Buyer Persona, Customer Avatar and ICP: Three Different Things

The ideal customer profile, known as the ICP, describes the type of company worth pursuing. The buyer persona describes the person inside that company and how they decide.

The customer avatar is a variant of the persona, coming out of direct response marketing, built more often on assumptions than on interviews. All three get used interchangeably in nearly every discussion. They answer questions that do not overlap.

TL;DR. The ICP answers “who is worth talking to”, the persona answers “how does this person decide”, and the avatar is the name the persona carries in the copywriting school.

The correct order is ICP first, persona second, because a profile of a person built before you know which company you want describes somebody who could work anywhere. The most expensive mistake sits elsewhere, though. A document full of demographic detail that explains no decision costs far more than the confusion of terms.

What is the difference, briefly?

The most useful distinction remains the one HubSpot put in a single sentence. “Personas tell you who you’re talking to. ICPs tell you which companies are worth talking to in the first place.”

Where does each concept come from?

The persona was not born in marketing. Alan Cooper introduced it as a software design instrument, and he writes explicitly that “The Inmates Are Running the Asylum, published in 1998, introduced the use of personas as a practical interaction design tool”.

In the same text, Cooper separates his concept from what marketers were already doing, noting that product marketing professionals had used similar entities for years to define demographic segments. The two traditions have been separate since the origin. That separation explains a good share of today’s confusion.

The marketing variant was systematised by Adele Revella, through the Buyer Persona Institute. She draws a distinction worth keeping, between the buyer persona and what she calls the “buyer profile”.

The profile describes a person’s characteristics and reveals little about how you influence them, while the persona reveals the buyer’s attitudes, concerns, decision criteria and journey. Her framework was published in the book Buyer Personas, in 2015. It is built on five information areas, meaning priority initiatives, success factors, perceived barriers, decision criteria and the buying journey.

The ideal customer profile comes out of the software-as-a-service world.

Lincoln Murphy defines it as “the type of customer that, within a clearly defined timeframe, you will dedicate sales and marketing resources to acquire”, and he is also the one who states the ordering rule, saying that persona development has to come after ICP development.

The customer avatar is the term used in direct response marketing, where it is treated openly as a synonym for the persona.

What separates it from Revella’s school is the method. The classic instruction in copywriting manuals says to not wait for surveys or interviews and to make assumptions where you have no data.

The approach makes sense when you sell a low-price product to a broad audience and can test quickly. It becomes dangerous in a B2B sale where one cycle runs for ten months.

Why do we start from WHY rather than from demographics?

A document that describes somebody answers the wrong question, because the useful question sounds entirely different. Why would this person move from where they are standing right now?

The doctrine we apply states it in one line. Communication starts from WHY, meaning from purpose, not from the description of the service, and the WHY reaches passive memory and the non-rational brain, which is where action gets triggered.

The structure is called the Golden Circle and it orders three rings, meaning WHY, HOW and WHAT. We wrote at length about how it works in the article on Simon Sinek’s Golden Circle.

Applied to the client profile, the rule changes what you look for in an interview. You ask what the person is trying to achieve, what stands in the way, and what they have already tried without success. Job title, age and city stay list data, useful for targeting and useless for the message.

The difference shows in the material that comes out. A profile built on WHAT produces a list of services anybody in the category could publish, while a profile built on WHY produces a message a competitor cannot sign.

The test is mechanical and takes a minute. You take the document, delete your company name, and hand it to somebody in your market. If they cannot guess who it is about, you have described the category.

How do we build the client profile, step by step?

The method runs in two directions at once, and they only get overlaid at the end.

The first direction starts from the top. You write down who you believe your clients are, then you ask a thorough research pass to validate or dismantle that hypothesis with market data.

The second direction starts from the bottom, inside the company. You take the DNA of the people who deliver and the methodology you actually practise. Then you ask who benefits from it most, without looking at the existing portfolio at all.

The ICP nucleus sits in the intersection of the two. What appears in both lists stays as a decision. What appears in only one goes into the test zone, with a small budget and a short deadline.

The principle we apply here reads as follows. You look for patterns of clients with a recurring need, and for the place where those decision makers are already gathered, so that you compress the time to the first clients.

The second half matters more than it sounds on a first reading.

A correct ICP, defined without knowing where those people gather, gives you a list and no channel to reach it through.

Our own band came out of exactly this overlay. Companies between three and thirty million euros in revenue, with the most productive zone between five and ten million.

The edges are not accidental. They come from a second principle we call the ideal client in the middle, and the logic behind it travels across markets. The client who is too large treats you as an execution supplier, while the one who is too small has neither the budget nor the structure to carry the decision.

