Lead Research and ICP: The First Build Phase of Any GTM Team

Lead research almost never fails because of the tools. It fails because nobody defined who you are actually looking for.

As a software entrepreneur, I build GTM teams. In every GTM team I build, the first step is the same. Not buying a tool. Not pulling a list. Defining the ideal customer profile. Who do we want, and how do we recognise them from the outside?

That is phase one. In the loop I teach in the framework on gtm.science it is called DECODE. Everything else comes after.

What you will take away

• Why the ICP comes before any lead research.
• How I define the ICP with a team in three steps.
• How the ICP turns into signals that tools can find.
• Why marketing and sales have to meet here.

The thesis: ICP first, list second

Quality over quantity. The line is in every sales book. Almost nobody does it. Most teams start with the question which database do we use? The right first question is who are we looking for and why are they worth it?

Without that answer, every list is a guessing game. With it, lead research becomes precision work.

Lead research without an ICP is collecting data with no direction.

🧨 The problem: prospecting without a definition

Prospecting, the act of finding potential customers, is a billion-dollar industry. Companies hire entire lead research teams. There are dozens of tools, many of them excellent. And still, for many, the pipeline dries up.

The reason: prospecting often sits only in sales. That is too late. The search for the right leads starts in marketing, with a shared definition. Companies with coordinated marketing and sales grow faster. The number that has circulated for years: 27 percent faster revenue growth when both teams work in sync. Defining the ICP is where that alignment begins.

Without that alignment you get what I see in many companies: marketing attracts leads sales would never touch. Sales researches accounts that will never benefit from your offer. Both work hard, and the pipeline stays thin anyway. The shared ICP definition isn't a workshop ritual. It's the working basis both teams can point to when priorities are debated.

🛠️ How I build the ICP with a team

I do this with every team in three steps. Not a workshop over weeks. One focused block.

• 1. Set the firmographics. Industry, size, region, maturity. Not who could buy, but who already buys and why. We look at the ten best existing customers and find the pattern.
• 2. Sharpen the buyer persona. Who sits at the table? Decision maker, champion, blocker. In B2B a group buys, not an individual. Each role has its own questions.
• 3. Derive the signals. How do I recognise the ICP from the outside, without asking? Which technologies do they run? Which roles are they hiring right now? Which triggers show that now is the moment?

Only after these three steps do we talk about tools. Not before.

In step one, the sorting matters: your best existing customers aren't the ones with the best-known logo or the biggest deal. They're the ones who stay, use the product and get results. That view protects you from building your ICP on prestige instead of substance. It's also why the definition starts with existing customers, not dream customers.

The 4 rules for an ICP that holds up

For the sharpening I like the discipline Mark Roberge, founding CRO of HubSpot, describes in The Science of Scaling. His ICP framework replaces gut feeling with checkable criteria. It comes from the US SaaS context, but it travels well because it doesn't assume market size, it assumes honesty.

Rule 1: every criterion must be visible from the outside, from publicly available data. Employee count, region, industry. Not budget, not urgency, not strategic priorities. What you can't check from the outside, your research can't find and your AI agent can't watch.

Rule 2: the ICP is based on retention and customer success, not on the easiest closes. Retention means how many customers stay. The question isn't who buys fast. The question is who stays and gets results. A segment with a high close rate and high churn is a trap.

Rule 3: the market behind the ICP should, per Roberge, cover roughly 3 years of customer and revenue targets. Not eternity. An ICP that can never become too small never was one.

Rule 4: the ICP stays honest through a visible change log and a clear three-way split. One segment you pursue proactively. One you serve only when it knocks on its own. And one you deliberately don't sell to. The third is the most important. It protects your team from deals that support pays for later.

Small enough to lead

Stijn Hendrikse adds three rules in T2D3 that point the same way. The ICP comes from your best existing customers, not from a market study. You pick the smallest market you can lead within a year. And you work with 3 personas at most.

The middle one is the most uncomfortable. Leading means: when your topic comes up in that segment, people know your name. You won't get there in a market like software for the mid-market. You will get there in a tightly cut niche you can penetrate in months rather than decades. Winning small beats swimming along big.

And the persona limit forces a decision many teams avoid: who do you address first? More personas doesn't mean more market. It means diluted messages and research that searches in every direction at once.

🤖 The tools, kept short

Once the ICP stands, the tools are almost an afterthought. LinkedIn and the Sales Navigator are the most current contact database for B2B, maintained by the leads themselves. Data providers like Apollo, Zoominfo or Clearbit deliver contact data at scale, at roughly 80 percent accuracy, strongest for large companies in North America. IP tracking shows you which companies visit your website. Platforms like Clay combine several sources.

But every tool is only as good as the question you ask it. The question comes from the ICP. A tool without a definition finds you more leads, not better ones.

A practical side effect of Roberge's first rule: criteria from public data are exactly the filters these tools are good at. Employee count, industry, region, technology in use. No tool in the world can model an ICP made of gut feeling.

The strongest argument: signals over volume

Here is the part that changes everything in 2026. Lead research used to mean enrich as many contacts as possible, then dial through them. Today it means react to signals.

A cleanly defined ICP turns into a list of signals that AI agents can watch around the clock. A new funding round. A leadership change. A newly adopted technology. An open role that hints at growth. I call that signal-based selling.

That is exactly the point of a context engine, the business context of your company prepared as a foundation for AI agents. Without that foundation the rule is: garbage in, garbage out. With it, a hyperlean team does the work of thirty people. A few pros plus AI agents. Three instead of thirty. That is Get Multiplayer.

How do you tell an account is ready to buy right now?

Never from a single signal. One signal is a guess, several are a pattern. Only when leadership changes, hiring patterns and engagement with your content stack on top of each other does a fitting account become this week's priority.

A new sales director alone means little. A new sales director at a company that is posting several sales roles and whose team engages with your posts is a different picture. The art isn't finding a signal. The art is stacking signals and only then acting.

Your own signals beat the purchased ones. Whoever reads your content or reacts to your posts has raised their hand. Which of this data belongs in your CRM is covered in the article on customer data in B2B sales.

And the ICP itself isn't a document, it's a hypothesis. With every team I check it once a quarter against the new customers: who did we win, who stays, who really uses the product? Your best customers shift faster than you think. How that becomes a running process is in the hub on signal-based selling.

🎢 Highs, lows, warning

✅ What works. An ICP based on real existing customers that translates into observable signals.

❌ What does not work. Starting with tool selection before the definition exists.

⚠️ Warning. This is never about aggressive or unethical methods. It is about relevance and real value. A precise ICP means disturbing fewer people, not more.

Lead research almost never fails because of the tools. It fails because of the missing definition. That is why the ICP is the first build phase I do with every team. Get this right and you stop feeding contact corpses into the CRM and start building a pipeline that compounds. Founder to founder: define the ICP before you pull the first list.