Six leads from 10,000 emails: how to A/B test cold email positioning when you don't know who you are

Sunny Ray · Mar 2, 2026 · 5 min read

Six leads out of 10,000 emails. That was the number on the screen when we sat down with a startup selling into construction. Roughly a 0.1% response rate on their first VC campaign.

The instinct in that moment is to rewrite the copy. New subject line, shorter opener, test version A against version B. I think that instinct is wrong more often than it is right, and this call is the cleanest example I have of why.

If you cannot say whether you are a design company, a construction company or a consumer app, an A/B test will not tell you. Three separate campaigns to three separate investor types will.

Positioning is a hypothesis, so test it like one

Ten minutes into the call it became clear the problem was not the email. The founders and Charles were disagreeing, in good faith, about what the company even was. AI design company? AI construction company? A consumer AI app? Each answer is defensible. Each one points at a different investor.

An AI fund wants to hear about the model and the data advantage. A construction tech investor wants to hear about job sites, contractors and how fast the product gets adopted on a crew. A consumer investor wants to hear about the homeowner and the viral loop. One email cannot speak to all three without sounding like it speaks to none of them.

So we stopped arguing and wrote it down as a test: three personas, three campaigns, three lists. Everyone on the call agreed in about a minute, which tells you how much time the argument had been wasting.

Why split campaigns beat a single A/B test

A classic A/B test holds the audience constant and changes the words. That works when you know who you are talking to and want to find the better sentence.

When the identity of the company is the open question, the audience is the variable. Changing the subject line while sending to the same mixed list of VCs is like testing two shades of paint on a house you have not decided to buy.

The split looks like this. Campaign one positions the company as AI design and goes only to AI-focused VCs. Campaign two positions it as construction tech and goes only to construction tech investors. Campaign three positions it as a consumer AI app and goes only to consumer-focused investors. Same founders, same traction, three stories, three lists. Whichever list replies tells you which story the market believes.

That is the real output of the test. You are not learning which email is better. You are learning which company you are, according to the people with the money.

The audience was half the problem

There was a second lesson in the six leads. The first campaign had leaned on an old-school audience, and as someone put it on the call: "The old school guys are hard to sell online."

The team expected tech-savvy audiences, VCs and angels, to respond three to six times better than traditional contractors. That is not an insult to contractors. They buy through relationships, trade shows and people they already know. A cold email from a startup they have never heard of is the weakest possible door into that world.

So part of the fix was simply pointing the machine at people who live in their inbox. An angel campaign went live alongside the VC campaign for the same reason.

Volume matters too

A 0.1% rate on 10,000 sends is six conversations. You cannot learn much from six. If you split that same volume three ways, each arm gets around 3,300 sends, and a small difference in reply rate becomes noise.

So we did the boring thing. Sixty additional mailboxes went up to double sending capacity. Positioning tests need enough sends per arm for a winner to be obvious, not arguable. If your test needs a statistician to read it, the arms are too small.

What the customer calls told us that the deck didn't

The most useful moment on the call had nothing to do with email. I came out of it more bullish on the company than I went in, and the reason was the founders' customer calls. When they described the outcome a customer actually gets, the before and after on a real project, the company stopped sounding like a category and started sounding like a business.

That is your tiebreaker when the test results are close. Which story do your customers tell when they describe you to someone else? Lead with that one, and point it at the investors who already fund that story.

How to run this yourself

Write the three or four identities your company could plausibly claim. For each one, name the investor type who funds that category and build a list of only those people. Write a separate campaign for each, with the same traction facts and a different first sentence about what you are. Give each arm enough volume to produce a clear winner. Then read the replies, not just the count. The questions investors ask tell you which story they heard.

If one arm wins clearly, that is your positioning for the raise. If none of them wins, your problem is not positioning. It is the offer, and that is a different conversation.

And if cold pitching is the part that keeps returning 0.1%, change the ask itself. An investor who ignores a pitch will often say yes to being a guest. See how a show gets you in front of investors.

Frequently asked questions