When a campaign isn't working, the founder's first instinct is to rewrite the copy. The team rewrites the subject line, then the opening line, then the body. The reply rate doesn't move. So the copy gets rewritten again. Three rounds in, everyone is exhausted and the calendar is still empty.
Nine times out of ten, when we step into a stuck campaign, the copy is fine. The ICP is wrong. Or — more often — the ICP is correct, but it's been operationalised in a way that points the campaign at people who can't or won't buy.
Below are the four ICP mistakes we see most often. We run this exact diagnostic before we write a single word of copy on a new engagement.
Mistake one: confusing buyer demographics with buyer triggers
An ICP statement that reads "Heads of Marketing at B2B SaaS companies, 50–500 employees, headquartered in the UK or US" is a description, not a strategy. It tells you who the buyer might be, but not why they would want what you sell right now.
The version of that ICP statement that actually drives campaigns reads something like: "Heads of Marketing at B2B SaaS companies who have just hired a Demand Generation lead in the last sixty days." That's a trigger — something you can detect in public data, something that creates urgency, something that maps to a specific moment in the buyer's life.
Demographic ICPs produce reply rates around 0.5–1.5%. Trigger-based ICPs, on the same offer with the same copy, produce 3–5%. The compounding effect over a quarter is enormous.
Mistake two: targeting the user, not the buyer
If you sell a tool that an analyst uses every day, the analyst is the user. The buyer is usually their boss — and the budget often sits two levels above that.
Founders make this mistake constantly. They write to the persona who would love the product, not the persona who would approve the cheque. The result is a flood of "yes, this looks great, let me share with my manager" replies that never convert. Or, worse, no replies at all, because the user doesn't have time to read cold emails about tools they're not buying.
Map the actual purchase decision before you target. Who has the budget line? Who has to sign off on changes to the existing stack? Who feels the consequences if the tool doesn't work? Often, that person is one or two layers above the user — and writes very different emails back.
Mistake three: assuming "growing companies" is a market
"We sell to high-growth startups" is one of the most common phrases we hear on a discovery call. It is also one of the least useful. "High-growth" is a description applied retroactively. You can't target a high-growth company; you can only target a company that has just done something that indicates growth.
The translatable signals are things like: just raised a Series A or B, just expanded to a new geography, just hired a senior leader, just released a new pricing page, just won a notable customer, just appeared on a hiring spree across 8+ roles. Those are observable and queryable. "High-growth" is not.
If you can't write down the specific public signal that says "this company has the problem you solve, right now", you don't have a targetable ICP yet — and your campaign will perform like your account list is randomly generated, because that's effectively what it is.
Mistake four: a TAM that's too small to be measurable
This one is more controversial. We frequently turn down engagements where the TAM is under 5,000 accounts. Founders push back: "But it's a high-quality 5,000! We'd be happy with even ten of them." The maths doesn't work that way.
Cold email reply rates plateau at 3–5% on a well-built campaign. Of those replies, perhaps a third are positive. Of those positives, around half become real meetings. Of the meetings, your win rate is whatever it is — call it 20% on a strong offer.
Run the funnel: 5,000 contacts × 4% reply × 33% positive × 50% meeting × 20% close gives you 6.6 customers across the entire TAM, before any account is contacted twice. If your business needs more than seven new customers a year, the TAM is too small for cold outbound to be your lead engine, regardless of how high-quality the contacts are.
This is also the most painful conversation we have on discovery calls. We've turned down founders we liked, with offers we believed in, because we couldn't see how cold email maths added up to enough meetings. In every case, retainer-model agencies will happily take their money. It's just that, after twelve months, those founders end up coming back to us asking why their pipeline didn't fill.
The four-question diagnostic
If your outbound is underperforming, run these four questions before you touch the copy:
- What public, observable trigger event puts a company on the target list? If you can't write it in one sentence, the ICP is too vague.
- Who in the company has the budget? Title and seniority, not "the C-suite".
- What specific tool, process or vendor are they likely using today, and why is it failing them right now? If you can't name what they're switching from, it's hard to write a compelling reason to switch.
- Is the addressable market large enough that you can lose 95%+ of it and still hit your number? Cold email is a low-yield channel by design. It works because the mathematics scale, not because every contact converts.
If three of those four answers are crisp, almost any reasonable copy will work. If two or fewer are crisp, no copywriting trick will save you — and you will spend a quarter rewriting subject lines while the real problem sits upstream.
The copy is the last 10% of an outbound campaign. The ICP is the first 90%. Most teams reverse the proportions.
Where to go from here
If you're stuck and you'd like a second pair of eyes on your ICP, drop us an email or book a discovery call. We'll run the four-question diagnostic with you live. If your ICP holds up, we'll tell you where to go from there. If it doesn't, we'll tell you that too — even if it means we don't take the engagement on.
The full ICP framework, including the trigger-event taxonomy and the LIA we use under UK GDPR, is in chapter two of the Cold Email Playbook 2026.