How to Build an Ideal Customer Profile That You and Your AI Can Actually Use

I bet there’s at least a handful of customers you wish you could clone.

You know the ones.

If you run a law firm, it’s the client whose case sat squarely in your wheelhouse, who signed the engagement letter without haggling over the retainer, and who sent their brother-in-law to you six months later.

If you’re in real estate, it’s the pre-approved buyer who knew what they wanted, took your advice, and closed in 45 days.

If you own a restaurant, it’s the table that comes back every other Friday and brings new friends each time.

If you sell online, it’s the customer who reorders every month and has never once opened a return.

If every new customer looked like that, you’d stop worrying about marketing.

Now look at where your marketing money actually goes.

Your website talks to everyone.

Your ads reach anyone who might buy.

Your lead form or booking page doesn’t ask a single question that would tell you whether the next inquiry looks anything like those customers.

So the budget lands on people who were never a fit, you or your team burn hours on consultations, showings, and carts that go nowhere, and the customers you’d clone find you mostly by luck.

Here’s the part I keep running into.

When I ask an owner to describe that best customer, the answer is almost always in their head and nowhere else.

Just about every business I talk to has skipped writing it down. And now they’re handing a vague audience to AI tools, which fill the gaps with a generic one, so the marketing gets more polished and less accurate at the same time.

Better copy won’t fix that. Neither will a bigger ad budget.

What fixes it is one written description of who fits, why they buy, and when to say no, and every person and tool in your business working from it.

That description is an ideal customer profile.

We made the ICP Asset 01 in our Brand Brain framework because every voice guide, offer, page, ad, and AI instruction depends on it.

Get it right and the rest of your marketing finally has something to aim at.

I wrote this guide for the owner or marketing lead who needs a usable ICP without a research team.

By the end, you’ll have a working profile built from your own customer records, your customers’ exact words, some outside audience research, and clear disqualifiers. You’ll also know where to put it in HubSpot, Canva, and your AI tools, how to test it, and how to keep it current.

A finished ICP should let anyone on your team answer six questions without asking you:

• Is this lead a strong fit?
• Which problem should we lead with?
• Which offer matches that problem?
• What proof will this buyer need?
• Where can we reach people like this?
• Which customers should we decline?

If your ICP can’t change one of those decisions, it isn’t done yet.

Open the FREE ICP template in our marketing-assets repository if you want to build alongside this guide.

---

What Is an Ideal Customer Profile?

An ideal customer profile defines the type of customer your business can serve well, deliver a strong result for, and earn a healthy return from.

Notice what that definition leaves out.

It isn’t the customer who spends the most, or the one who’s easiest to reach, or the one you happen to like.

It’s the customer where the work goes well for both sides. Those are often the same people, but not always, and I’ve watched that gap sink more than one profile.

One file, two views.

The ICP describes the company, household, or customer situation that fits your offer.

A short buyer summary inside it describes the person who feels the problem, compares options, and makes or influences the purchase. Keep both, and don’t confuse them.

If you run a personal injury firm, the customer view might be a person injured in a car accident in your county within the last 30 days, with medical treatment underway and a clear liable party. The person view might be someone in pain, worried about bills, unsure whether they even have a case, and looking for proof that you’ll return their calls.

If you’re a real estate agent, the customer view might be a pre-approved buyer looking in three specific neighborhoods within a set price range on a 90-day timeline. The person view might be a first-time buyer who’s nervous about overpaying and wants someone to explain each step before it happens.

If you sell to other businesses, the customer view might cover industry, revenue, employee count, location, software, and buying process. The person view might cover title, goals, pressure, objections, and where they go for advice.

The buyer summary supports the ICP. It doesn’t replace a full buyer persona.

How Is an ICP Different From a Buyer Persona?

I hear these two terms used interchangeably all the time, and it costs people.

An ICP tells you which customer to pursue.

A buyer persona helps you talk to the people involved in that decision. You need the first before the second is worth writing.

Document	Main Question	Useful Fields	Decision It Supports
Ideal customer profile	Which customer should we pursue?	Fit, need, value, triggers, disqualifiers	Targeting and qualification
Buyer persona	How does this person decide?	Role, goals, questions, objections, language	Messaging and content
Market segment	Which group shares a trait?	Industry, location, behavior, product use	Planning and reporting

Build the ICP first.

Add a persona when different people inside the same ideal company need different messages, like an owner who approves the budget and a manager who has to live with the result.

Keep that person-level work separate so it can’t blur your fit rules.

Once the fit is clear, our buyer persona guide walks you through the people inside that buying decision.

Does an ICP Work for a Consumer Business?

Yes. If you sell to households instead of companies, swap the company details for the buying conditions that affect fit.

A restaurant might define visit occasion, distance, party size, dietary need, average spend, and how likely the guest is to come back.

A family law firm might define case type, urgency, jurisdiction, ability to pay the retainer, and how the client found you.

A jewelry store might define purchase occasion, budget, how much research the buyer does, design preference, and service expectations.

An online store might use product need, order frequency, margin, return behavior, and customer lifetime value.

The fields change. The job doesn’t: point your business at the customers you can serve well.

What Makes an Ideal Customer Profile Useful?

A useful ICP contains evidence and decision rules. A weak one contains adjectives.

“Homeowners who want to sell” doesn’t tell your team who qualifies, what problem to lead with, or what evidence the buyer needs.

“Homeowners in three specific zip codes who have owned for more than seven years, are relocating for work, and need to sell before they buy” gives them something to act on.

