Most founders validate backwards. They build the thing first, launch it, then go looking for people who care. That is a brilliant way to spend six months and a chunk of savings learning something a 30-minute conversation would have told you for free.

There is a faster order of operations. Before you write a single line of code, you find 50 people who fit your customer profile, send them a short cold email, and get them on a call. If real buyers will give you 20 minutes to talk about a problem they actually have, you have a signal. If nobody bites, you just saved yourself the build.

Here is the full playbook, using Claude and Apollo end to end.

Step 1: Define your ICP, then build a list of 50

Before you touch a contact database, get crystal clear on who you are actually trying to reach. "Founders" is not an ICP. "Heads of ops at US logistics companies with 50-200 staff who still run dispatch on spreadsheets" is an ICP.

Use Claude to pin it down. Paste this in:

I'm validating a startup idea: [one-line description of the idea and the problem it solves]. Help me define a precise Ideal Customer Profile for cold outreach. Give me: job titles, company size band, industry/vertical, geography, and 3-4 observable signals that tell me this person feels the problem acutely. Keep it specific enough that I could filter for it in a contact database.

What good output looks like: a shortlist of 2-3 exact job titles (not ten), a clear company-size band, a region, and a few "trigger" signals you can actually search for, such as a recent funding round, a specific tool in their stack, or a hiring spree. If Claude hands you something vague, push back with "narrow this to the single segment most likely to pay for a fix today."

Now build the list in Apollo. Apollo gives you access to a 210M+ contact database, and you can filter on job title, company headcount, industry, location, and a stack of other attributes. Punch in the ICP Claude defined, aim for 50 contacts, and prioritise getting a verified work email for each one.

Verify before you send. A list with dead addresses tanks your deliverability before your copy gets a fair shot. Apollo flags verified emails, so lean on the verified ones and bin anything marked risky.

Done looks like: 50 named humans who genuinely match the profile, each with a verified work email, in a simple sheet or Notion table. No "close enough" entries.

Step 2: Write a 5-line cold email ending in a one-word question

The job of this email is not to sell. It is to start a conversation. The single biggest mistake here is pitching your idea in the email itself, because then people react to your solution instead of telling you about their problem.

Keep it to five lines, plain text, no images, no links, no tracking pixels. End with one question they can answer in a single word, because a one-word reply has almost no friction, and a reply is all you need to open the door.

Use Claude to draft it. Try this prompt:

Write a 5-line cold email to [ICP job title] at [type of company]. Goal: start a conversation about [the problem], not pitch a product. Rules: plain text, no links, conversational, no buzzwords, opens with something specific to their role, and ends with a single question they can answer in one word (like "yes/no" or "weekly/monthly"). Give me 3 variations.

What a good one reads like:

Hi [Name],

Quick one. I keep hearing that [specific role] teams are still [doing the painful thing] by hand.

I'm digging into whether that is actually as annoying as people say, or whether it's a non-issue.

You'd know better than me.

Is [the painful thing] something your team deals with weekly, monthly, or basically never?

Notice it asks for a one-word answer, makes no demands, and never mentions a product. The subject line should be short and human (2-4 word subject lines hit the best open rates), something like "quick question" or "[their company] + [their pain]".

A few practical notes: