An ideal customer profile (ICP) describes the companies most likely to buy from you, stay and grow. The fastest way to build one is to study the deals you've already won. Pull your closed-won accounts, group them by what each company does and who it sells to, then compare how each group performs. The groups that close more often, faster and bigger are your ICP.
Most ICPs don't start that way. They start in a meeting where someone says "mid-market SaaS, 200 to 1,000 employees," everyone nods, and that sentence quietly becomes next quarter's targeting.
What an Ideal Customer Profile Is
An ideal customer profile is a description of the company, not the person, that gets the most value from your product and brings you the most value back. It covers what the company does, who it sells to, its size and region, the tools it runs and the moments that make it ready to buy.
A good ICP is specific enough to rule companies out. "B2B companies that want to grow" describes everyone. "Software companies that sell to regulated industries, have a sales team of at least ten and run a modern CRM" describes a market you can count, rank and work.
What's the Difference Between an ICP and a Buyer Persona?
An ICP describes a company and a buyer persona describes a person inside it. Your ICP tells you which accounts to pursue; your personas tell you who to talk to once you're there and what they care about. You need both, in that order, because the perfect persona at the wrong company still won't buy.
Why Most ICPs Miss
Most ICPs miss because they describe the customer a team hopes to have, not the customers who actually signed. They're built from opinion, written in firmographic shorthand and never checked against results.
Three problems show up again and again:
- Lumped businesses. Industry codes and headcount bands lump unrelated businesses together, so a "software, 201 to 500 employees" ICP covers companies with nothing in common.
- Averages that hide winners. The ICP is built from averages, which hides the segments where you actually win.
- No check against losses. Nobody tests the ICP against lost deals, so traits that sound important but don't separate wins from losses survive for years.
The cost is real. CSO Insights research on documented ICPs, reported in 2011, found that companies with a defined, documented ideal customer profile won 51% of their forecast deals, against 42% for companies without one. And getting the market itself wrong is worse: when CB Insights reviewed 111 startup post-mortems, 35% of founders named no market need as a reason they failed, according to CB Insights' analysis of startup post-mortems.
Start With Your Closed Deals
Pull every closed deal from the last 18 to 24 months: won, lost and open. Won deals show you who buys. Lost deals show you which traits don't matter. Open deals show you whether the pattern is still holding today.
Go back far enough to get a meaningful sample, but not so far that the deals describe a product you no longer sell. If you changed pricing or launched a major product in that window, weight the newer deals more heavily. Tag the wins that churned within a year as their own group, because a deal that closed and left is closer to a loss than a win for ICP purposes.
What Can Closed-Lost Deals Tell You About Your ICP?
More than most teams expect. Lost deals show which traits don't predict a win, so they keep your ICP from filling up with assumptions. Look for three patterns:
- Traits shared by wins and losses. If a trait appears just as often in both, it isn't part of your ICP, however sensible it sounds.
- Segments you keep losing. A segment with plenty of opportunities and a low win rate usually belongs in your disqualifiers, not your target market.
- Losses that came back. Accounts that were lost on timing, then reopened, often fit perfectly. They belong in your ICP with a timing note attached.
Ocean.io learns from lost deals as well as wins, so traits that show up on both sides stop pushing scores up.
Group Wins by What Companies Do
Group your won deals by what each company actually sells, who it sells to and how it makes money. This is the step that separates a useful ICP from a generic one.
Industry codes won't do this for you. Two companies can share a code and run completely different businesses: one sells payroll software to restaurants, the other sells data infrastructure to banks. Grouped by code, they look identical. Grouped by what they do, they're obviously different customers who buy for different reasons. That's the idea behind B2B customer segmentation beyond firmographics.
Doing this by hand means reading each customer's website and tagging it, which works for fifty deals and breaks at five hundred. Ocean.io reads what each company actually does and who it sells to, then maps every company you've sold to into segments automatically. Its Discovery view shows where your win rate and revenue cluster, so the groups come from your data rather than from someone's memory of last year's best deals.
Compare Segments, Not Averages
Once your wins are grouped, compare the groups on the numbers that matter: win rate, average deal size, sales cycle and retention. The segments that win more often, faster, bigger and stay longer are the core of your ICP.
| Segment (illustrative) | Win rate | Average deal | Sales cycle | Retained after 12 months |
|---|---|---|---|---|
| Software sold to banks and lenders | 31% | $38,000 | 62 days | 94% |
| Payments and fintech platforms | 24% | $27,000 | 71 days | 90% |
| Healthcare software | 18% | $24,000 | 96 days | 85% |
| Developer tools | 9% | $14,000 | 48 days | 71% |
An average across all four would hide everything this table shows. The blended win rate might look like 20%, which describes none of the segments. The first two are your ICP. The third is a maybe. The fourth closes fast but churns, which is exactly the kind of segment that inflates pipeline without building revenue.
Then check the edges against your lost deals. If a trait shows up just as often in losses as in wins, it isn't separating anything, however sensible it sounds. Keep only the traits that clearly push win rate up.
How Many Closed-Won Deals Do You Need for an ICP?
There's no fixed number, but patterns get easier to trust once a segment has a few dozen wins behind it. With fewer, treat the ICP as a hypothesis: write it down, test it against lost deals and your open pipeline, and update it as new deals close. A small, honest ICP beats a large, made-up one.