The same goes for “anyone who needs a lawyer” versus “a person injured in a car accident in the last 30 days who is already receiving treatment.”

Our own ICP at imFORZA is a good example of why this matters.

We serve companies in several industries, so an industry label alone would tell us almost nothing.

Our profile defines the buying conditions instead: the person running the business signs the check, already pays for several marketing tools, and wants a steadier flow of customers.

That description forces choices I’d otherwise dodge.

I write for the person who approves the budget, not a vague marketing audience. I lead with customers, time, and money before I explain how anything works.

And when the fit isn’t there, I recommend a smaller starting point instead of pushing a full-service offer on someone who doesn’t need it yet.

Start With Fit, Need, and Value

Every strong profile answers three separate questions, and it’s worth keeping them separate:

1. Fit: Can you serve this customer with the offer, location, capacity, and experience you have?
2. Need: Does the customer have a problem urgent enough to act on?
3. Value: Is the relationship likely to work for both sides after price, service cost, repeat business, and support needs?

A customer can fit your service area and have no urgent need.

Another can have a serious need but require work you shouldn’t sell.

And a large invoice can hide poor margin, late payment, heavy support, and no chance of repeat business.

I’ve seen all three get mistaken for “ideal.”

Your best customer is the one who gets real value from your work while fitting the way your business already operates.

Include an Anti-ICP

Disqualifiers protect your time and, just as often, the customer.

Saying no early spares both of you a bad engagement.

Write down the conditions that make a poor fit, such as:

• a budget below the minimum required to deliver the work properly
• a location outside your service area
• a need your offer doesn’t solve
• a buying timeline your team can’t meet
• a business model that conflicts with your process
• a request for a guarantee you can’t honestly make
• a documented pattern of payment, communication, or scope problems

Don’t turn the anti-ICP into a list of people you dislike.

Write objective conditions someone else could apply without reading your mind.

Keep Customer Evidence Separate From Audience Signals

This is the habit that separates a real ICP from a confident guess.

Label each important statement as one of four things:

• Verified customer evidence: supported by customer records, interviews, sales calls, invoices, or support history
• Observed audience signal: aggregated behavior reported by SparkToro or another outside research source
• Working hypothesis: a possible meaning that still needs direct customer or campaign validation
• Unknown: a useful question that still needs research

These labels stop a polished research report or a strong opinion from quietly becoming company policy.

Say three customer calls mention missed follow-up as the reason leads were lost.

That’s direct customer evidence.

If an outside research tool shows that the same audience also visits other cideries and orchard destinations in the region, that’s an observed audience signal.

The idea that a regional food-and-drink publication will outperform a general lifestyle publication stays a working hypothesis until a campaign or a customer interview backs it up.

---

How Do You Create an Ideal Customer Profile?

Create an ideal customer profile by choosing one offer and one decision, reviewing your best and worst customers, finding the patterns behind them, writing the customer’s job and trigger, recording pain and decision criteria, capturing exact customer language, and drafting a first version with its unknowns marked.

You don’t need a perfect customer database for this.

You do need to know which lines came from records, which came from outside research, and which are still assumptions.

Here are the seven steps I use to turn that evidence into a first version you can test on a real decision.

To keep each step concrete, I’ll build one profile from beginning to end for a real client of ours, High Limb Cider.

High Limb makes craft hard cider in Plymouth, Massachusetts. It runs a taproom, sells online and through retailers in Massachusetts, Rhode Island, and Connecticut, and sells to bars, restaurants, and package stores through a separate trade side of the business.

That mix of local business and online store is why I picked them. Most of the owners I talk to are running some version of it.

I’ll use only what’s public about High Limb (the site, the taproom, where it ships, what customers say in reviews) plus the audience research we ran for this guide.

Where the profile needs their private records, I’ll say so and mark the field unknown rather than make something up. That’s the same discipline I’d ask of you.

Step 1: Choose One Offer and One Decision

Pick the offer the ICP will support.

By offer I mean one service, package, or product line that a customer buys for one reason, such as a maintenance plan, a website rebuild, or monthly marketing management.

If two offers need different fit conditions, triggers, proof, or disqualifiers, they need separate profiles. If they share the same customer and buying conditions, one profile can cover both.

High Limb is a clean example of why this rule exists.

On one side it sells cider to people: a visit to the taproom, a four-pack at a local store, an order shipped to the house.

On the other side it sells to businesses: a bar or restaurant putting High Limb on draft, a package store stocking the cooler, a distributor covering a state.

Those two customers have different triggers, different proof, different objections, and different people doing the buying. One profile covering both would be too vague to help anyone.

So the ICP in this walkthrough supports the consumer side: taproom visits, online orders, and retail purchases in the three states High Limb serves.

Its job is to help decide which visitors and subscribers deserve the follow-up, which message leads, and when to route someone to the trade side instead. The trade customer gets its own profile.

Finish this sentence before you gather any data:

We need this ICP to help us decide which [leads, customers, accounts, markets, or campaigns] deserve [sales time, marketing budget, service capacity, or product attention] for [specific offer].

That sentence is the job description for the document. Put it at the top of the file so nobody has to guess what the profile is for.

Step 2: Build Your Customer Evidence List

Start with customers you already know.

Pull records from your CRM, invoices, booking system, sales notes, support history, and call transcripts.

