An ideal customer profile (ICP) describes companies your offer can serve well and profitably: their problem, operating conditions, buying roles and essential requirements. Build it from customer outcomes and won and lost deals, then use it to prioritize accounts, qualify inquiries and plan campaigns. Assess company fit, buying timing and missing information separately.
Two manufacturers work in the same industry and have the same headcount. One needs a packaging line you can integrate. The other needs a capability your solution doesn’t support. A working ICP tells them apart before your sales team finds out the hard way, and helps you choose the next step.
Download the ICP template · Download the evidence worksheet · Get the worked deal comparison
What is an ideal customer profile?
An ideal customer profile is a practical description of an account that is a strong match for a particular offer. It should explain why that company can benefit, why the work makes commercial sense for you and which observable characteristics help you recognize it.
“US manufacturers with 50 to 500 employees” gives you a starting list. A working profile goes further: food producers with a packaging bottleneck, equipment your team can integrate and enough potential project value to justify installation and support. Headcount is useful only if it helps predict something that matters.
Customer value belongs in the definition. Salesforce’s ICP guide recommends examining lifetime value and retention alongside customer goals and buying triggers. A large contract that is expensive to deliver or leaves the customer struggling should prompt a closer look.
ICP vs. target market vs. buyer persona
| Tool | What it describes | The decision it supports |
|---|---|---|
| Target market | The broader set of businesses that could need your category | Where to look for demand |
| ICP | Companies with a problem, operating conditions and economics that suit your offer | Which accounts deserve priority |
| Buying committee | The roles involved in a particular purchase | Who needs to participate and what can block approval |
| Buyer persona | A research-based description of a person or role, including their priorities and concerns | What to explain to each participant |
For an automation project, the ICP describes the plant. The committee might include a plant manager, process engineer, food safety specialist and finance lead. The engineer’s persona helps you prepare a technical discussion. It does not establish whether the plant is a viable account.
An ideal client profile usually serves the same purpose for a service business. Use the wording your team understands, and keep the underlying decisions explicit.
A nine-field ideal customer profile template
Write one profile for a defined offer and market. If the offer is “packaging-line integration,” avoid mixing evidence from spare-parts orders into the same analysis.
Use these nine fields as a compact front sheet. Keep the supporting records in the evidence worksheet, where each criterion has a source, date, counterexample and next check.
| Field | What to write | Where to check it |
|---|---|---|
| 1. Company and operating context | Industry, geography, business model and relevant scale or complexity | CRM, company research, sales records |
| 2. Problem and desired outcome | The job the company needs done and what a useful result looks like | Buyer interviews, calls, implementation reviews |
| 3. Delivery requirements | Technology, operating conditions, capabilities and constraints needed for success | Product or delivery team, technical discovery |
| 4. Commercial fit | Project or contract economics, cost to serve and purchasing route | Finance, delivery records, renewal history |
| 5. Buying committee | Users, sponsor, approver and potential blockers | Actual deal contacts and interviews |
| 6. Trigger and timing | The event that creates a reason to act, and how to verify it | Buyer confirmation, dated public signals |
| 7. Disqualifiers | Confirmed conditions under which this offer cannot work | Product limitations, delivery constraints, economics |
| 8. Observable signals | What you can research before a conversation, and what remains unknown | Company sources, CRM, relevant public records |
| 9. Evidence and review | Supporting and conflicting records, confidence, owner and next review | The linked evidence worksheet |
A field such as “has budget” needs care. The cost of your solution may be incompatible with a company’s economics, which is a fit issue. A suitable company waiting for next year’s capital approval has a timing issue. “Budget not recorded” is missing information.
How to build an ICP from customer evidence
Start with customers who received value and were worth serving. Then compare their characteristics with relevant losses and disappointing wins. This tests whether a familiar trait actually helps you choose.
1. Define the offer and comparison group
Choose a period that covers a meaningful sales cycle and a reasonably consistent offer. Include won deals, lost deals and decisions to do nothing. Review customer outcomes after the win when enough time has passed.
A useful export contains an account identifier, offer, source, opportunity dates, outcome, value, loss reason and the candidate characteristics you want to examine. Add delivery cost, retention or repeat work where they apply. Keep unknown values visible.
