ICP sales works when you start with the company

In sales, ICP stands for ideal customer profile. Gong, Salesforce and Revenue Grid all use it the same way: a description of the kind of company that buys, gets value and stays. In a wider business sense it's a filter for where your team spends its hours. The unit is what matters. An ICP is a company.

A buyer persona is a different tool. It describes the individual role inside that company, with a job title, goals and pain points. You need both, but the order matters. According to Prospeo's guide, the ICP should be reviewed quarterly, while personas need a review twice a year. The company profile is the one that moves first when your market does.

Here's why the claim deserves your attention. Prospeo reports that companies with a strong ICP see 68% higher win rates and are 67% more likely to exceed quota. That's one vendor's blog claim, so read it as a direction, not a promise. The logic still holds. If the company is a bad fit, the best-written persona won't close the deal.

Take a worked case. Say you sell heatmaps to B2B SaaS. A US team and a team in Bengaluru can write the same company line: B2B SaaS, 50 to 500 employees, already running Google Analytics. The persona comes after, and it might be a head of growth at one company and a founder at another. The company line decides who gets a call. The persona decides what you say on it.

It doesn't apply everywhere. If every deal is a one-off that a founder closes personally, a formal ICP adds little. You're better off writing down why your last few customers said yes.

Do this today: write your ICP in one sentence naming an industry, a size band and one tool the company already uses. Keep the persona on a separate page.

Salesforce lists the building blocks as firmographics, technographics and behavioural patterns. Firmographics are the facts on a company profile: industry, revenue, employee count. Technographics are the tools the company already runs. Behaviour is what the company does around you. Most write-ups spend their words on the first two.

Technographics sharpen a profile faster than you'd expect. For a heatmap seller, the research brief suggests companies that use GA4 but haven't put a CRO stack in place. A company with analytics and no testing tools has the data and the gap. A company with a full CRO stack already has an answer, even if it's a rival's.

Now the part that earns its keep. Prospeo calls the negative ICP a high-impact upgrade: a written list of companies you won't sell to. Suppose you sell churn-reduction software. The brief's profile is recurring-revenue businesses with 100 to 1,000 customers, $10K to $100K in monthly recurring revenue and churn above 5%. The negative ICP is simple: a company with no subscription model has no churn to reduce, however large it is.

Why does it matter? Prospeo cites an average B2B SaaS conversion rate of about 1.2% and says that filtering out bad-fit prospects can move that number. The source doesn't say by how much. So the honest claim is that it helps, and you'll learn how much by measuring your own deals.

The trade-off is real. A negative ICP built from a handful of customers can exclude a company that would have bought. If you have few customers, add the deals you lost and write down why. Lost deals are a second dataset, and they're often more honest than the wins.

Decide this week which kinds of company you'll stop calling, write them next to your ICP sentence, and tell the team.

Third-party lists tell you who fits on paper. Your own site tells you who acted. That difference is the gap in most ICP advice, and you can close it without buying a list.

Three things are worth reading from your own traffic, and all three are about companies. Which companies came back. Which pages they read, for example pricing, comparison pages and docs. And in what order they read them. A blog post alone isn't a buying signal. A return to pricing after docs looks much more like one.

Take two companies, both inside your ICP. The first reads a blog post and leaves. The second visits pricing, then docs, then pricing again. This is an illustration, not a customer story, but the order is the signal. A company that goes from pricing to docs is checking whether it can use the product. A second look at pricing reads like a cost check before a decision. The second company is further along, and it goes to the top of your list.

Here's a method you can run this week.

1. List the companies that reached your site last month.
2. Mark the ones that match your ICP sentence.
3. Of those, mark any that read pricing, a comparison page or docs.
4. Note which came back, and read the order of their pages.
5. Rank them by priority, with matching, returning companies first.

That's a priority order, not a temperature label. A company that matches and returns goes first. One that matches and read only a blog post goes next. One that doesn't match waits.

You need a tool that resolves the company behind a visit in the first place. PageFox identifies the company behind a visit where available. Where a visit can't be resolved, treat it as unknown and don't guess.

When this doesn't work: low traffic. If your site is small, the same company rarely returns, and the order of pages means little. Start with firmographic and technographic signals from an enrichment tool, and come back to behaviour as traffic grows.

One caution on tracking. Privacy requirements vary by jurisdiction and your consent setup, so use your own legal review before you read traffic by company.

Marketing builds its lists from the same profile that sales calls from. When the two teams hold different definitions, marketing hands over leads that sales ignores. If you're the one doing the selling, write one profile and one negative ICP, and have both teams point at it.

Don't assume a profile built in one market fits another. A 50 to 500 employee band that works for a US SaaS seller may fit Indian companies differently, and only your own closed deals will tell you. The sources behind this article don't split their advice by country. None gives India-specific benchmarks for company data coverage, privacy rules or team size.

If you quote deal sizes or budgets, write both currencies. Keep the dollar figure first. Use the day's rate, which was 1 USD = ₹96 on 1 October 2026.

Check your data before you trust it. Lists and enrichment tools may cover one market better than another, and no source here says which. Test a small sample of companies from your target market before you build a quarter of outreach on any list.

If you read site traffic by company to find accounts, settle your consent setup with a lawyer who knows the market you sell into. This article isn't legal advice.

When this doesn't apply: if you sell only in one country, the split doesn't matter, and one currency is enough. Then spend the saved effort on the review habit below.

Here's the strongest case against all of this. A tight ICP can make you miss a company that would have bought. That's the trade-off, and it's real. A profile built from your past customers describes who bought before, not who might buy next.

Prospeo gives a rule for it. If under 30% of your wins are referral-driven, lean on psychographics, meaning the attitudes and priorities of the buyer, rather than strict firmographics. Put simply, when your customers don't cluster by size, don't force them into a size band. If your pipeline is mostly referrals, or you're early and still learning who buys, keep the profile loose and review it more often.

There's an open question the brief leaves too: when data is thin, which field comes first, tech stack or revenue? Start with the tech stack if your product depends on one. It's easy to check from the outside, whereas revenue for a private company is often a guess.

Review the whole profile quarterly, as Prospeo advises. A profile that was right in January can be wrong by October if your product or the market moved. Bring last quarter's closed deals, the lost ones too, and the list of companies that read your pricing page. If the companies reading pricing don't match your ICP, the profile is the thing to change.

Put a date in the calendar for the next review, one quarter from today.

PageFox is the wrong choice for a site with little traffic: it tells you which companies are already visiting, it does not bring new visitors.

Write your ICP as one sentence about a company, with a negative ICP beside it, before you buy any list. Then check it against the companies that reached your pricing page last month.