If your customer list is small, review every record. When the records allow, here’s the starting set I ask for:

• five customers you’d gladly clone
• five customers you wouldn’t pursue again
• five qualified opportunities you lost
• every churned or refunded customer with a clear reason
• every strong referral source with a known customer outcome

Then score each customer on the facts that matter to your business.

Fit Signal	What to Record
Customer outcome	Did the customer get the result the offer is built to deliver?
Revenue quality	Did the work produce healthy revenue after delivery costs?
Repeat value	Did the customer renew, reorder, expand, or refer?
Service load	How much time, support, revision, or escalation did the account require?
Payment fit	Did the customer accept the price and pay as agreed?
Process fit	Could your team deliver through its normal process?
Relationship fit	Did communication stay clear, respectful, and productive?

A simple three-point score is plenty for a first pass:

• 3, strong fit: the evidence says you’d pursue this customer again
• 2, mixed fit: the relationship worked, but one important part created cost or risk
• 1, poor fit: the result, margin, payment, process, or relationship makes this a customer you’d decline

Add one sentence explaining each score.

The note matters more than the number, because the note tells you which condition to carry into the ICP.

For a consumer business like High Limb, the records are the point-of-sale system, the online store, the email and text list, and the reviews.

A customer scores a 3 when those records show a taproom visit, a sign-up, an online order a few weeks later, and a second order after that, ideally with a friend’s first order traced back to them.

A customer scores a 1 when the records show a single visit on a promotion, a shipping address outside the three states High Limb can ship to, and nothing since.

The note says which of those facts drove the score.

Don’t let the largest invoice win by default. You’re looking for the customer your business served well without bending itself out of shape.

You should now have a comparison set: strong-fit customers, poor-fit customers, lost deals, and churn reasons, each with a score and a note.

Step 3: Find Patterns Among Your Best-Fit Customers

Now compare your clone group with the customers you’d decline.

Look for conditions that showed up before the purchase:

• the problem that triggered the search
• the cost of leaving that problem alone
• the words they used to describe it
• the previous option that failed
• the person who felt the pressure
• the person who approved the money
• the proof that reduced doubt
• the condition that made delivery easy or hard

Treat age, gender, job title, company size, and location as useful only when they change fit or buying behavior. Decorative details don’t belong in the profile.

End this step with two short lists.

Write “Our best-fit customers usually…” and “Our poor-fit customers usually…”

Every line should point back to a record, a call, an invoice, or a support history.

If the records show you what happened but not why, interview customers from the clone group.

Five is a practical starting point, but if you have fewer, talk to every one who’s willing. Ask about one real purchase:

1. What was happening when you decided to look for help?
2. What had you already tried?
3. Who else was involved in the decision?
4. What almost stopped you from choosing us?
5. What result told you the decision had worked?
6. Where did you look for information while comparing options?

Questions about a real decision give you details you can check.

Questions about what someone might do in the future give you opinions.

For High Limb, the lines I’d expect to test read something like: lives within driving distance of Plymouth or inside the three ship-to states, first touch was a taproom visit or a tasting, joined the list at the bar rather than from an ad, came with a group, and placed the first online order in the weeks after the visit.

Every one of those is checkable against the point-of-sale timestamps, the sign-up source, the postal code, and the order history. Until that check runs, they’re hypotheses, and the profile should say so.

Step 4: Write the Customer’s Job and Trigger

The job explains what progress the customer wants. The trigger explains why the problem became urgent.

I use this format:

When [trigger or situation], I want to [action or progress], so I can [business or personal outcome].

For High Limb’s consumer customer, the public evidence points to a job like this:

When the weather turns and we’re planning a weekend out, I want to find a cidery nearby that’s worth the drive, so I can spend an afternoon somewhere with a good atmosphere and bring home something I can’t get at the grocery store.

That sentence gives your marketing a problem, a moment, and an outcome. “Likes craft cider” gives you none of those.

The trigger sits inside that sentence.

For High Limb, it’s the first cool weekend of fall, or friends and family coming to town, when someone starts searching for a place to go.

The outside research later in this guide backs that up: the searches this audience runs cluster around “near me,” tastings, fall events, and state-by-state “best cideries” lists, and they spike in September and October.

My working hypothesis, and the one High Limb’s order history has to confirm, is that the online order is the second job, not the first: keep drinking at home what we found on the trip.

Step 5: Record Pain, Objections, and Decision Criteria

Pain points should describe consequences.

“People don’t know about us” is vague.

“Someone tried our cider at a friend’s house, liked it, and gave up when they couldn’t find it at their local store” is something you can write to.

These fields are easy to blur together, so here’s how they differ for High Limb’s consumer customer:

• Trigger: the first cool weekend of fall, or visitors coming to town, starts a search for somewhere to go
• Pain: the cider they liked is hard to find again, so the second purchase never happens
• Job: find a place worth the drive, then keep drinking what they found at home
• Objection: “Where do I even get it?” and, for a growing number of people, “Is it gluten-free?” and “How many calories?”
• Decision criteria: atmosphere, food, what’s on the draft list right now, whether it ships to their state
• Proof need: current draft list, a retailer locator that works, recent reviews from people who went

Two of those lines came straight from public evidence.

The gluten-free and calorie questions show up in the search demand around hard cider, so High Limb answers them on its site.

The atmosphere and food criteria come from what customers say in reviews, which we’ll get to in the next step.

The “where do I get it” objection is the one every small producer with a local footprint and an online store runs into, and it’s why the Find Us page exists.

For each pain, record:

• what happens
• who feels it
• what it costs in money, time, risk, or missed opportunity
• what the customer has already tried
• what would make the problem urgent

Then add the objection, your honest response, and the proof required. Don’t write a rebuttal that promises more than you can prove.

Write the objection in the customer’s words.

Follow it with the plain answer you can defend and the proof that supports it.

If the proof doesn’t exist, mark it as a gap instead of writing a stronger claim.

I’d rather see an honest gap in a profile than a confident line nobody can back up.

Decision criteria tell you what the buyer will compare.

Price may matter, but so can speed, location, experience, compatibility, direct access, contract terms, implementation help, and evidence from a similar customer.

Step 6: Capture the Customer’s Exact Language

Copy the words. Don’t polish them.

Pull language from:

• sales and discovery calls
• support tickets and chat
• reviews of your business
• reviews of your alternatives
• lost-deal notes
• customer interviews
• relevant public discussions

Tag each quote with its source and date. Strip out names, phone numbers, email addresses, and account details before the file goes anywhere near an AI tool.

Here are two real entries pulled from public reviews High Limb displays on its own site:

“Love the atmosphere of this place. Cozy, with gathering around the fire pits.” Public review, taproom guest, 2026. Theme: decision criterion (atmosphere).

“Great place to go with friends, especially for girls’ night out!” Public review, taproom guest, 2026. Theme: occasion and who’s involved.

Notice what those two lines already tell you that a demographic never would.

The purchase is social. The setting is part of the product.

And “cozy” and “fire pits” are words High Limb can put on a page, in an ad, and in the instructions it hands an AI tool, because a customer said them first.

Your ICP should summarize the pattern. Keep the full quote library in a separate customer language file so a writer can trace the language back to evidence. Our template calls that file VOICE-OF-CUSTOMER.md, but a shared document with the same structure works just as well.

Step 7: Draft the First Version Before It Feels Complete

Write one primary profile with the evidence you have.

Mark the unknown fields instead of filling them with a plausible story.

The first version should be useful enough to qualify a lead, choose a message, and reject an obvious mismatch.

Research can improve it after the file starts helping with real decisions. It won’t get better sitting in a folder waiting to be perfect.

Hand the draft to one person who didn’t help write it.

Ask them to qualify a real lead or choose a homepage message with it.

Every point of confusion becomes a field to clarify, not another paragraph of background.

At this point the file should explain who fits, why they buy, what they need to believe, and when you should say no. That’s a complete first ICP, and you can start using it right away.

Outside research answers a different question: where people like this spend attention and how they describe the problem.

The next section shows how we answered it for High Limb. If you have no budget for research tools yet, skip ahead to the template and come back later.

---

How Do You Research an ICP With SparkToro?

Your core ICP is already usable.

This section adds outside evidence about where people like your best customers spend attention and how they describe their work.

SparkToro is the audience research tool I reach for here.

It shows how a group describes itself, what it searches for, and where it spends attention online. SparkToro explains that its audience data combines anonymized, aggregated website-visit data from research panels with public LinkedIn profile information.

One rule governs everything in this section.

Your customer records tell you which relationships worked.

SparkToro can reveal the language, interests, searches, and places that attract a related audience. It cannot prove that anyone bought, renewed, paid on time, or fit your delivery process.

Mark every SparkToro finding as an observed audience signal until direct evidence supports it.

You have three routes:

1. Run the research yourself in the SparkToro app.
2. Connect SparkToro to an AI assistant so the assistant requests reports for you.
3. Skip SparkToro and use customer interviews, reviews, search data, and trade sources instead.

The steps below work for the first two routes. If you take the third, the same discipline applies to any outside source.

Step 1: Describe People, Not a Topic

“Hard cider marketing” is a topic. “Adults in New England who drink craft cider and visit cideries” is an audience.

Start with the people you identified from your customer evidence. Add the age range, location, habits, and buying behavior that change the decision. For a business selling to companies, add role, industry, and company size instead.

For High Limb, we used:

Adults aged 25 to 55 in Massachusetts, Rhode Island, and Connecticut who drink craft hard cider, visit local cideries, taprooms, and orchards, and buy craft beverages from small producers online.

Use that description in SparkToro’s audience research app.

If you’d rather go the connected route, SparkToro offers an MCP connector.

MCP is a standard connection that lets an AI assistant request SparkToro reports without you copying data by hand. It requires a paid SparkToro subscription, so check the plan details first.

The connection changes how you request the report. It doesn’t change how you inspect or validate the findings.

If you use the connected route, start with:

Create one SparkToro audience report for the following people in the United States: adults aged 25 to 55 in Massachusetts, Rhode Island, and Connecticut who drink craft hard cider, visit local cideries, taprooms, and orchards, and buy craft beverages from small producers online.

Create the report once. Reuse it for each section so you don’t burn account quota rebuilding the same audience.

Step 2: Check Who the Report Actually Found

Read the demographics, location, and professional signals before you use a single recommendation.

They appear in the demographics section of the report.

This is where the High Limb report split in two, and it’s worth walking through because it’s the kind of thing that fools people.

The behavior side of the report was excellent.

The websites this audience visits were almost all regional and almost all on point: other cideries and orchard destinations with taprooms across Massachusetts, Rhode Island, and Vermont, cider trade and enthusiast publications, the cider industry association, well-known New England craft breweries, and a couple of online craft-beverage delivery shops.

