought, which decides which people

Most outbound fails on the second one.

The advice to build an ICP from your closed-won deals is now mainstream, and it's good advice. Rox lays out the account version well: take 24 months of closed-won, keep the top 20 to 30% by close speed, deal size and retention, document what they share, and add exclusions from churned and lost deals, which they say can remove 20 to 40% of a typical list. That gives you a good account ICP. Which companies.

It doesn't tell you who to email inside them, and that's the gap a cold campaign falls into. Forrester puts the average buying decision at 13 people across two or more departments in 89% of purchases. The slide names one of them, usually the most senior, usually the signer. Thirteen people were in the deal. You've written down one, and it's the one least likely to answer a stranger.

In my own LinkedIn campaigns the obvious top title, the CMO, rarely replied. They don't feel the day-to-day pain. The senior-most person one level below them does, and that's who answered. Going one level down changed who replied more than any rewrite did.


Read the titles off the deals you've won

Where your real ICP is hiding: closed-won deal contacts, customer success email threads and renewal contacts point to the job titles that actually buy

Three places in your own systems that show who really buys.

The title ICP is already in your systems. Pull three exports:

  • Every contact on a closed-won deal, not only the primary one. The person who found you and the person who ran the evaluation are usually not the person who signed.

  • The people your customer success team actually emails. They're the ones using what you sold, and the ones who'll run the next evaluation somewhere else.

  • The contacts who sign renewals. A renewal is a second purchase made with full knowledge of the first. If Rox filter accounts by retention, this is the title-level equivalent: the people who bought twice.

Group the titles into families, because the same job goes by many names, and count. You'll find a few families that keep appearing and at least one you'd never have put on the slide. Watch the seniority as well as the function. If most of the buyers sit one level below the title you were aiming at, that isn't a data problem. That's the answer.

If you don't have many customers yet, use the nearest thing: the titles that took demos and stayed engaged, and the titles that replied to your first outreach. Treat it as a draft and replace it as deals close.


TAM is a list, not a slide

Outbound TAM equals ICP job titles multiplied by target accounts, giving a list of real people you can reach

A list of real people, not a market-size slide.

Once you have the titles, the outbound universe is one line:

TAM = ICP job titles × target account list

Not the market-size number in a pitch deck. The real people you can reach, which is the only TAM outbound cares about. Pick the accounts first, and there are three honest ways to do it: mirror the customers you have, go up market to bigger versions of them, or go into a segment you haven't sold to yet. Then apply the title filter to those accounts in an enrichment tool like Clay and export the matches. One export like that came back with 4,200 rows: job title, company, city, state, country, follower count, LinkedIn URL.

Keep every column. City and country tell you later where to host a dinner, because your biggest cohort of buyers is your city. Follower count builds the second list.

The second list is the people your buyers listen to: the same titles, the account filter removed, a broad company filter instead, and a follower threshold around 10,000. Fifty to seventy people. You won't sell to most of them. They're who you connect with first on LinkedIn so your posts show up next to theirs, who you invite to speak, and whose posts tell you what your buyers read before you write a word.


Who not to email

A list built from the right titles can still be full of the wrong people, and the rules for removing them matter as much as the rules for finding them. Mine, applied to every list before anything goes out:

  • The email domain is the truth signal, not the company field. People change jobs, and the CRM company field is the last thing anyone updates. The domain they use now says where they work now.

  • Free and personal domains go into a permanent exclusion set.

  • Check the LinkedIn profile against the name. Enrichment tools attach the wrong person more often than you'd think, and a wrong-person email is the fastest way to look careless.

  • When in doubt, remove the record. Your sending domain's reputation is the asset. Bad records spend it.

  • Deduplicate on the lowercased email, and keep the higher-priority copy: customers over pipeline over event contacts over cold.

On LinkedIn, add three more. Filter on where the person is, not where the company is headquartered: an India-headquartered company can have its marketing leader in New York. Add an experience gate when seniority matters. On the founder campaign above, the first wave got feedback that the founders looked too young, so we kept only founders whose current role was at least their third job, which removed 104 people from a list of 1,147. And don't trust keyword scores for company quality. Scoring companies on description keywords ranked unknowns above category leaders. Check the top of the list by hand.

One sequence per company at a time, and never more than two contacts in one company in the same week. Nobody likes hearing from the same vendor twice.


The list tells you when the ICP is wrong

You don't have to guess whether the titles are right. The campaign tells you, if you watch three things.

Acceptance, on LinkedIn: if fewer than about one in five people accept the request, fix the list or the profile before sending more. Replies: if they keep pointing you to someone else in the company, that person's title belongs in your ICP, and the one you wrote probably doesn't. And meetings: if they keep landing with people who can't buy or can't influence a purchase, the titles are wrong however good the reply rate looks.

At the account level, track two numbers over a quarter. Coverage is the share of target accounts you've reached with at least one contact. Depth is how far you've got with the ones that matter, scored on every touch across every channel, with in person worth more than a reply and a reply worth more than an open. A good title ICP moves both. A wrong one gives you coverage with no depth: plenty of accounts touched, and nobody in them who matters.

Most teams write the buyer they want and then rewrite the email three times when that buyer doesn't answer. The buyer you have is already in your closed-won deals, one level below where you were looking.

The step-by-step Clay build, the database setups and the scoring model for every contact are in The 90-Day GTM Sprint to Make B2B Outbound Work.


FAQs

How do you define an ideal customer profile for B2B outbound?
Build two. An account ICP from your closed-won companies, and a title ICP from the people on those deals, the contacts your customer success team emails and the people who sign renewals. Outbound sends to the second.

Should outbound target the most senior decision-maker?
Usually not. The most senior title rarely answers a cold message. The senior-most person one level below them feels the problem day to day and replies more often.

How many people are involved in a B2B buying decision?
Forrester's State of Business Buying 2024 found an average of 13 people, with 89% of purchases involving two or more departments.

How do you calculate TAM for outbound?
Multiply your ICP job titles by your target account list in an enrichment tool. The result is a list of reachable people, not a market-size estimate.

How do you clean a B2B contact list before outbound?
Treat the email domain as the truth signal, exclude free email domains, check each LinkedIn profile against the name, remove anything doubtful, and deduplicate on the lowercased email. Expect about half of a raw export to go.

What if a startup doesn't have enough customers to define its ICP?
Use the titles that took demos and stayed engaged and the titles that replied to early outreach, and replace that draft with real buyer data as deals close.

AUTHOR

Aakash Sinha

Aakash Sinha

AI-Native Fractional GTM Operator, AlterStride

Builds and runs go-to-market motions for B2B companies rather than advising from the sidelines. First hire and zero-to-scale operator across HR, legal, and sales tech, with a track record of taking businesses from early customers to 200-plus in North America.

Copyright © 2026 AlterStride.

All Rights Reserved.

Field notes on B2B GTM in the AI age, by Aakash Sinha.

Field notes on B2B GTM in the AI age, by Aakash Sinha.

Copyright © 2026 AlterStride.

All Rights Reserved.