One data point beats twenty

Most cold emails that claim to be personalised are personalised on the same five facts. A funding round. A new hire. Someone changing jobs. A tech stack. Headcount. Every tool sells those fields, so every vendor in your market uses them, often in the same week.

The result is predictable. Your prospect gets a message about their new funding, then another, then another. None of those messages is wrong. None of them is worth a reply.

The campaigns that work for us start somewhere else. We look for one data point that is specific to the market we are selling into, that the prospect can check for themselves, and that ties directly to what our client sells.

One of our clients is a Dutch SEO agency that sells to online shops. The standard pitch in that market is some variation of "we'll improve your Google rankings". Every shop owner has heard it dozens of times.

Instead, we looked for online shops without an HTML sitemap in their footer. It is a small technical gap. The owner can spot it in ten seconds by scrolling to the bottom of their own homepage. And it ties to work an SEO agency does and can measure.

Every email opened with that single observation and asked one question. The campaign achieved a reply rate of 8.79% and generated 33 qualified opportunities.

The sitemap was not a brilliant insight. It worked because it was true, verifiable and relevant all at once.

We give every possible signal a score from 1 to 10 before we build anything.

• 8 to 10. A yes-or-no gap that the prospect can check for themselves and that ties directly to the offer. This is sitemap level.
• 5 to 7. A strong indication: a pattern in job ads, an adjacent tool, or an announced expansion without the capacity to deliver it.
• 3 to 4. The right profile, but no behaviour. The company fits, but nothing says why now.
• 1 to 2. Basic fit only.

We only write a specific email from 7 upwards. Below that, we send a well-written generic message and don't pretend to know more than we do.

We score honestly, because the score sets the tone of the email. An inflated score produces an email that claims more than the data can back up, and prospects notice. An honest 4 beats a false 8.

Rarely from a sales database. In our projects, they came from:

• The prospect's own website, read the way a customer would read it. A page that is missing, or a team page with nobody in a region where the company claims to be active.
• Trade association member lists. For a market research fieldwork provider, we built the target list from the member lists of healthcare research associations, not from a contact database.
• Public registers. Licences, company registers, tenders.
• Job ads, which often describe a problem more honestly than the company's website does.

One rule helps here: say why a source reveals the signal. "Have an AI agent read their website" is a choice of tool, not a choice of source. The real choice is which page, and what you expect to find there.

1. Can the prospect check it in under a minute? If not, it reads as a guess.
2. Does it tie to what you sell? A hiring signal is noise, unless your offer solves the problem that job ad points to.
3. Does it apply to enough of your list? A brilliant signal that exists for 5% of your market is a segment, not a campaign.

If a data point passes all three tests, one sentence about it does more than a paragraph about your product.