The search behavior was the same story: “near me” searches for cideries and tastings, fall cider events, and “best cideries in” each of the three states, with demand peaking in September and October.

The profile side did not hold up.

The location data read like a national audience, with the three states we named barely registering. And the largest professional group in the role data was real estate agents, with customer service and sales jobs next.

Nobody drinks cider because they sell houses.

What happened is that SparkToro’s location setting works at the country level, so the states in our description shaped the behavior data but not the profile data.

And the professional fields describe people who keep public work profiles, which tells you very little about a consumer audience built around a weekend habit.

We didn’t throw the report out.

We kept the website, publication, and search themes as observed audience signals, because they matched what High Limb already knows about its customers.

We left the location and role data out of the profile entirely and noted why.

That’s the right response to an imperfect report: keep the signal that survives the check, label its limits, and don’t let the charts you can’t explain into the file.

Before you use any report, run this check:

• Does the location data match where your customers actually are?
• Do the roles or demographics match the customer you described, or do they describe something else?
• Are unrelated groups shaping the findings?
• Which sections should you exclude?
• What wording should change before you run the report again?

If the profile layer is wrong, tighten the description or set that layer aside and lean on behavior. Name the actual habits, places, and websites your verified customers use. Don’t keep a section because the charts look nice.

Our correction also changed the next marketing decision.

For High Limb, the report argues for showing up where people search for a cidery to visit this weekend (the Google Business Profile, the taproom and Find Us pages, fall event listings, and the regional cider publications) before spending anything on broad lifestyle or national audiences.

We’d test that first and let the response decide. We wouldn’t move budget on the report alone.

Step 3: Inspect Four Sections

You don’t need every SparkToro section for an ICP. Start with four:

1. Demographic and role signals: Check whether the report matches the customer you intended to study, and set aside the parts that don’t.
2. Bio language: Find the words people use to describe their work, interests, or identity. For a consumer audience this section is often thin or off-target, as it was for High Limb, so don’t force it.
3. Websites visited: Identify the destinations, publications, associations, stores, and communities that earn attention.
4. Search behavior: Find the places, occasions, comparisons, and questions the audience researches.

For High Limb, the useful patterns grouped around regional cideries and orchards with taprooms, cider publications and the industry association, New England craft breweries, online craft-beverage retailers, and searches about finding a cidery, a tasting, or a fall event nearby.

Those themes belong in hypotheses about occasions, content, partnerships, and distribution. They don’t prove why any specific customer bought.

For a connected AI assistant, request those sections from the existing report:

Use the audience report you already created. Show me the role and company signals, bio language, websites visited, and search behavior. Do not create a new report. Separate observed data from your interpretation.

Step 4: Translate Signals Into Decisions

Don’t paste a list of domains or keywords into your ICP and call it research.

Run each signal through this four-part note:

Field	What to Write
Source	Where the signal came from
Observation	What pattern appeared
Possible meaning	What the pattern may say about the customer
Validation	Which customer record, interview, or campaign can confirm it

Here’s one from the High Limb work:

Field	High Limb Working Note
Source	SparkToro audience report
Observation	Other regional cideries, orchards, and craft breweries with taprooms appear together, and “near me” tasting searches peak in early fall
Possible meaning	The customer is choosing a destination for an outing, not a brand of cider, and compares High Limb with every other place worth the drive that weekend
Validation	Check point-of-sale and sign-up timestamps by month, and ask taproom guests where else they went this season

The observation is a reported audience signal. The possible meaning is a working hypothesis. Keep those labels intact.

Step 5: Use Search Behavior to Improve the Buying Job

Look for clusters, not isolated keywords.

For High Limb, searches for cideries near me, cider tastings, orchard tastings, fall cider events, and the best cideries in each state all point to one larger job: pick a place worth the drive this weekend.

That’s the job I wrote in Step 4, and it came from this cluster plus the reviews, not from a guess about what cider drinkers want.

It also raised the most useful question in the whole exercise.

Almost none of the search demand is about buying cider online. So does the online order follow the visit, or is it a separate customer?

Only High Limb’s order history can answer that, which is why it sits in the unknowns instead of the profile.

Write the job in customer language, then verify it in calls and records. Don’t force your offer into every search theme. Some signals will describe related problems you don’t solve, and that’s fine.

Step 6: Use SparkToro’s Persona Output as a Draft

If your plan includes Persona Builder, SparkToro can generate editable personas from audience reports. It combines observed demographics and behavior with an AI-written layer of context.

That output can speed up the person-level part of your research.

Review every pain point, buying trigger, and journey against your records before it goes into the ICP. SparkToro reports aggregated audience behavior.

Your business records show which customer relationships actually worked.

Step 7: Save Conclusions, Sources, and Questions

Your final ICP should stay readable.

Summarize the conclusions that affect decisions, then keep a short source note for each one.

Here’s what that looks like for High Limb, using only what we can actually stand behind:

• Verified customer evidence: Customers describe the taproom by its atmosphere, the fire pits, the food, and going with a group. Source: public reviews, 2026.
• Verified customer evidence: Online orders can only ship within Massachusetts, Rhode Island, and Connecticut, so anyone outside those states has to visit or find a retailer. Source: the online store.
• Observed audience signal: Regional cideries, orchards, and craft breweries with taprooms appear together, and searches for a cidery or tasting nearby peak in September and October. Source: SparkToro audience report.
• Working hypothesis: The customer is choosing a destination for an outing first and a cider brand second. Needs a seasonality check on point-of-sale and sign-up data plus a short guest survey.
• Unknown: Whether first online orders follow a taproom visit or come from people who’ve never been. Needs the order history matched against the sign-up source and postal code.