For the win-rate comparison below, count closed no-decision outcomes among losses and keep “no decision” as a separate reason. Exclude open opportunities from that calculation.
Record characteristics as they were known before the outcome. A detail added after a successful implementation may explain satisfaction, but it would not have helped you qualify that account at the start.
When historical deals are a poor guide
Past deals also reflect whom you approached, which channels you used and what you could deliver at the time. If you advertised only to multi-site plants, the CRM cannot tell you how unapproached single-site companies would have responded. A missing segment is not evidence of poor fit.
A new product, market, price or delivery capability may change the relevant criteria. Keep the old profile as a dated baseline and test the new segment separately through buyer conversations and new opportunities. Compare outcomes after a full sales cycle, keeping the offer and acquisition source visible. Do not mix the new segment into the old win-rate comparison before checking whether the groups are comparable.
2. Check who benefited, including after the sale
Ask delivery and customer success what happened after the contract. Did customers achieve the intended result? Did onboarding take far more work than expected? Did they renew, expand or buy again?
Separate a good purchase from a good relationship. For project work, examine delivery effort and follow-on demand. For subscriptions, examine adoption and retention over comparable periods. A new customer has not yet had the same chance to renew as an older one.
3. Compare each proposed trait with an alternative
Here is a hypothetical dataset of 24 closed opportunities for the same integration offer. It demonstrates the calculation, not Wonderflow client results. Each trait is a separate view of the same 24 opportunities, so the rows across the two comparisons must not be added together.
| Proposed trait | Won | Lost | Closed opportunities | Win rate |
|---|---|---|---|---|
| Food and beverage company | 10 | 10 | 20 | 50% |
| Other manufacturer | 2 | 2 | 4 | 50% |
| Two or more sites | 9 | 3 | 12 | 75% |
| One site | 3 | 9 | 12 | 25% |
Win rate = won opportunities ÷ (won + lost opportunities).
Ten of the twelve wins came from food and beverage companies. That sounds persuasive until you see that ten losses came from the same industry. In this dataset, its win rate is identical to the alternative. The industry label alone does not distinguish more likely wins.
The multi-site pattern is more promising: nine wins out of twelve opportunities, compared with three out of twelve for single-site companies. It earns further investigation.
Before making “two or more sites” a rule, check deal source, project scope, margin and what customers achieved. Perhaps the multi-site opportunities came through an experienced partner, while single-site inquiries came from broad advertising. Site count might be standing in for a better source or a different project.
Keep the three single-site wins in view. They are counterexamples to a blanket ban. The initial action is to prioritize multi-site research and investigate the exceptions, then test the criterion on new opportunities. This small sample does not establish cause or a reliable prediction of future win rates.
A complete evidence record
Keep the comparison and the decision together. Here is the site-count record filled in for the same hypothetical dataset. The date identifies when we prepared this teaching example, not when real deals occurred.
| Evidence field | Filled record |
|---|---|
| Criterion | Number of production sites |
| Rule or hypothesis | Prioritize research into multi-site companies, while investigating viable single-site projects |
| Why it matters | Multiple sites may support a larger or repeatable project. This is a hypothesis to test. |
| Supporting evidence | Multi-site: 9 won + 3 lost = 12 closed opportunities, 75% win rate |
| Contradictory evidence | Single-site: 3 won + 9 lost = 12 closed opportunities, 25% win rate. The three single-site wins challenge a hard exclusion, and three multi-site losses show the trait is insufficient. |
| Source and date | 24 hypothetical deal records, teaching example prepared October 22, 2026 |
| Confidence | Provisional. Small selected sample, with acquisition-source differences not ruled out. |
| How to verify on a new account | Confirm site count and the actual project, rather than inferring need from company size |
| Decision | Give multi-site research priority. Site count alone cannot qualify or disqualify an account. |
| Owner | Sales lead, an illustrative role in this exercise |
| Next review | When the next comparable cohort has had a full sales cycle. Compare both groups, source mix and delivery outcomes. Log new counterexamples as they appear. |
The evidence worksheet also contains completed records for service area, technical requirements, project economics and buying timing, plus blank rows for your own criteria. It distinguishes a pattern in deals from a supplier’s chosen operating rule.
