Ideal customer profile for outbound: how to define it and test it with campaigns

An ideal customer profile (ICP) is a short, written description of the companies most likely to buy from you, stay with you and be worth serving. For outbound, define it in a few concrete criteria based on your best current customers, then test it: give each segment its own calling campaign, use the same script for all of them, and compare meetings per call attempt and the reasons people say "not interested". The segment that books more meetings with fewer calls wins. A profile you never test is only an opinion.

What an ideal customer profile is, and what it is not

An ICP describes companies. It answers: which industry, which size, in which situation, and who signs. A buyer persona describes the people inside those companies: their role, their worries, the words they use. You need both, but they are different tools.

An ICP is also not a list of everyone who could possibly buy. Almost any company "could" use a bookkeeping tool or a cleaning service. The ICP is about who buys fastest, pays well and stays. That is a much smaller group.

Why the wrong customers cost more than they bring

Selling to anyone who says yes feels like progress. It often is not. In a 2026 Harvard Business Review article based on interviews with startups, Eric Janssen and Brian Denenberg call this "sales debt". Poor-fit customers tend to need discounts, extra support and custom work, and they churn faster. The sales team then has to replace them, often with more poor-fit customers.

The article also offers a simple test you can use in your own team: ask people across sales, service and finance whether they would be happy with ten more customers just like a given one. If the answer is no, that customer is not in your ICP, even if the deal was large.

For a calling team there is a second cost. Every call to a company that will never buy is a call not made to one that might. Our article on how many calls it takes to win a customer shows how strongly a better meeting rate lowers the number of calls you need.

The five layers of a B2B profile

A classic 1984 HBR article by Benson Shapiro and Thomas Bonoma, How to Segment Industrial Markets, still gives the clearest structure. They describe five "nested" criteria, from easy to see to hard to see. Here they are, translated for an outbound team:

1. Demographics. Industry, company size, location. Easy to find in any lead list.
2. Operating variables. How the company works: the technology it uses, the services it delivers, whether it has a team for the problem you solve.
3. Purchasing approach. Who buys, how centralised buying is, whether they prefer existing suppliers.
4. Situational factors. Urgency, order size, specific applications. Is something happening right now?
5. Personal characteristics. The person: their attitude to risk, their loyalty to current suppliers.

The authors also noted that many industrial companies used segmentation to explain results afterwards rather than to plan. That is exactly the trap with an ICP: it becomes a slide, not a rule for which leads get called. They also warned that the model should not be applied like a recipe. Use the layers that matter for your offer.

Add urgency: who needs it now?

Layers one and two give you a list. They do not tell you who to call first. In a 2023 HBR article, Neil Baron and Rod Griffith argue that classic profiles miss urgency of need. They sort prospects into four groups: urgent, non-urgent, currently met (they have a solution that works for now) and no need. Their example: job ads for a role your product supports can signal that a company is struggling right now.

For cold calling, urgency signals are gold. Think of a new location, a recent merger, a vacancy in the team you help, or a new law that hits their sector. Add one or two of them to your ICP as a priority layer.

Start from your best customers, not from a wish list

Your current customers already hold most of the answers. A 2024 HBR article by Frank Cespedes, Scott Peterson and Daniel Weinfurter describes how one IT services firm did this. It looked at which clients were most and least profitable, which were likely to renew and which red flags pointed to clients it could not serve well. From that it built its ideal customer profile and its qualification rules. Note the limits: this is one company, and one of the authors had consulted for it.

You can do a simple version in an afternoon:

1. List your last 20 to 30 customers.
2. Mark each one: profitable or not, stayed or left, easy or hard to serve.
3. Look at the best ten. What do they share? Industry, size, the trigger that made them buy, the role of the person who signed.
4. Look at the worst five. What do they share? Those traits go on your "do not pursue" list.

Write down who not to call

The "no" list is as important as the "yes" list. In a 2025 HBR article, Scott Edinger urges leaders to tell sales teams which business they should not pursue, and to describe the ideal client in detail: industry, size and circumstances. He also stresses the level of authority needed to buy. A perfect company with the wrong contact still ends in a no.

Write your ICP as hypotheses you can test

Do not write one perfect profile. Write two or three competing guesses, each specific enough to build a lead list from. An example for a fictional company that sells a planning tool to service businesses:

For every segment, also write who to ask for (for example the operations manager, not the receptionist) and one urgency signal to sort the list.

How to test your ICP with calling campaigns

Now let the phone decide. The setup matters more than the size of the test.

1. One campaign per segment. Never mix segments in one list, or the average hides both the winner and the loser.
2. Same script, same callers. Change only the segment. If caller A calls segment A and caller B calls segment B, you are testing callers, not segments.
3. One outcome per call. Not reached, conversation, call back, meeting, not interested, invalid number.
4. A reason for every "not interested". Pick from a short fixed list, so you can count them.
5. Enough calls. Wait until each campaign has a few hundred attempts and dozens of real conversations. A difference of one or two meetings means nothing.

Example: three campaigns compared

The numbers below are made up to show how to read a comparison. They are not a benchmark.

How to read this:

• A books three times as many meetings per call as B. It is the strongest ICP candidate. Its main objection ("already has a tool") is a competitor issue you can work on in the script.
• B is a poor fit. "Too small, no need" is a profile problem, not a script problem. Move it to the "do not pursue" list.
• C picks up the phone well but talks to the wrong person. Do not drop the segment yet. Change the role you ask for and run it again.

That last point is why the reasons matter as much as the rates. The rates tell you which segment works. The reasons tell you why the others do not, and whether that is fixable. Our guide to win/loss analysis goes deeper into learning from every no.

Look past the first meeting

A meeting is a leading signal, not the finish line. After a few months, check which segment also turned meetings into customers, and which of those customers stayed. A segment that books many meetings but rarely buys is a warning sign, just like the sales debt described above.

Tell the team, and update it every quarter

An ICP only works if callers know it. Put the criteria and the "do not pursue" list where callers see them: at the top of the script or in the call notes. Then review it every quarter, or sooner when a campaign result clearly says something else.

Ringlane helps with the testing part. You import each segment as its own campaign, callers log one outcome per call with a single key, and every "not interested" gets a reason. The reports then show connect rate, conversation-to-meeting rate, meetings as a share of attempts and the reasons for not interested per campaign, side by side. For what each of those numbers means, see our guide to cold calling KPIs. If you buy lead lists to build your segments, read our article on B2B lead lists and the GDPR first.

Common ICP mistakes

• Too broad. "SMEs in Belgium" is a market, not a profile.
• Only firmographics. Size and industry give you a list. Urgency and the right contact give you meetings.
• Built from hope. The segment you would like to win is not always the one that buys.
• Mixed campaigns. If segments share one list, you can never tell which one worked.
• Never revisited. Markets, offers and teams change. The profile has to change with them.

Frequently asked questions

Sources

1. Harvard Business Review: How to Segment Industrial Markets (Shapiro and Bonoma, 1984)
2. Harvard Business Review: Let the Urgency of Your Customers' Needs Guide Your Sales Strategy (Baron and Griffith, 2023)
3. Harvard Business Review: The Risks of Prioritizing Short-Term Revenue over Customer Fit (Janssen and Denenberg, 2026)
4. Harvard Business Review: 4 Steps That Can Optimize Your Sales Process (Cespedes, Peterson and Weinfurter, 2024)
5. Harvard Business Review: Winning the Right Customers Isn't Just a Sales Issue (Edinger, 2025)