Save only the findings that affect a decision. This format keeps the research useful without making it sound more certain than it is.

You now have two sets of inputs: direct customer evidence and outside audience signals.

The template below turns them into one file without erasing the difference.

---

What Should an Ideal Customer Profile Template Include?

An ideal customer profile template should record fit, buying triggers, the customer’s job, pain points, objections, decision criteria, and disqualifiers.

It should also hold exact language, research habits, evidence status, an owner, and a change history.

Keep these fields in one structured file so your team and AI tools work from the same evidence.

I keep ours in Markdown because the file stays readable, searchable, linkable, and easy to update. Your team can trace a claim to its source, and an AI tool can read the same structure without pulling text out of a designed document.

A shared document with the same headings works too.

The completed example below shows what finished looks like. The blank template follows it.

What Does an Ideal Customer Profile Example Look Like?

This is High Limb’s consumer profile after the seven steps and the SparkToro research, built from public evidence.

The fields that need their private records are marked unknown on purpose. That’s what a first version should look like: honest about what it knows, specific about what it still has to find out.

Offer supported: High Limb’s consumer cider business: taproom visits, online orders, and retail purchases in Massachusetts, Rhode Island, and Connecticut. The trade side (bars, restaurants, package stores, distributors) has its own profile.

Decision supported: Decide which visitors and subscribers get which follow-up, which message leads on the site and in email, and when to route someone to the trade side instead.

Primary customer: An adult within driving distance of Plymouth or inside the three ship-to states who treats a cidery visit as a weekend outing and wants to keep drinking what they found once they’re home.

Buyer summary: The person planning the outing. Reviews show the visit is social: friends, a night out, family in town. Whoever picks the destination is the buyer for the group.

Job to be done: When the weather turns and we’re planning a weekend out, I want to find a cidery nearby that’s worth the drive, so I can spend an afternoon somewhere with a good atmosphere and bring home something I can’t get at the grocery store.

Trigger: The first cool weekend of fall, or visitors coming to town, starts a search for somewhere to go.

Pain: The cider they liked is hard to find again, so the second purchase never happens.

Objection: “Where do I even get it?” plus “Is it gluten-free?” and “How many calories?”

Decision criteria: Atmosphere, food, what’s on the draft list right now, recent reviews, and whether it ships to their state.

Disqualifier: Lives outside the three ship-to states and isn’t planning a visit, so there’s no way to serve them beyond a retailer they may never be near. Also anyone asking about wholesale, draft accounts, or distribution: real customer, wrong profile, route to trade.

Research habit: Searches for cideries, tastings, and cider events nearby, especially in early fall. Visits other regional cideries, orchards, and craft breweries with taprooms. Reads cider publications and follows the brands it likes.

Evidence map: Ship-to states, the taproom, the product lines, and the Find Us retailers come from High Limb’s site and store. Atmosphere, food, and the social occasion come from public reviews. Destinations, publications, and search patterns come from SparkToro and stay observed audience signals. The idea that the outing comes first and the online order second is a working hypothesis.

Unknowns: Whether first online orders follow a taproom visit or come from people who’ve never been. Which of the three states produces the most repeat orders. Whether the fall spike in searches shows up in High Limb’s own visits and sign-ups. What share of taproom guests come from outside the ship-to states. Every one of those has a record that can answer it.

Don’t add personal trivia unless it changes the buying decision.

A useful profile needs fit, triggers, jobs, proof, and disqualifiers, not a stock photo or invented lifestyle details.

You’ll notice this one has no age, no income, and no job title, because none of the evidence we have says those change whether someone drives to Plymouth.

Copy the Template

Our public ICP template gives you the complete working folder: the master ICP.md, a Voice-of-Customer library, a buying-committee template, and a filled example.

The filled example in the folder is a fictional business-to-business one, so you can see how the same format handles a company-level customer with several people involved in the purchase.

A buying committee is the list of people who influence one purchase, such as the owner who approves the budget and the operations lead who researches options.

The shorter version below has the fields you need to start in an existing document. Copy it first, then fill only the fields your evidence supports. Leave the rest marked unknown.

# Ideal Customer Profile

Last updated: [YYYY-MM-DD]
Owner: [Name and role]
Offer supported: [Specific offer]
Decision supported: [What this ICP helps you decide]

## Primary Customer
[One sentence naming the person, account, need, and fit]

### Fit
- Customer or company type:
- Size, budget, or buying condition:
- Industry or category:
- Geography:
- Business model or purchase pattern:
- Tools or process signals:

### Buyer
- Role:
- Responsibility:
- Decision authority:
- Other people involved:

### Job to Be Done
When [situation], I want to [progress], so I can [outcome].

### Trigger Events
- [Specific event that makes the problem urgent]

### Pain Points
1. [Problem, consequence, and evidence]

### Objections
- Objection:
- Honest response:
- Proof required:

### Decision Criteria
1. [Criterion the buyer will compare]

### Disqualifiers
- [Objective condition that makes the customer a poor fit]

### Language
- "[Exact customer quote]" (source and date)

### Content and Research Habits
- Searches:
- Websites and communities:
- Trusted sources:
- Preferred formats:

### Evidence Status
- Verified customer evidence: [statement, source, and date]
- Observed audience signals: [pattern, source, and date]
- Working hypotheses: [possible meaning and planned test]
- Unknown: [open question and how you will answer it]

### Where This ICP Is Used
- HubSpot: [fields or profile section that use this ICP]
- Canva: [campaign brief or Brand Voice guidance]
- Other systems: [location and purpose]

### Scheduled Listening and Review
- Approved public sources: [forums, review sites, and social searches]
- Scheduled research tool: [tool and saved instructions]
- Review cadence: [when a person checks the findings]
- Evidence queue: [where unverified findings wait for review]

## Change Log
- [YYYY-MM-DD]: [What changed, why, and source]

Run the Five-Decision Test

A profile isn’t finished because every field has text in it.

