Seven ICP mistakes that quietly kill outbound

The targeting errors that make good lists behave like bad ones: vague industries, missing exclusions, wrong seniority, unbuildable signals and profiles written for humans rather than systems.

8 min read · Published September 5, 2026 · Updated September 16, 2026

The short answer

The short version.

Most bad lists are bad briefs. The data engine did exactly what the profile asked for.
Exclusions do more work than inclusions, and almost nobody writes enough of them.
Job titles are unreliable across company sizes, so target function and seniority together.
If a criterion cannot be found in a database or on a website, it cannot be used to build a list.

When a list underperforms, the vendor gets blamed first and the brief gets read last. In practice, most disappointing lists are faithful executions of a profile that was never specific enough to execute.

These are the seven failures we see most often, and the fix for each.

One: an industry so broad it means nothing

Professional services covers a law firm, a staffing agency and a bookkeeper. Technology covers a two person app studio and a payments processor. A broad industry label guarantees a list full of businesses that share a category and nothing else.

Fix it by naming what the business does in plain language, then listing three companies you already sell to well. Those examples do more targeting work than any taxonomy code.

Two: no exclusions

A profile without exclusions will always fill with competitors, vendors, agencies and adjacent industries.

Inclusions describe the middle of your market. Exclusions protect the edges, and the edges are where lists go wrong. Without them you will receive your own competitors, companies that sell to you, agencies that serve your buyers, and the recruiters and software vendors that orbit every vertical.

Write at least six exclusions. Include competitor names, the categories that share your keywords, company types that cannot buy, and any department that will never own the problem.

Named competitors and their subsidiaries.
Vendors and resellers who serve your buyers.
Adjacent categories that share vocabulary with yours.
Company types that structurally cannot buy, such as franchises or public bodies.

Three: targeting titles instead of functions

Titles are wildly inconsistent across company sizes. The person who buys your product might be Head of Operations at forty staff, Director of Field Services at four hundred and VP Operations at four thousand. Pinning a list to one exact title cuts most of your real market out.

Target a function plus a seniority band instead, then let the matching engine map local title variants onto it. Also name the titles you never want, because junior coordinators are the fastest way to fill a list with people who cannot act.

Four: a company size range that hides two businesses

One to five hundred employees is not a segment. Under fifty staff the owner decides, spends fast and answers the phone. Over two hundred there is a committee, a procurement step and a longer cycle. Those need different lists, different openers and different offers.

Split them. Run each as its own profile with its own seniority, and compare connect and meeting rates after three hundred dials each.

Five: signals that cannot be found

Profiles often include criteria no system can verify, such as unhappy with their current provider or planning to expand next year. Intent like that is inferred from behaviour, not stored as a field, so asking for it either returns nothing or silently gets ignored.

Convert soft criteria into observable proxies: technology in use, hiring activity for a relevant role, location count, fleet size, certifications, service keywords on the website. Those are findable and repeatable.

Six: geography written loosely

North America is not a geography if you only sell in three provinces or six states. Loose geography inflates volume and wrecks connect rates, because reps end up dialling across time zones they never staff.

Name the states, provinces or metro areas you actually cover, and name the ones to avoid. If you sell nationally but staff one time zone, say so, because pacing delivery around calling hours matters more than list size.

Seven: never revising the profile after the first month

A profile is a hypothesis. After the first few hundred dials you know things you could not know at signup: which segments answer, which titles engage, which exclusions were missing. Teams that update the brief monthly compound their results. Teams that set it once blame the data all year.

Keep a simple log of the records your reps flagged and why. Patterns appear quickly, and each pattern becomes an exclusion or a refinement. Our guide to defining an ICP walks through the full brief structure.

Key takeaways

Name what the business does, not the industry category it sits in.
Write at least six exclusions before you write another inclusion.
Target function plus seniority, never a single exact title.
Replace unbuildable intent criteria with observable proxies.

See the data behind the advice

Five ICP matched prospects with mobile numbers, direct dials and work emails, free. Same pipeline that fills a paid account, no credit card.

Get 5 free sample leads See pricing

Frequently asked questions

What makes a good ideal customer profile for buying leads?

One a system can execute without guessing: a plain language description of the business, named example customers, a function and seniority band, a tight geography, a narrow employee range and at least six exclusions.

Should I target job titles or job functions?

Functions with a seniority band. Titles vary too much across company sizes, so exact title targeting removes large parts of your real market while letting junior lookalikes through.

How often should I update my ICP?

Monthly for the first quarter, then quarterly. Use flagged records and connect rate by segment as the evidence, not opinion.

Keep reading

Buying