MCP

Ideal customer profile

TL;DR

  • An ideal customer profile describes companies, not people. It is a filter you can apply to a database, which means every criterion has to be observable from outside.
  • Build it from customers you already have and already like, not from the market you wish you had.
  • Narrow beats broad by a wide margin in the published data — smaller companies and tighter segments reply several times more often than large ones.

What an ICP has to be

The test of an ICP is whether two people applying it to the same list produce the same list. That rules out most of what appears in a typical profile — "innovative", "growth-minded", "understands the value of design" — and leaves criteria you can actually filter on: industry, headcount range, geography, funding stage, technology in use, business model, and observable behavior like hiring for a role or opening a location.

The reason to be strict about this is not tidiness. Every unobservable criterion becomes a judgment call made row by row, which either does not happen at scale or happens inconsistently, and either way your list stops matching your thesis.

Build it backwards from real customers

The reliable method is to start from your best existing accounts and look for what they share. Best means high value, fast to close, low to serve, and still around — not merely large. Pull twenty of them and look for shared observable attributes, then check the pattern against accounts that churned or never closed. What separates the two groups is your profile; what both groups have in common is not a criterion, it is just a description of your market.

For a first campaign with no customer history, the profile is a hypothesis rather than a finding. Say so, write it into the campaign thesis (Campaign planning workflow), and treat the campaign as a test of it.

Narrow really does win

The data is unusually consistent here. A 7.5M-email corpus (2026) found companies of 0-10 employees replying at 0.72% against 0.22% for enterprises above 10,000 — a threefold spread with nothing to do with copy [1]. A separate 20M+ email corpus (2026) reports lists under 50 contacts replying at 5.8% against 2.1% for lists over 1,000 [2].

Read the second finding carefully, because the causation runs backwards from how it is usually quoted: small lists do not cause high reply rates. Small lists are what you get when the criteria are tight, and tight criteria are what causes the reply rate. Truncating a broad list to 50 names buys nothing.

Segment while defining

A profile that survives contact with a database usually turns out to contain two or three meaningfully different groups — different problems, different buying triggers, different language. Split them now rather than writing one email that addresses all of them weakly: Segmentation and tiering.

Write down the exclusions too

An explicit not-list is as valuable as the profile: companies too small to pay, too large to reach, in regimes where the outreach is not lawful (Staying legal in a nutshell), competitors, existing customers, and anyone already suppressed. Every one of these that reaches a send is a reply you have to apologise for — see Disqualification and fit.

References

  1. Belkins — Cold email response rates (7.5M emails sent in 2025, updated Jun 2026)
  2. Woodpecker — Cold email statistics (20M+ email corpus, updated Jun 2026)