It’s finished when two reviewers can use the same evidence to reach a clear decision, or name the same missing fact.

Give the ICP to one person who didn’t write it. Give the same file to an AI assistant. Ask both to answer:

1. Should we pursue this lead?
2. Which problem should the homepage lead with?
3. Which proof should the sales page show?
4. Which content topic should we create next?
5. Which channel deserves a small test?

Use this prompt with a real, redacted prospect or campaign:

Read the ICP below, then evaluate this prospect or marketing decision. Return: fit or no fit, the three ICP facts that drove the answer, missing evidence, the best next action, and the exact section of the ICP that needs clarification if the decision is uncertain. Do not invent facts that are not in the file.

Compare the evidence and reasoning behind the answers. The test passes when both reviewers cite the same fit rules, name the same missing evidence, and invent nothing.

If they disagree, look at why.

A disagreement caused by missing prospect information belongs in your next sales question.

A disagreement caused by an unclear threshold or disqualifier belongs in the ICP.

Once the file can carry those decisions, its next job is to become part of the daily work and stay current.

---

How Do You Use and Maintain an ICP?

Use the ICP anywhere someone chooses a customer, message, offer, proof point, or channel.

Put the ICP Into Daily Marketing Decisions

Website: Keep the ICP next to the page brief. Check whether each headline, service description, proof point, and call to action matches the buyer’s job, trigger, objection, and decision criteria. A practice-area page, a neighborhood page, a menu page, and a product page all get the same check.

Content and search: Start with the questions your ideal customer asks while feeling the problem, comparing options, and trying to reduce risk. Use the customer’s own language wherever it matches verified search demand.

Paid ads: Target only the fit conditions the platform can actually reach. Lead with the trigger or consequence, then send the click to a page built for the same customer and offer.

Sales and intake: Add fit, trigger, need, authority, timing, and disqualifier fields to the CRM or intake form. Use dropdowns for facts you need to compare and free-text notes for context. A law firm’s intake questions, a Realtor’s buyer consultation, and an online store’s post-purchase survey are all places these fields belong.

Customer response: Route each inquiry by need and urgency. Give whoever responds, whether that’s a person or an AI agent, the right proof and objection guidance before the first reply goes out.

AI tools: Load the same ICP and Voice-of-Customer files before you ask for an email, page, ad, or article. Keep those files attached or connected so you’re not rewriting the audience inside every prompt.

Here’s the instruction I use with the full files:

Before you draft, read /01-icp/ICP.md and /01-icp/VOICE-OF-CUSTOMER.md. Name the primary customer, job to be done, trigger, pain, objection, proof need, and disqualifier that apply to this task. If the files do not support a decision, stop and list the missing evidence. Write to one reader and use only language supported by the files.

Keep One Master File

Pick one master file. Link to it from your CRM, project templates, writing instructions, and AI workspace.

Don’t copy the full ICP into six tools. Copies drift. Give each tool the master file or a maintained connection to it.

Assign one owner. That person doesn’t decide the ICP alone. The owner collects evidence, runs the review, records changes, and keeps duplicate versions from spreading.

Put the ICP Into HubSpot, Canva, and Your Other Working Tools

The ICP only earns its keep when it shows up at the point where your team qualifies a lead, chooses a message, or builds a campaign.

Add the parts each system needs, and keep the complete logic and evidence in the master file.

HubSpot: HubSpot should carry the fields your team needs while qualifying a lead.

A Super Admin or Partner Admin can add the profile to HubSpot’s AI context so its AI tools use it as business context. Go to Agents > Agent Hub, open the Context tab, then the Customers tab, and add the ICP in the Ideal Customer Profiles section. The same tab has a separate Personas section for the people inside that buying decision, which lines up with the ICP and buyer summary split in this guide. HubSpot documents the steps in Manage AI context.

Map your measurable fit conditions, such as industry, revenue, geography, and customer type, to company properties. The default Ideal Customer Profile Tier property stores the resulting Tier 1 to Tier 3 classification. If your subscription includes HubSpot’s account-based marketing tools, you can assign the tier manually or with a workflow and add buying roles to contact records. HubSpot lists the subscription requirements.

The tier is the result of your rules. It isn’t the rulebook. Link the HubSpot setup back to the master ICP so anyone can see why a company got its classification.

For a consumer business the fields are simpler, and High Limb is a good example of how little it takes.

The one fact that changes the next step for a new subscriber is the postal code. Inside the three ship-to states, the follow-up leads with the online store.

Outside them, it leads with the Find Us page and an invitation to visit the taproom. A second field, whether the person asked about a wholesale or draft account, routes them to the trade side entirely.

Two fields, pulled straight from the ICP’s fit conditions and disqualifiers, and every automated email already knows which customer it’s talking to.