The middle has all three. It has money, it has internal structure, and it has somebody accountable for the result. That is why it pays on time and decides within a predictable window.

Below roughly two and a half million, the decision gets taken on other criteria and the cycle becomes unpredictable. Our declared anti-fit is the company that asks only for ads and refuses the positioning conversation. The methodology has nothing to rest on there.

What breaks when you confuse them?

The budget breaks, and the mechanism is always the same.

A company builds a document describing “Andrew, 42, procurement director, enjoys fishing”, then uses it to decide where the money goes. That document contains no information capable of changing a decision. Age and hobby do not explain why somebody signs, and a shared hobby has never rescued a weak proposal.

Clayton Christensen said this in a form I have not managed to improve on since. “The fact that you’re 18 to 35 years old with a college degree does not cause you to buy a product.”

Description stays description, and a profile built out of descriptions produces campaigns that speak nicely and convince nobody.

The second consequence appears at list level. A company that skips the ICP and starts directly with the persona ends up addressing the same job title at companies of completely different size and situation. The same message goes to a two-million company and to a hundred-million one.

The result shows in the reply rate. The message looks like the cause, although the cause actually sits in the list.

In what order do they get built?

The ICP gets built first, from your data rather than your wishes. You look at the clients who stayed longest, brought the highest margin and required the least delivery effort. Then you find what they have in common in size, industry and situation.

The persona gets built second, from interviews with people who went through a real decision, including those who chose somebody else.

Refusals explain more than acceptances, because they show you the criterion you were failing.

The third step, which almost everybody skips, is writing down what your client is not. An ICP band with no floor and no ceiling stays a wish list. Real segmentation means giving something up, and if your document excludes nobody, it will save you nothing.

How many messages do you need when thirteen people decide?

A single message for everybody looks like economy and behaves like surrender, and in B2B it gets paid for directly in the conversion rate.

The rule we apply is called a different message per decision-maker, and it starts from a simple observation. In B2B several people buy, each carrying a different fear, and every fear has its own source and its own vocabulary.

We applied it at Bigal, a Romanian manufacturer of aluminium façades and a good stand-in for mid-market industrial companies anywhere in Europe.

The company ranks fifth in its market by margin and fifteenth by revenue, a contrast read from the public financial statements filed with the Romanian tax authority for the 2024 financial year, in our own analysis of the 19 active companies in its category (NACE 2512).

The second contrast sits on their own website, in plain view.

The company has executed more than 1,300 projects, of which 22 are published in the portfolio on its own site, all of them anonymised, with no building name and no developer name.

This is the situation we meet most often. The proof already exists and it has been sitting inside the company for years, uncommunicated. The differentiator everybody hunts for is simply waiting to be brought into the light.

The move we proposed was to break the single message into three, along the three decision makers who appear in every façade project.

The architect buys the preservation of design intent, meaning the guarantee that the façade they drew reaches the building as they conceived it. The builder buys predictability on site, because every week of delay comes out of their pocket. The developer buys the conversion of cost into sale value.

The anti-pattern is segmenting by company size or by industry, instead of segmenting by role in the decision.

It gives a sense of rigour and leaves the message identical for everybody.

A second reference point comes from ADF, a joinery manufacturer. The company records roughly 390 brand searches per month, against 4,400 for QFort, its largest domestic competitor, figures we measured against Google search volumes in June 2026.

That distance does not close with a better message aimed at everybody. It closes with separate messages to the installer, to the architect and to the owner paying the bill, because the three are looking for different things. We covered the mechanism in the article on branding in B2B.

What do you do when thirteen people decide?

This is where the classic persona model runs into difficulty, and the data shows it clearly. Forrester measured, in its January 2026 report, that a typical buying decision involves thirteen internal actors and nine external influencers. Procurement becomes a decision maker in 53% of cycles.

A profile written for a single person cannot describe a process like that.

Gartner’s proposed answer is to move attention from people to jobs. Their framework, published in 2025, describes six “buying jobs” the group has to complete, meaning problem identification, solution exploration, requirements building, supplier selection, validation and consensus creation.

The practical idea stays simple, because thirteen separate profiles help nobody. What helps is clarity on what the group has to resolve together, plus the material that serves each job in turn.

The most neglected job of all is consensus creation.

The person you convinced has to convince somebody else inside the company, often somebody who has never spoken to you. If you give them nothing to do it with, the process stops there. We wrote about this breaking point in the article on the real buying journey as well.

Which layer of the company does the client profile sit on?

The methodology we run has four layers, plus the AI layer that binds them. The client profile does not sit on all of them equally.