4. Ask buyers what changed
Interview customers, lost buyers and people who stayed with their existing approach. Include the people who used the solution as well as those who approved it. There is no universal interview count that makes an ICP valid.
Useful questions follow the purchase:
- What happened before you started looking?
- How were you handling the problem, and what stopped working?
- Which alternatives did you seriously consider?
- Who became involved, and what did each person need to approve?
- What nearly stopped the decision?
- What happened after implementation?
Record the buyer’s actual words when you have them. An AI-written summary or a salesperson’s interpretation should remain distinguishable from a quotation.
5. Classify loss reasons before creating exclusions
A loss can expose a poor match, a timing problem or something you need to improve.
| What happened | Interpretation to investigate | Initial action |
|---|---|---|
| The required integration is unsupported | This offer may be unable to deliver | Confirm the requirement, then disqualify if it is essential |
| The project was postponed until a site expansion | The account may fit, with later timing | Record the event and follow-up date |
| The buyer chose a supplier with better implementation proof | Your evidence or offer may need work | Examine the objection and competing proposal |
| Sales never reached the relevant stakeholder | The buying process was incomplete | Repair the qualification or follow-up process |
| No reliable reason was recorded | The outcome is known, the explanation is not | Investigate before changing the profile |
“Lost” is an outcome. It does not explain why the company was unsuitable.
Two filled B2B ideal customer profile examples
These profiles are our worked examples for two hypothetical offers. They show how to write usable fields. Their criteria require validation before use in a real business.
Example 1: an automation integrator
Offer: a standard end-of-line palletizing package for sealed cartons at food and beverage plants.
For this exercise, the integrator offers one defined configuration. The service boundary, supported configuration and commercial threshold below are hypothetical business rules, not industry benchmarks.
| Field | Filled profile |
|---|---|
| Company and operating context | Food and beverage plants within 150 road miles of a service base in Columbus, Ohio. Multi-site operators receive research priority, without excluding single-site projects. |
| Problem and desired outcome | A bottleneck in palletizing sealed cartons. Confirm required throughput and operating hours during technical discovery. |
| Delivery requirements | A dry operating area and an existing Siemens S7-1200 controller with a documented interface the delivery team can verify. This standard offer does not cover other controller configurations or direct-washdown operation. |
| Commercial fit | Expected project contribution margin must be at least 30% of the quoted price after equipment, engineering, installation, travel and included first-year service costs. Confirm the purchasing route separately. |
| Buying committee | Plant manager sponsors the project, process engineer confirms the interface, finance approves capital spending. Verify the actual people and approval sequence. |
| Trigger and timing | A new packaging line or a confirmed bottleneck, with an identifiable investment decision. Unapproved funding means timing needs qualification. |
| Disqualifiers | An essential requirement outside the standard configuration, a location outside the service boundary, or a scoped quote below the contribution threshold. Consider a different offer separately. |
| Observable signals | Expansion news, plant location and relevant technical hiring start research. Controller configuration, the actual task and approval status need confirmation. |
| Evidence and review | The separate 24-opportunity hypothetical dataset suggests a multi-site hypothesis. The service, configuration and contribution limits are chosen exercise rules. Review delivery scope and cost before a quote, and the hypothesis after new outcomes mature. |
The 24-deal file does not record controllers, locations or costs, so it cannot validate these package conditions. It supplies a site-count hypothesis to investigate, not evidence for the whole profile.
For this rule, expected project contribution margin = (quoted price − scoped delivery and service costs) ÷ quoted price. It is a project-level estimate, not net profit or customer ROI. A hypothetical $200,000 quote with $140,000 of those costs leaves $60,000, or 30%, meeting the threshold exactly. Recheck the estimate when scope or included costs change.