Canva: Canva is where your team applies selected parts of the ICP, not where the complete profile lives. Put the customer fit, main pain, and offer into the Target audience or ICP section of Canva’s go-to-market strategy template so every campaign brief starts from the same customer.

Use Canva’s Brand Voice only for short language and communication guidance. Its 500-character limit fits the customer’s vocabulary and tone needs, not customer evidence or fit rules. Team members can apply that guidance when they use Magic Write. Customer evidence, disqualifiers, sources, and change history stay in the master file.

For any other CRM, ad platform, design tool, or AI workspace, look for fields named ideal customer, target audience, buyer, brand context, or selling profile. Copy only the fields that tool needs, keep a link back to the master file, and add the tool to your ICP change checklist. That keeps the working copies useful without letting them turn into competing sources of truth.

Schedule an ICP Listening Loop

Your customer records explain what already happened.

Public conversations can show you new language, complaints, alternatives, and questions before they ever reach a sales call.

A listening loop is a scheduled check of those public conversations. It collects possible updates in a review queue. It should never rewrite the ICP on its own.

You can run it at three levels:

1. Manual: Once a month, read new reviews of your business and your alternatives, skim the trade forums and communities named in your ICP, and paste anything useful into a simple queue document with its source and date.
2. Assisted: Paste your ICP into an AI assistant and ask it to summarize what it finds in the sources you name, with links, so you can review the findings faster.
3. Connected: Use an AI agent that can run on a schedule, monitor sources you approve, and save citations. Depending on what you already use, that could be Grok, Hermes, OpenClaw, Perplexity Computer, or a similar research agent. Point it at the Reddit communities, X discussions, LinkedIn posts, review sites, and trade forums where your ideal customer talks about the problem or your brand.

Start at the level you can sustain.

And start with the sites, communities, search terms, and alternatives already named in the ICP. A broad alert for every mention creates more noise than evidence.

For the assisted or connected route, run a prompt like this:

Monitor the approved public sources below for new conversations from people or companies that match our ICP. Focus on mentions of our brand, the problem we solve, complaints, desired outcomes, alternatives, buying triggers, objections, and the exact words people use.

For each new finding, return:
1. the exact quote or a short faithful excerpt
2. the source URL and publication date
3. why the source appears to match the ICP
4. the theme, such as trigger, pain, objection, decision criterion, complaint, review, or competitor
5. whether it is direct brand evidence or broader category evidence
6. the ICP field it might inform

Compare the findings with the prior report and remove duplicates. Do not change the ICP. Add the findings to an evidence queue for human review. Do not collect private-group content or personal information that is not needed for the research.

Treat every result as an observed audience signal, not a verified customer fact.

The ICP owner reviews the source, removes poor matches, and looks for repeated patterns. A useful pattern becomes a customer-interview question, a CRM check, or a small campaign test. It enters the verified section only when direct customer or transaction evidence supports it.

Platform access matters. Use the platform’s official integrations, approved connectors, public search, native alerts, or source lists you’re allowed to monitor.

LinkedIn’s User Agreement limits automated collection, so don’t scrape discussions or private groups to fill the report.

Save each approved finding with its source and date.

At the next ICP review, ask which language changed, which complaint repeated, which alternative showed up, and whether any fit rule or marketing decision should change.

That turns maintenance into a repeatable research habit instead of a rewrite from memory.

Review It When Evidence Changes

My working default is to review the ICP after every 10 closed deals or once a quarter, whichever comes first.

Adjust that when your deal volume or buying cycle calls for it. Also review the file when:

• a new offer attracts a different buyer
• repeated qualified deals are lost for the same reason
• service costs or margins change
• a new geography or channel changes the customer mix
• customer language shifts across calls and reviews
• SparkToro or another audience source reveals a behavior worth testing

The review should answer four questions:

1. Which statement gained evidence?
2. Which statement lost evidence?
3. Which decision changed because of it?
4. What should the team test before the next edit?

Record every material change in the file.

A change log is how you tell a market shift from one unusual customer.

Measure Whether the ICP Improves Decisions

An ICP earns its place by improving the work, not by looking complete.

Pick a small set of outcomes tied to its job. Record a baseline from a period long enough to reflect your normal buying cycle, then compare results by fit level.

Assign the fit level as soon as sales has enough information to qualify the lead.

Don’t wait until the deal is won or lost.

Knowing the result colors the label and makes the profile look more accurate than it was.

• percentage of qualified leads that match the profile
• close rate and sales cycle by fit level
• renewal, repeat purchase, referral, or return rate by fit level
• delivery time and service cost by fit level
• content and campaign results for the primary customer
• number of unclear decisions found in the five-decision test

Don’t credit the ICP for every improvement. Use the results to find where the profile predicts fit and where it still misses.

---

Frequently Asked Questions About Creating an ICP

Build Your Ideal Customer Profile

Your first ICP doesn’t need every answer.

It needs enough evidence to guide one real decision and enough honesty to show where evidence is missing.

Build it from the records you already have.

Test it on one real lead, page, campaign, or content decision. Turn every unclear answer into the next research question.

Open the FREE ICP starter folder to copy the full ICP.md, VOICE-OF-CUSTOMER.md, and BUYER-COMMITTEE.md templates.

You can use this guide and the template to keep the work in-house.

If you’d rather not, we can structure the first version for you or connect it to the tools you already use.

And when the ICP is one piece of a larger marketing problem, we can build and maintain the broader AI marketing system around it.

---