L1 is performance marketing, where cost per lead and return on budget get measured. L2 is the revenue layer, meaning the commercial process and how an opportunity turns into a contract. L3 is AI orchestration, meaning the infrastructure that holds the company’s context. L4 is positioning and category architecture.

The ICP gets decided at L4, because it states which category you play in and who you refuse.

The persona serves L2, because it describes how the commercial conversation is carried.

A wrong ICP, however, gets paid for at L1, in the advertising budget, two or three years after the moment it was written. That is where the bad number surfaces. The decision had been taken long before, in a meeting where nobody mentioned cost.

L3 is where the document becomes useful to a machine.

An ICP written in free text stays a nice story. The same ICP written with clear edges becomes an automatic filter in the prospecting list and in lead qualification.

The order matters more than the content of any single layer. A company that starts at L1, with the budget on channels, pays for years for a positioning decision it never took.

What does it look like written down?

A useful ICP fits in a few lines and contains edges rather than adjectives. Here is the form we use, applied to a company selling industrial equipment.

The company we pursue is a manufacturer with revenue between five and thirty million euros, with its own production and with exports making up at least a quarter of income, at a moment when its margin has fallen for at least two consecutive financial years.

We do not pursue distributors without their own production, companies below three million, or companies that already have an internal marketing department of more than five people, because the decision there gets taken differently and the cycle becomes too long.

The second part, meaning the list of those you refuse, is the part that produces the saving. It speaks on the day a company that looks good and does not fit picks up the phone. It tells you the correct answer is no.

The corresponding persona looks equally simple. The person who signs is the founder or managing director, who built the company on relationships and on the product. He has no time for theory and measures everything in what remains afterwards.

Their priority right now is to stop the margin decline without cutting quality. Their main barrier is that they have already worked with somebody who promised results and delivered reports, so they arrive with justified distrust. Their decision criterion is whether they can see a mechanism or a list of services.

Notice what is missing from the description above. There is no age, no city and no mention of weekend hobbies, because none of those elements change how you write to them or what you show them.

How do you test whether your document is good?

Through three questions, all with verifiable answers.

First. Can the document say no? If it does not produce at least one category of client you refuse, then it describes a wish.

Second. Could a competitor put the same document on their intranet without a single change? If yes, then you have described the category. It is the same competitor test we apply to brand messaging, and it works just as well here.

Third. Has the document changed a decision in the last month? An ICP that is written and unused has the same value as one never written. The practical test is to look at the last five opportunities that came in and check whether anybody ran them through the filter before investing time.

Documents that fail these tests are not wrong because of their format. They were built out of assumptions in a meeting instead of sales data and interviews. The difference shows precisely when somebody asks them to produce an uncomfortable decision.

What does it look like at a 3M+ EUR company?

Simpler than it appears, provided you do not copy templates. You need an ICP written in two lines, stating the size, industry and situation of the company you pursue, plus the list of those you refuse.

You need one persona for the person who signs and a short description of the person who can block, because at this company size the blockage most often comes from finance or from an operations director who was never in the conversation.

The sources are already in your house.

The last ten proposals you won and the last ten you lost contain nearly everything, and if you read them looking for real reasons, what comes out is more useful than any purchased research.

This document is not a deliverable to tick off. It becomes useful only at the moment you use it to say no to an opportunity that looks good and does not fit. That remains the only test that matters.

The connection between this clarity and how the market perceives you is described at length in the article on cognitive ownership.

Frequently asked questions

How many personas do I need?

One for the person who signs and one for the person who can block. Above two, the documents stop being read by anybody, and an unread document influences no decision.

Can I use demographic data at all?

Yes, as a list filter. Company size and industry help you choose who to contact, while the reason somebody buys comes from criteria and barriers.

Is that statistic about 71% of companies exceeding targets with documented personas real?

The figure circulates as though it were from last year, but it comes from a Cintell study in 2016. If you use it, state the year, otherwise you repeat an error that everybody repeats.

How long does it take to build the ICP?

A few days, if the sales data is at hand. The analysis is short. The hard part is the decision to exclude segments where you already have clients and find it hard to set aside.

Are ICP and target market the same thing?

No. The target market describes the whole space where you could sell, while the ICP describes the part of that space you choose to pursue now, with the resources you have.

How do I know whether I need separate messages per decision maker?

Look at who signs and at who can stop the project. If the two have different objectives, as an architect and a developer do, a single message will convince at most one of them.

If you want to see what your real ideal client band looks like, starting from what you already have in sales, you can request a positioning analysis. It connects the client profile to how the market perceives you. The two break together and get repaired together.