Apply the profile to two companies
Both companies below are hypothetical food producers with 150 employees and two sites. Those shared filters put both on the research list. The next checks lead to different decisions.
| Check | Company A | Company B |
|---|---|---|
| Service location | 80 road miles from the Columbus base | 110 road miles from the same base |
| Task and configuration | Sealed cartons in a dry area. The delivery team confirms the existing S7-1200 interface is within the standard package. | Sealed cartons in a dry area. The existing controller is Allen-Bradley ControlLogix, and the buyer requires it to remain. The delivery team confirms this is outside the standard package. |
| Economics | Preliminary scoping: $200,000 price, $140,000 costs, 30% expected contribution margin. Recheck after confirming throughput and operating hours. | Not scoped. The essential configuration fails first, so a margin has not been estimated. |
| Timing | Planned first-quarter 2027 project. Capital approval is pending. | An active project does not resolve the configuration mismatch. |
| Decision | Continue discovery to confirm throughput, operating hours and the investment approval process. Fit remains provisional, and this is not yet a sales-qualified opportunity. | Decline this standard offer and explain the incompatible requirement. Assess a custom supplier or a separate offer if appropriate. |
For Company A, confirm throughput and operating hours, update the scoping estimate and establish the capital-approval step. For Company B, more visits or a larger budget would not remove the essential mismatch. Neither decision uses site count as proof of fit.
Example 2: B2B financial consolidation software
Offer: recurring software for finance teams combining results across legal entities.
| Field | Filled profile |
|---|---|
| Company and operating context | Multi-entity businesses with a central finance team and recurring consolidation work. Entity count and reporting complexity matter more than a headcount band alone. |
| Problem and desired outcome | Repeated spreadsheet reconciliation delays consolidated reporting. The team wants a repeatable close with traceable adjustments. |
| Delivery requirements | Supported accounting systems, accessible source data and reporting requirements the product can actually meet. |
| Commercial fit | The cost of implementation and subscription makes sense relative to the workload. The customer has someone able to own adoption. |
| Buying committee | Controller or head of finance as sponsor, CFO as economic approver, accounting users and IT for access and integration. |
| Trigger and timing | Acquisition, a new entity, a new finance leader or a change to reporting requirements. Establish whether it creates an active project. |
| Disqualifiers | A required unsupported system, no consolidation use case, or a requirement for outsourced bookkeeping rather than software. |
| Observable signals | Public acquisition news and finance hiring can start research. They do not establish a confirmed need, budget or implementation date. |
| Evidence and review | A starting hypothesis. Validate with interviews and early implementations, then examine adoption, delivery effort and renewals. |
The two profiles share a structure, but the reasons to prioritize, disqualify and follow up are different. That difference is what makes each profile useful.
Use fit, timing and evidence to choose the next action
You can begin without a points system. Give each researched account three separate fields:
- Fit: meets the essential conditions, fails a confirmed condition, or remains unknown.
- Timing: active buying situation, later or inactive, or unknown.
- Evidence: source and date for the conclusion, including the next question.
In the worked example, Company A needs further qualification and Company B fails a confirmed requirement. A technically suitable company whose project is postponed belongs in follow-up. An account with an unknown controller configuration needs investigation before a fit decision.
If you introduce ICP scoring, start with explainable rules. A confirmed technical blocker should not be canceled out by points for employee count or website visits.
HubSpot’s scoring documentation distinguishes scores based on attributes from scores based on engagement. That separation is useful even if your system is a spreadsheet. Engagement tells you someone interacted. It does not establish that you can serve their company.
An ICP also does not make every matching contact sales-qualified. The contact’s role, problem and buying situation still need checking. See our guide to MQL vs. SQL and lead qualification for that handoff.
Turn the ICP into account research, content and campaigns
Keep the reasoning when moving from the profile into tools. Some criteria can become targeting settings. Others need a conversation, a form field or a check by the delivery team.
LinkedIn: choose accounts, then buying roles
Build a named-company list from the company criteria and verify the entries. Add identifiers such as company website or LinkedIn Page URL to improve matching. Then consider the relevant job functions or seniority levels.
LinkedIn’s current company-list requirements distinguish the upload threshold from audience eligibility: at least 300 rows are required for an upload, and at least 300 matched member accounts are needed to use the list in an active ad set. Adding role filters can narrow the final audience further, so check its eligible size.
A list of companies does not tell LinkedIn which plants have a packaging project. Research that separately. If the final audience is too small, assess another route to those accounts or a broader segment whose fit you can justify. Adding unsuitable companies merely to meet a threshold defeats the profile.
Review any available audience expansion or automated targeting settings against the purpose of the campaign. Our LinkedIn Ads service explains how that account selection connects to campaign work.