Write Your Ideal Customer Profile Down
A written ICP is one the whole revenue team can use, challenge and improve. Keep it to one page, organized around the traits that actually separated wins from losses.
| Field | What to write | Example (illustrative) |
|---|---|---|
| What the company does | Business model and what it sells | Software sold to regulated financial institutions |
| Who it sells to | Its own customers | Banks, credit unions and lenders |
| Size | The range where you've won | 50 to 1,000 employees |
| Region | Where you can sell and support | US and Canada |
| Tools it runs | Systems your product needs | A modern CRM and a cloud data warehouse |
| Team signals | Signs it can buy and use you | Sales team of ten or more, growing |
| Timing moments | When it's ready to talk | New funding, new sales leader, sales hiring |
| Disqualifiers | What rules a company out | Agencies, outsourced sales, no sales team |
| Buying committee | Roles that show up in won deals | VP Sales, RevOps lead, CFO |
The disqualifiers row matters as much as the rest. Writing down who you don't sell to saves more rep time than any single targeting trait, and it's usually the row teams skip.
The buying committee row can come from evidence too. Ocean.io reads the job titles on your won and lost deals, learns which roles show up when you win, then finds people in those roles at each company it recommends.
What Does a Negative ICP Look Like?
A negative ICP describes the companies you should never chase, and it saves more rep time than any single targeting trait. Build it from the same evidence as the ICP itself: segments with low win rates, accounts that churned fast and companies that bought for the wrong reason.
Typical negative ICP traits include:
- Wrong business model. Agencies or resellers when you sell to end users, or the reverse.
- No team to use the product. For example, no sales team when your product exists to help one.
- Fast churners. Segments that close quickly but leave within a year.
- Long, low-value cycles. Segments that take twice as long to close for half the deal size.
In Ocean.io, keyword exclusions remove any company whose site or description matches your disqualifiers, so negative-ICP companies never reach your reps in the first place.
What Do ICP Examples Look Like by Industry?
Here's how an ICP reads in four industries where B2B targeting is especially hard. The details are illustrative, but the shape is what a strong ICP looks like: specific about the business, the buyer and the disqualifiers.
Software Sold to Financial Institutions
Companies that sell compliance, risk or data software to banks, credit unions and lenders. 50 to 1,000 employees, a sales team of ten or more, and a security or compliance certification already in place. Disqualify consultancies and companies that only resell other vendors' products.
Payments and Fintech Platforms
Companies that move money for other businesses: payment processing, payouts, expense management or embedded finance. Growing sales teams, recent funding and a product sold to finance leaders. Disqualify consumer-only apps.
Logistics and Transport
Companies that run their own fleets, warehouses or freight operations, or sell software to companies that do. Regional and national operators with a dedicated operations team. Disqualify single-location operators and freight brokers when you sell to asset owners.
Recruitment and Staffing
Agencies and platforms that place candidates for other businesses, with a business development team that sells to employers. Disqualify in-house recruiting teams when your product is built for agencies.
Put Your ICP to Work
An ICP only pays off when it changes what your team does. Use it in four places: sizing the market, scoring accounts, routing inbound and building territories.
Start by sizing the market your ICP describes. Count the companies that match it with hard filters and keyword exclusions, and you'll know how many accounts each segment holds and what they're worth. In Ocean.io, filters give you an exact company count, and recommendations project what each segment is worth at your own opportunity and win rates.
Then score every account against the ICP and give each one a tier. Ocean.io's recommendations come with the reasons matched, the segment, a score and a tier from 1 to 4, so a rep can see why an account made the cut.
Next, route inbound by fit. Ocean.io scores each new lead against the segments that win, shows the fit score and the reason behind it, and writes segment scores back to your CRM, so the best leads reach a rep first. Finally, carve territories by ICP value rather than by account count, and push the ranked accounts into Salesforce or HubSpot so reps start from the same picture.
How Often Should You Update Your ICP?
Review it every quarter and rebuild it at least once a year. Rebuild sooner after a pricing change, a major product launch or a move into a new market, because each of those changes who wins. The strongest ICPs never stop updating: every new deal, won or lost, is more evidence. Ocean.io keeps learning from your own deals, so the segments and scores move as your market does.
What to Do Next
- Build your ICP from won, lost and open deals, not from a workshop.
- Group wins by what companies actually do, and compare segments instead of averages.
- Put the ICP to work in sizing, scoring, routing and territories, and keep it learning.
See the ICP hiding in your own closed deals: book a demo.
Frequently Asked Questions
What is an ICP in sales?
In sales, an ICP, or ideal customer profile, describes the type of company most likely to buy, stay and grow with you. Sales teams use it to decide which accounts to prospect, how to prioritize them and which deals to walk away from early.
What should an ideal customer profile include?
It should include what the company does, who it sells to, its size and region, the tools it runs, the signals that show it can buy, the moments that make it ready and clear disqualifiers. Every trait should be one that separates your won deals from your lost ones.
Can a company have more than one ICP?
Yes. Many B2B companies have two or three ICP segments that buy for different reasons. Treat each one as its own profile with its own message and play, rather than blending them into one average description that fits none of them well.
What if we don't have many closed-won deals yet?
Start with the wins you have, add your most engaged lost deals and your best open opportunities, and treat the result as a hypothesis. Write it down, test it in targeting for a quarter and update it as new deals close.


