Notes from the build/

The three receipts

I sent fifteen cold emails that carried evidence and got two replies, after thirty well-written ones that carried claims got zero. This is the framework that came out of that split, and why the most important step is deleting the findings that fail verification.

By

On 13 August I sent fifteen cold emails in one afternoon. Each one opened with a guest review the recipient's hotel group had left unanswered, named, dated, and quoted. Two people replied within a day, and one of them, a manager at a golf estate, wrote back a sentence I have been thinking about since: "Strangely, we have not seen the review you are referring to. You are welcome to send through the free report."

Two replies from fifteen cold emails is a good afternoon in any market. The interesting number sits underneath it though. In the three months before that afternoon I had sent roughly thirty cold emails of a different kind, offers, gifts, introductions, written as carefully as I know how, and the combined reply count across all thirty was zero. The fifteen that worked carried evidence the reader could check, and the thirty that failed carried claims the reader had to take on faith. Once I saw the split that cleanly, I rebuilt my whole outbound system around it, and this is the write-up of what came out.

2 to 0

Replies from fifteen cold emails that carried checkable evidence, against thirty carefully written ones that carried claims. Same sender, same market, same effort. The split is the whole framework.

Claims are free, and every reader knows it

I sell marketing and measurement services, which means I compete in the most flooded inbox category on earth. In Southern Africa, where much of my work is, the flood is worse: unemployment is high, everyone with a laptop offers digital marketing, and a business owner receives some version of "we help brands unlock growth" several times a week. The owner cannot tell a real operator from the flood by reading the copy, because copy is exactly the thing the flood is also producing. So everything gets deleted together, and honestly, that is the rational response.

The messages that survive this environment share one property, and it has nothing to do with writing quality. The first line contains a fact about the reader's own business that they can verify themselves, in one click, from a source they trust more than they trust you. The golf estate manager did not reply because my email was elegant. She replied because I had named a real review on her own listing, and when she went looking for it, the claim held.

The claim

  • "We are a results-driven agency helping businesses maximise ROI through data-driven strategies."
  • Unverifiable, and identical to the last thirty messages in the inbox.
  • Deleted on sight, rationally.

The receipt

  • "You are running Google ads today, and the page they land on has no conversion tag in its source."
  • Checkable in one click, from a source the reader trusts more than you.
  • Nobody else wrote it, because nobody else looked.

The framework

I now hold every cold message to a standard I call the three receipts. A message goes out only when it carries all three, and a guess on any one of them kills the send.

The first is the situation receipt: proof of the thing you noticed, in the prospect's own words. Their LinkedIn post announcing a new website. Their job ad for a sales rep. Their agency's portfolio page naming them as a client. You quote it verbatim and you keep the URL, because the quote is what earns the right to the rest of the message.

The second is the money receipt: proof that money is moving through the problem you solve. In my trade this one is almost unfair, because Google publishes it. The Ads Transparency Center is a public record of every ad an advertiser runs, with dates, and it answers in seconds whether a business is spending right now. "Last shown: today" turns a cold email from an opinion into a report. Other trades have their own registries: a salary in a job post, a sponsorship on an event page, a paid tool visible in a page's source.

The third is the gap receipt: proof of the defect, read from a public artifact rather than asserted. A website's source code shows whether anything measures the traffic the ads are buying. If the site loads Google Tag Manager, the container itself is public, and you can read whether a conversion tag is configured inside it. Zero hits on that check means the business is paying for clicks that nothing can count, and that single verified fact carries an entire message with no adjectives attached.

The part that matters most is the deleting

Here is the part I would skip if I were writing this to impress rather than to be useful. In one sweep my tooling flagged seventy websites as running ads with broken measurement. That is seventy potential messages, each with a finding that looks devastating. The temptation to send all seventy is enormous, because the finding is the whole pitch and the pitch was working.

I checked inside the tag containers before sending, and thirty of the seventy turned out to have conversion tags loading dynamically, invisible to a surface scan but there. Thirty findings deleted, thirty messages never sent.

30 of 70

Candidate findings deleted in a single sweep after checking inside the tag containers. Each one looked devastating in the pitch, and each one was wrong. The deleted findings are the framework.

It hurt to do and it was the most important step in the system, because the framework only works if the reader who checks your claim finds it true. One wrong claim and you are disproven in one click, by the exact person you asked to trust you, in a message whose entire argument was that you can be checked. The deleted findings are the framework.

It generalizes further than I expected

I built this for paid media work, and then kept noticing that the chain has nothing to do with paid media. If you sell anything that touches a company's online presence, their receipts are already public. A web designer can read a prospect's mobile speed score and find the dead links on a checkout flow. A reputation service can find the one-star review that has sat unanswered for four months, which is where my own first working batch came from. An SEO consultant can watch a site relaunch and then watch the old URLs start returning 404s, taking years of rankings with them. An IT security firm can read a company's DMARC record in public DNS while that company hires salespeople whose entire pipeline is outbound email. Every trade has a registry where the money shows and an artifact where the gap shows. The work is finding yours.

A worked example, and why speed became the offer

My favourite constructed signal is one I call the migration break. Companies post "our new website is live" on LinkedIn constantly, proud, dated, public. What almost nobody in the congratulations thread knows is that relaunches quietly break measurement: tags rarely survive the rebuild, and the old URLs often die without redirects. So the sequence writes itself.

Anatomy of a migration break
  • MONThe proud post: 'our new website is live'
  • TUEAds Transparency Center: ads still running, last shown today
  • TUEView source on the new site: no conversion tag survived the rebuild
  • 48HThe repair window, before remarketing decay and bidding starvation compound

They are buying visits the new site cannot count, days after celebrating it.

The email opens with congratulations, states exactly what was checked, and offers the full write-up free within 48 hours. The 48 hours stopped being a sales device once I understood the mechanics. Every day a tag is missing, the remarketing audience shrinks, and those people are gone for good. Google's automated bidding runs on conversion data, degrades without it, and re-learns slowly after the fix. The decay is technical, which means speed of repair genuinely changes the client's outcome, and saying that plainly is the most credible sentence in the message.

What I owe to other people's work

The insistence on the offer over the copy comes from Eric Nowoslawski, who publishes real numbers from a volume of cold email I will never send, and whose data keeps showing that a strong offer with no personalisation beats a weak offer with all of it. My sample sizes are tiny next to his, fifteen emails where he has millions, and I want that gap on the record. What I think I am adding is the receipts discipline: the rule that every line must survive the reader checking it, and the deleting that the rule demands.

The honest limits

This does not scale, and I have stopped wishing it would. The verification is manual at the exact point where it matters, my lists are dozens of companies where the volume players run hundreds of thousands, and a week of this work produces maybe forty sendable messages. But the flood cannot follow me here, because the flood exists precisely to avoid this kind of work, and the reader can feel the difference in one line. The golf estate manager asked me to send the report because every sentence in my email held when she checked it. That is the entire method. Send things that hold.

Get the next essay in your inbox.

Long-form on travel, AI, and the people the platforms were not built for. One email a week. No noise.

← Back to writing