Google Search: translate the problem into intent
For the integrator, “palletizing robot integration” is a candidate commercial search. “Palletizing robot jobs” suggests a different task. These are examples of intent to investigate, not measured keyword recommendations for this industry.
Use the ICP to decide which searches to research, what the landing page must explain and what the inquiry form should establish. A service-area requirement belongs in location planning and qualification. The number of production sites may need a form question.
Apply negative keywords to unwanted search intent with care. Excluding the broad word “parts” could also remove a buyer searching for an integration project involving existing parts. Google’s negative keyword guidance explains that negative match types behave differently from positive keywords and do not cover close variants automatically.
Customer Match is another distinct mechanism. Its supported matching data includes customer contact information. A list of company names or domains alone is not a Google Search account-targeting list.
For the campaign workflow, see Google Ads.
Pages and sales: show the proof the buyer needs
The process engineer needs equipment compatibility and implementation details. Finance needs a credible project scope and purchasing discussion. Put relevant proof where each person can find it, and ask only the qualifying questions needed to choose the next step.
Then connect accepted opportunities and outcomes back to the original inquiry. Enhanced conversions for leads supports attribution of imported offline lead conversions. Whether a particular event should guide bidding depends on its meaning, reliability, volume and delay.
How to validate and update your ICP
Date and save the profile before testing it on new opportunities. Assign fit using information available at that point, record the reasoning and keep exceptions visible.
Choose a review window long enough for the relevant outcomes to mature. A long industrial sales cycle and a short software purchase need different windows.
| Check | Comparison | What could change your decision |
|---|---|---|
| Qualified demand | Relevant inquiries and accepted opportunities from each fit group | The filters attract the right-looking companies but the wrong use case |
| Sales outcomes | Won ÷ closed opportunities, deal value and sales cycle | The proposed priority group performs no better, after checking source and timing |
| Delivery economics | Contribution after the relevant delivery and support costs | Larger contracts absorb more work than their value justifies |
| Customer outcomes | Adoption, intended results, repeat work or retention | Companies buy but struggle to benefit |
| Exceptions and unknowns | Good customers outside the profile and unresolved records | A rule is too narrow, or missing data is being mistaken for poor fit |
This is an operational comparison, not a controlled experiment. Changes in channel, salesperson, price or offer can explain a difference. Check them before crediting the ICP.
Review the profile when evidence changes: a new offer, a recurring loss reason, delivery difficulties or a new segment that succeeds. Assign one person to keep it current, with input from sales, marketing, delivery and customer success.
No customer history yet? Start with an explicit hypothesis for one use case. Interview potential buyers and test a small set of accounts. Track what supports or contradicts each assumption as you learn. You can prioritize research before you have enough data to justify broad advertising.
Questions about ideal customer profiles
How many ICPs should a B2B company have?
Use separate profiles when the customer problem, delivery requirements, economics or buying process require different decisions. Different products may share a profile. A single product may need more than one. Split only when the distinction changes action.
Does an ICP work for B2C?
The account-level approach here is designed for B2B. For consumer offers, the unit may be a person or household. Use relevant customer research and segmentation rather than copying company fields into a consumer profile.
Can AI create an ICP?
AI can organize supplied evidence, summarize patterns and flag missing fields. Give it source references and require it to separate observations from assumptions. It cannot establish a customer’s budget, motives or satisfaction by inventing plausible answers. A generated persona is not a buyer interview.
Is a buying signal proof that an account is in market?
Treat it as a reason to investigate. An acquisition, job opening or funding announcement may concern a different team or project. Confirm the problem, relevant people and timing before treating the signal as an active opportunity.
Who owns the ICP?
Choose a named owner who can maintain the document and resolve inconsistent rules. Sales, marketing, delivery and customer success contribute different evidence. A profile that affects all four teams needs their input, even when one person maintains it.
Sources and scope
Sources checked on October 22, 2026. The template, two filled profiles, account decisions and deal calculation are our explanatory tools. The companies and deal data are hypothetical.
The primary sources linked above support the specific platform and method statements beside them. The deal comparison illustrates how to test a criterion. It does not establish that those criteria work for a real company or that an ICP alone improves results.
