You have a product that's ready (or close), zero paying customers, and a question nobody on the team can answer with confidence: who do we sell to first? The founder has three favorite segments, an investor is pushing a fourth, and your first sales hire wants to go after anyone with a budget. What usually follows is a prospect list that's way too broad, generic messaging, and three months later nobody knows whether the problem is the product, the segment, or the approach.
Without customers, there's no ICP in the classic sense, the kind you derive from analyzing your existing base. What you have is an ICP hypothesis. And a good hypothesis is one you can test and throw out fast.
Swap the "ICP" for explicit hypotheses
The first mistake is writing a final ICP on a slide. The second is leaving the hypothesis implicit, living in the founder's head. Ideally, you have two to four competing hypotheses, each written so it can be proven wrong.
A useful hypothesis has four parts:
- Segment: industry, company size (headcount or revenue range), and geography. Example: fintechs with 50 to 300 employees in Brazil.
- Trigger: what's happening at the company that makes the problem urgent right now. A recent funding round, expansion into another country, new regulation, a leadership change.
- Pain and current alternative: what problem you solve and how they handle it today (spreadsheets, a competitor, an in-house team, nothing).
- Buyer and user: who feels the pain, who signs the contract, and whether they're the same person.
If you can't fill in the trigger, be suspicious of the hypothesis. A segment without a trigger produces friendly conversations that never turn into pipeline.
How to choose which hypotheses to test first
You don't have the time or budget to test ten segments. Prioritize with three simple criteria, scoring each from 1 to 5:
- Pain intensity: does the problem cost money, create risk, or just annoy people?
- Reachability: can you build a list of at least 200 to 300 accounts and find the decision-makers on LinkedIn and in contact databases?
- Likely sales cycle: can a company like this buy in weeks, or does it need a buying committee, legal review, and an annual budget cycle?
Pick the two or three hypotheses with the highest total. Strategic segments with a 12-month cycle can wait: right now you need fast learning, not a shiny logo on your website.
Designing the test: outbound as an experiment
Outbound prospecting is one of the best labs for a new product, because you control who gets the message and what it says. But it only works if you treat it like an experiment, with isolated variables.
Minimum structure for each test
- Sample size: somewhere between 100 and 200 accounts per hypothesis, with two or three contacts per account. Anything less and every result is noise.
- Same cadence across all hypotheses: same channels (email, LinkedIn, phone, WhatsApp where it makes sense), same number of touches, same spacing. If the cadence changes, you can't tell what caused the difference.
- Hypothesis-specific messaging: this is where everything changes. The message needs to speak to that segment's trigger and pain. Generic messaging tests your copywriting, not the segment.
- Fixed window: four to six weeks, depending on volume.
What to measure beyond meetings
With a new product, booked meetings aren't the only signal. Also track:
- Reply rate, split into positive, negative, and "not the right time."
- Recurring objections, word for word.
- Who replies: the title you targeted, or someone who forwards you to another team?
- In the meeting: does the prospect recognize the pain without you having to explain it? Do they ask about pricing? Do they ask for a next step?
Pro tip: set up a "pain signals" spreadsheet and have whoever runs the calls and meetings log the exact phrase the prospect used to describe the problem. After 15 to 20 conversations, those phrases become the copy for your next cadences, and almost always a much better message than anything marketing wrote.
Kill criteria: when to cut or double down
Without criteria set in advance, every team finds a reason to keep its favorite segment alive. Agree on the numbers before you start. The ranges below are rough benchmarks for B2B tech and should be adjusted to your deal size and market:
- Cut: positive reply rate below 1% after the full cadence, dominant objections along the lines of "we don't have that problem," and meetings where the prospect doesn't recognize the pain.
- Adjust and retest: decent reply rate, but the wrong title is responding or the objection is about timing. Usually the segment is right and the trigger or the buyer is wrong.
- Double down: positive replies consistently above the average of your other tests, meetings where the prospect asks for a proposal or pilot, and objections concentrated on price or integration (a sign the pain is real).
Notice that the strongest criterion isn't volume, it's conversation quality. A segment with fewer meetings where half ask for a pilot is worth more than one with twice as many exploratory calls.
From test to your first real ICP
When a hypothesis clears the bar, go deeper: expand the sample, test variations in company size and geography within the segment, and track whether meetings move through the funnel. Your first five to ten customers from that group will give you what was missing, real data on sales cycle, deal size, and retention, and that's when your ICP stops being a hypothesis.
This process also works outside your home market. In engagements we've run at Nord, for example, tightly defined segments generated 137 meetings with decision-makers for a fintech in Spain and 32 meetings a month for a logistics company in the same market. In both cases, the gains came from a sharp focus, not bigger lists.
Next steps
This week, write two or three ICP hypotheses with segment, trigger, pain, and buyer. Score them, pick the best ones, define your kill criteria, and build a list of 100 to 200 accounts for each. In six weeks you'll have data, not opinions, to decide where to focus your team. If you don't have the bandwidth to run these tests with discipline, consider a dedicated pre-sales operation, in-house or outsourced, like Nord's.





