Your ICP is too precise to reach: how narrow should your ICP be for cold email

Sunny Ray · Mar 3, 2026 · 5 min read

The tighter your filters, the more likely you are emailing the most spam-protected people on earth. A precise ICP feels smart in a spreadsheet. In a live campaign it starves you.

We had this exact conversation on a weekly call with a developer-tools startup selling to engineering teams. Their sweet spot was clear and correct: Series A startups hitting scaling problems and code complexity. Nobody on the call disputed that. The trouble was the list we built to reach them. As it was put on the call: "The audience has been quite small."

Small audiences are not automatically bad. Small audiences that do not answer are fatal.

The sweet spot was right, the list was wrong

There is a difference between who buys and who you can reach. Founders blur the two all the time.

The ideal customer for this company was real. A Series A team with a growing codebase and engineers drowning in complexity feels the pain. But the campaign tried to find those exact teams through filters, and the filters cut the market down to a sliver before a single email went out.

When the list is that thin, every other weakness gets magnified. One bad week of deliverability, one mailbox that needs warming, one subject line that misses, and you have nothing to learn from because there were never enough sends to begin with.

Technology filters: precise, tiny, unreliable

The main culprit was technology filtering in Apollo. On the call we said it plainly: the technology filters are quite limiting. You pick a stack, a tool, a framework, and the database returns the companies it has tagged with that technology.

Two problems. The tagging only catches what the data provider can detect, so you lose real buyers the database never classified. And stacking a technology filter on top of stage, headcount and role multiplies the shrinkage. Each filter sounds reasonable on its own. Together they leave you with a list you could email by hand in an afternoon.

Spam-filtered personas never reply

The second problem was who sat inside that sliver. Part of the target list was quants and other heavily protected technical people. These inboxes are built to say no. Aggressive filtering, assistants, and a learned allergy to anything that smells like sales.

You can write the best cold email of your life and it will not matter if it lands in a folder nobody opens. The narrower and more senior-technical your persona, the more of your list behaves this way.

So you end up with the worst of both: a small list, and a large share of it unreachable.

Go broader, on purpose

The decision on the call was to widen deliberately. Stop filtering by technology and niche segment. Target VP Engineering and CTO roles at bigger tech companies, where the pain of scaling code is just as real and the people are reachable at volume.

This is the part founders resist, because it feels like giving up on precision. It is not. You are moving precision from the filter to the message. A broad list with a sharp message beats a tiny list with a perfect filter, because the broad list gives the message enough chances to work.

Invite the engineers instead of pitching them

The other move was changing the ask for the most technical people. Strong engineers rarely take a sales call. On that same call the team noted that high-quality engineers were far more willing to join a podcast.

That matches everything we see. An engineer who would never book a demo will happily spend 25 minutes talking about how they solved a hard problem. The conversation about what they are struggling with now comes naturally after the recording stops. You did not pitch them. You gave them a stage, and they told you their pain unprompted.

The message changed with the list. The new campaign was positioned around codebase optimisation and streamlining, a problem every engineering leader at a growing company recognises, instead of a pitch tuned to one narrow stack. Broader list, broader pain, same product. The expectation on the call was higher reply rates from exactly that combination, and the plan included booking me as a guest on the client's own show, so the company could use the same invitation mechanic on its buyers.

Set a kill date before you start

The last decision was the one I would copy first. The broader campaign ran as a one-month trial, with an agreed plan to pull the plug if there was no meaningful traction.

A kill date protects you from two failures. It stops you abandoning a campaign after a quiet week, and it stops you dragging a dead one along for a quarter because nobody wants to call it. Write down what meaningful traction means before you hit send: replies, booked calls, a specific number. Then honour it.

How narrow is too narrow?

My rule of thumb: if you can describe your ICP precisely but cannot build a list of a few thousand reachable people from it, your ICP is a customer profile, not a targeting plan. Keep the profile for your messaging. Build the list one level up, at the role and company type that contains your buyer, and let the replies tell you who inside that group cares.

Then check who is actually inside the list. If a large chunk of it sits behind spam filters and assistants, change the offer for those people before you change anything else.

If your buyers are senior people who ignore sales email, an invitation to a show is often the only door that opens. See how a show gets you in front of enterprise buyers.

Frequently asked questions